kanshan/TensorRT-LLMs
1
0
Fork 0
mirror of https://github.com/NVIDIA/TensorRT-LLM.git synced 2026-01-14 06:27:45 +08:00
Code Issues Actions 1 Packages Projects Releases Wiki Activity
1d509f9205
TensorRT-LLMs/_modules/tensorrt_llm/models/mllama/model.html
Kaiyu Xie 1d509f9205 Update GitHub pages in root to v1.0.0rc6
2025-08-07 06:26:14 +00:00

2214 lines
262 KiB
HTML
Raw Blame History

<!DOCTYPE html>
<html lang="en" data-content_root="../../../../" >
<head>
<meta charset="utf-8" />
<meta name="viewport" content="width=device-width, initial-scale=1.0" />
<title>tensorrt_llm.models.mllama.model &#8212; TensorRT-LLM</title>
<script data-cfasync="false">
document.documentElement.dataset.mode = localStorage.getItem("mode") || "";
document.documentElement.dataset.theme = localStorage.getItem("theme") || "";
</script>
<!--
this give us a css class that will be invisible only if js is disabled
-->
<noscript>
<style>
.pst-js-only { display: none !important; }
</style>
</noscript>
<!-- Loaded before other Sphinx assets -->
<link href="../../../../_static/styles/theme.css?digest=8878045cc6db502f8baf" rel="stylesheet" />
<link href="../../../../_static/styles/pydata-sphinx-theme.css?digest=8878045cc6db502f8baf" rel="stylesheet" />
<link rel="stylesheet" type="text/css" href="../../../../_static/pygments.css?v=8f2a1f02" />
<link rel="stylesheet" type="text/css" href="../../../../_static/styles/nvidia-sphinx-theme.css?v=df3ac72c" />
<link rel="stylesheet" type="text/css" href="../../../../_static/copybutton.css?v=76b2166b" />
<link rel="stylesheet" type="text/css" href="../../../../_static/autodoc_pydantic.css" />
<link rel="stylesheet" type="text/css" href="../../../../_static/togglebutton.css?v=13237357" />
<link rel="stylesheet" type="text/css" href="../../../../_static/custom.css?v=95073da6" />
<!-- So that users can add custom icons -->
<script src="../../../../_static/scripts/fontawesome.js?digest=8878045cc6db502f8baf"></script>
<!-- Pre-loaded scripts that we'll load fully later -->
<link rel="preload" as="script" href="../../../../_static/scripts/bootstrap.js?digest=8878045cc6db502f8baf" />
<link rel="preload" as="script" href="../../../../_static/scripts/pydata-sphinx-theme.js?digest=8878045cc6db502f8baf" />
<script src="../../../../_static/documentation_options.js?v=5929fcd5"></script>
<script src="../../../../_static/doctools.js?v=9a2dae69"></script>
<script src="../../../../_static/sphinx_highlight.js?v=dc90522c"></script>
<script src="../../../../_static/clipboard.min.js?v=a7894cd8"></script>
<script src="../../../../_static/copybutton.js?v=65e89d2a"></script>
<script>let toggleHintShow = 'Click to show';</script>
<script>let toggleHintHide = 'Click to hide';</script>
<script>let toggleOpenOnPrint = 'true';</script>
<script src="../../../../_static/togglebutton.js?v=4a39c7ea"></script>
<script>var togglebuttonSelector = '.toggle, .admonition.dropdown';</script>
<script>var togglebuttonSelector = '.toggle, .admonition.dropdown';</script>
<script>DOCUMENTATION_OPTIONS.pagename = '_modules/tensorrt_llm/models/mllama/model';</script>
<script>
DOCUMENTATION_OPTIONS.theme_version = '0.16.1';
DOCUMENTATION_OPTIONS.theme_switcher_json_url = './_static/switcher.json';
DOCUMENTATION_OPTIONS.theme_switcher_version_match = '1.0.0rc6';
DOCUMENTATION_OPTIONS.show_version_warning_banner =
false;
</script>
<link rel="icon" href="../../../../_static/favicon.png"/>
<link rel="index" title="Index" href="../../../../genindex.html" />
<link rel="search" title="Search" href="../../../../search.html" />
<meta name="viewport" content="width=device-width, initial-scale=1"/>
<meta name="docsearch:language" content="en"/>
<meta name="docsearch:version" content="1.0.0rc6" />
</head>
<body data-bs-spy="scroll" data-bs-target=".bd-toc-nav" data-offset="180" data-bs-root-margin="0px 0px -60%" data-default-mode="">
<div id="pst-skip-link" class="skip-link d-print-none"><a href="#main-content">Skip to main content</a></div>
<div id="pst-scroll-pixel-helper"></div>
<button type="button" class="btn rounded-pill" id="pst-back-to-top">
<i class="fa-solid fa-arrow-up"></i>Back to top</button>
<dialog id="pst-search-dialog">
<form class="bd-search d-flex align-items-center"
action="../../../../search.html"
method="get">
<i class="fa-solid fa-magnifying-glass"></i>
<input type="search"
class="form-control"
name="q"
placeholder="Search the docs ..."
aria-label="Search the docs ..."
autocomplete="off"
autocorrect="off"
autocapitalize="off"
spellcheck="false"/>
<span class="search-button__kbd-shortcut"><kbd class="kbd-shortcut__modifier">Ctrl</kbd>+<kbd>K</kbd></span>
</form>
</dialog>
<div class="pst-async-banner-revealer d-none">
<aside id="bd-header-version-warning" class="d-none d-print-none" aria-label="Version warning"></aside>
</div>
<header class="bd-header navbar navbar-expand-lg bd-navbar d-print-none">
<div class="bd-header__inner bd-page-width">
<button class="pst-navbar-icon sidebar-toggle primary-toggle" aria-label="Site navigation">
<span class="fa-solid fa-bars"></span>
</button>
<div class="col-lg-3 navbar-header-items__start">
<div class="navbar-item">
<a class="navbar-brand logo" href="../../../../index.html">
<img src="../../../../_static/nvidia-logo-horiz-rgb-blk-for-screen.svg" class="logo__image only-light" alt="TensorRT-LLM - Home"/>
<img src="../../../../_static/nvidia-logo-horiz-rgb-wht-for-screen.svg" class="logo__image only-dark pst-js-only" alt="TensorRT-LLM - Home"/>
<p class="title logo__title">TensorRT-LLM</p>
</a></div>
</div>
<div class="col-lg-9 navbar-header-items">
<div class="me-auto navbar-header-items__center">
<div class="navbar-item">
<div class="version-switcher__container dropdown pst-js-only">
<button id="pst-version-switcher-button-2"
type="button"
class="version-switcher__button btn btn-sm dropdown-toggle"
data-bs-toggle="dropdown"
aria-haspopup="listbox"
aria-controls="pst-version-switcher-list-2"
aria-label="Version switcher list"
>
Choose version <!-- this text may get changed later by javascript -->
<span class="caret"></span>
</button>
<div id="pst-version-switcher-list-2"
class="version-switcher__menu dropdown-menu list-group-flush py-0"
role="listbox" aria-labelledby="pst-version-switcher-button-2">
<!-- dropdown will be populated by javascript on page load -->
</div>
</div></div>
</div>
<div class="navbar-header-items__end">
<div class="navbar-item navbar-persistent--container">
<button class="btn search-button-field search-button__button pst-js-only" title="Search" aria-label="Search" data-bs-placement="bottom" data-bs-toggle="tooltip">
<i class="fa-solid fa-magnifying-glass"></i>
<span class="search-button__default-text">Search</span>
<span class="search-button__kbd-shortcut"><kbd class="kbd-shortcut__modifier">Ctrl</kbd>+<kbd class="kbd-shortcut__modifier">K</kbd></span>
</button>
</div>
<div class="navbar-item">
<button class="btn btn-sm nav-link pst-navbar-icon theme-switch-button pst-js-only" aria-label="Color mode" data-bs-title="Color mode" data-bs-placement="bottom" data-bs-toggle="tooltip">
<i class="theme-switch fa-solid fa-sun fa-lg" data-mode="light" title="Light"></i>
<i class="theme-switch fa-solid fa-moon fa-lg" data-mode="dark" title="Dark"></i>
<i class="theme-switch fa-solid fa-circle-half-stroke fa-lg" data-mode="auto" title="System Settings"></i>
</button></div>
</div>
</div>
<div class="navbar-persistent--mobile">
<button class="btn search-button-field search-button__button pst-js-only" title="Search" aria-label="Search" data-bs-placement="bottom" data-bs-toggle="tooltip">
<i class="fa-solid fa-magnifying-glass"></i>
<span class="search-button__default-text">Search</span>
<span class="search-button__kbd-shortcut"><kbd class="kbd-shortcut__modifier">Ctrl</kbd>+<kbd class="kbd-shortcut__modifier">K</kbd></span>
</button>
</div>
</div>
</header>
<div class="bd-container">
<div class="bd-container__inner bd-page-width">
<dialog id="pst-primary-sidebar-modal"></dialog>
<div id="pst-primary-sidebar" class="bd-sidebar-primary bd-sidebar">
<a class="navbar-brand logo" href="../../../../index.html">
<img src="../../../../_static/nvidia-logo-horiz-rgb-blk-for-screen.svg" class="logo__image only-light" alt="TensorRT-LLM - Home"/>
<img src="../../../../_static/nvidia-logo-horiz-rgb-wht-for-screen.svg" class="logo__image only-dark pst-js-only" alt="TensorRT-LLM - Home"/>
<p class="title logo__title">TensorRT-LLM</p>
</a>
<div class="sidebar-header-items sidebar-primary__section">
<div class="sidebar-header-items__center">
<div class="navbar-item">
<div class="version-switcher__container dropdown pst-js-only">
<button id="pst-version-switcher-button-3"
type="button"
class="version-switcher__button btn btn-sm dropdown-toggle"
data-bs-toggle="dropdown"
aria-haspopup="listbox"
aria-controls="pst-version-switcher-list-3"
aria-label="Version switcher list"
>
Choose version <!-- this text may get changed later by javascript -->
<span class="caret"></span>
</button>
<div id="pst-version-switcher-list-3"
class="version-switcher__menu dropdown-menu list-group-flush py-0"
role="listbox" aria-labelledby="pst-version-switcher-button-3">
<!-- dropdown will be populated by javascript on page load -->
</div>
</div></div>
</div>
<div class="sidebar-header-items__end">
<div class="navbar-item">
<button class="btn btn-sm nav-link pst-navbar-icon theme-switch-button pst-js-only" aria-label="Color mode" data-bs-title="Color mode" data-bs-placement="bottom" data-bs-toggle="tooltip">
<i class="theme-switch fa-solid fa-sun fa-lg" data-mode="light" title="Light"></i>
<i class="theme-switch fa-solid fa-moon fa-lg" data-mode="dark" title="Dark"></i>
<i class="theme-switch fa-solid fa-circle-half-stroke fa-lg" data-mode="auto" title="System Settings"></i>
</button></div>
</div>
</div>
<div class="sidebar-primary-items__start sidebar-primary__section">
<div class="sidebar-primary-item">
<nav class="bd-docs-nav bd-links"
aria-label="Table of Contents">
<p class="bd-links__title" role="heading" aria-level="1">Table of Contents</p>
<div class="bd-toc-item navbar-nav"><p aria-level="2" class="caption" role="heading"><span class="caption-text">Getting Started</span></p>
<ul class="nav bd-sidenav">
<li class="toctree-l1"><a class="reference internal" href="../../../../overview.html">Overview</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../quick-start-guide.html">Quick Start Guide</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../key-features.html">Key Features</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../torch.html">PyTorch Backend</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../release-notes.html">Release Notes</a></li>
</ul>
<p aria-level="2" class="caption" role="heading"><span class="caption-text">Installation</span></p>
<ul class="nav bd-sidenav">
<li class="toctree-l1"><a class="reference internal" href="../../../../installation/containers.html">Pre-built release container images on NGC</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../installation/linux.html">Installing on Linux via <code class="docutils literal notranslate"><span class="pre">pip</span></code></a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../installation/build-from-source-linux.html">Building from Source Code on Linux</a></li>
</ul>
<p aria-level="2" class="caption" role="heading"><span class="caption-text">LLM API</span></p>
<ul class="nav bd-sidenav">
<li class="toctree-l1"><a class="reference internal" href="../../../../llm-api/index.html">LLM API Introduction</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../llm-api/reference.html">API Reference</a></li>
</ul>
<p aria-level="2" class="caption" role="heading"><span class="caption-text">Examples</span></p>
<ul class="nav bd-sidenav">
<li class="toctree-l1 has-children"><a class="reference internal" href="../../../../examples/index.html">LLM Examples Introduction</a><details><summary><span class="toctree-toggle" role="presentation"><i class="fa-solid fa-chevron-down"></i></span></summary><ul class="simple">
</ul>
</details></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../examples/customization.html">LLM Common Customizations</a></li>
<li class="toctree-l1 has-children"><a class="reference internal" href="../../../../examples/llm_api_examples.html">LLM Examples</a><details><summary><span class="toctree-toggle" role="presentation"><i class="fa-solid fa-chevron-down"></i></span></summary><ul>
<li class="toctree-l2"><a class="reference internal" href="../../../../examples/llm_inference.html">Generate text</a></li>
<li class="toctree-l2"><a class="reference internal" href="../../../../examples/llm_inference_async.html">Generate text asynchronously</a></li>
<li class="toctree-l2"><a class="reference internal" href="../../../../examples/llm_inference_async_streaming.html">Generate text in streaming</a></li>
<li class="toctree-l2"><a class="reference internal" href="../../../../examples/llm_inference_distributed.html">Distributed LLM Generation</a></li>
<li class="toctree-l2"><a class="reference internal" href="../../../../examples/llm_guided_decoding.html">Generate text with guided decoding</a></li>
<li class="toctree-l2"><a class="reference internal" href="../../../../examples/llm_logits_processor.html">Control generated text using logits processor</a></li>
<li class="toctree-l2"><a class="reference internal" href="../../../../examples/llm_multilora.html">Generate text with multiple LoRA adapters</a></li>
<li class="toctree-l2"><a class="reference internal" href="../../../../examples/llm_speculative_decoding.html">Speculative Decoding</a></li>
<li class="toctree-l2"><a class="reference internal" href="../../../../examples/llm_runtime.html">Runtime Configuration Examples</a></li>
<li class="toctree-l2"><a class="reference internal" href="../../../../examples/llm_sampling.html">Sampling Techniques Showcase</a></li>
<li class="toctree-l2"><a class="reference internal" href="../../../../examples/llm_mgmn_llm_distributed.html">Run LLM-API with pytorch backend on Slurm</a></li>
<li class="toctree-l2"><a class="reference internal" href="../../../../examples/llm_mgmn_trtllm_bench.html">Run trtllm-bench with pytorch backend on Slurm</a></li>
<li class="toctree-l2"><a class="reference internal" href="../../../../examples/llm_mgmn_trtllm_serve.html">Run trtllm-serve with pytorch backend on Slurm</a></li>
</ul>
</details></li>
<li class="toctree-l1 has-children"><a class="reference internal" href="../../../../examples/trtllm_serve_examples.html">Online Serving Examples</a><details><summary><span class="toctree-toggle" role="presentation"><i class="fa-solid fa-chevron-down"></i></span></summary><ul>
<li class="toctree-l2"><a class="reference internal" href="../../../../examples/curl_chat_client.html">Curl Chat Client</a></li>
<li class="toctree-l2"><a class="reference internal" href="../../../../examples/curl_chat_client_for_multimodal.html">Curl Chat Client For Multimodal</a></li>
<li class="toctree-l2"><a class="reference internal" href="../../../../examples/curl_completion_client.html">Curl Completion Client</a></li>
<li class="toctree-l2"><a class="reference internal" href="../../../../examples/deepseek_r1_reasoning_parser.html">Deepseek R1 Reasoning Parser</a></li>
<li class="toctree-l2"><a class="reference internal" href="../../../../examples/genai_perf_client.html">Genai Perf Client</a></li>
<li class="toctree-l2"><a class="reference internal" href="../../../../examples/genai_perf_client_for_multimodal.html">Genai Perf Client For Multimodal</a></li>
<li class="toctree-l2"><a class="reference internal" href="../../../../examples/openai_chat_client.html">OpenAI Chat Client</a></li>
<li class="toctree-l2"><a class="reference internal" href="../../../../examples/openai_chat_client_for_multimodal.html">OpenAI Chat Client for Multimodal</a></li>
<li class="toctree-l2"><a class="reference internal" href="../../../../examples/openai_completion_client.html">OpenAI Completion Client</a></li>
<li class="toctree-l2"><a class="reference internal" href="../../../../examples/openai_completion_client_for_lora.html">Openai Completion Client For Lora</a></li>
<li class="toctree-l2"><a class="reference internal" href="../../../../examples/openai_completion_client_json_schema.html">OpenAI Completion Client with JSON Schema</a></li>
</ul>
</details></li>
</ul>
<p aria-level="2" class="caption" role="heading"><span class="caption-text">Model Definition API</span></p>
<ul class="nav bd-sidenav">
<li class="toctree-l1"><a class="reference internal" href="../../../../python-api/tensorrt_llm.layers.html">Layers</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../python-api/tensorrt_llm.functional.html">Functionals</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../python-api/tensorrt_llm.models.html">Models</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../python-api/tensorrt_llm.plugin.html">Plugin</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../python-api/tensorrt_llm.quantization.html">Quantization</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../python-api/tensorrt_llm.runtime.html">Runtime</a></li>
</ul>
<p aria-level="2" class="caption" role="heading"><span class="caption-text">C++ API</span></p>
<ul class="nav bd-sidenav">
<li class="toctree-l1"><a class="reference internal" href="../../../../_cpp_gen/executor.html">Executor</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../_cpp_gen/runtime.html">Runtime</a></li>
</ul>
<p aria-level="2" class="caption" role="heading"><span class="caption-text">Command-Line Reference</span></p>
<ul class="nav bd-sidenav">
<li class="toctree-l1"><a class="reference internal" href="../../../../commands/trtllm-bench.html">trtllm-bench</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../commands/trtllm-build.html">trtllm-build</a></li>
<li class="toctree-l1 has-children"><a class="reference internal" href="../../../../commands/trtllm-serve/index.html">trtllm-serve</a><details><summary><span class="toctree-toggle" role="presentation"><i class="fa-solid fa-chevron-down"></i></span></summary><ul>
<li class="toctree-l2"><a class="reference internal" href="../../../../commands/trtllm-serve/trtllm-serve.html">trtllm-serve</a></li>
<li class="toctree-l2"><a class="reference internal" href="../../../../commands/trtllm-serve/run-benchmark-with-trtllm-serve.html">Run benchmarking with <code class="docutils literal notranslate"><span class="pre">trtllm-serve</span></code></a></li>
</ul>
</details></li>
</ul>
<p aria-level="2" class="caption" role="heading"><span class="caption-text">Architecture</span></p>
<ul class="nav bd-sidenav">
<li class="toctree-l1"><a class="reference internal" href="../../../../architecture/overview.html">TensorRT-LLM Architecture</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../architecture/core-concepts.html">Model Definition</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../architecture/checkpoint.html">TensorRT-LLM Checkpoint</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../architecture/workflow.html">TensorRT-LLM Build Workflow</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../architecture/add-model.html">Adding a Model</a></li>
</ul>
<p aria-level="2" class="caption" role="heading"><span class="caption-text">Advanced</span></p>
<ul class="nav bd-sidenav">
<li class="toctree-l1"><a class="reference internal" href="../../../../advanced/gpt-attention.html">Multi-Head, Multi-Query, and Group-Query Attention</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../advanced/gpt-runtime.html">C++ GPT Runtime</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../advanced/executor.html">Executor API</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../advanced/graph-rewriting.html">Graph Rewriting Module</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../advanced/lora.html">Run gpt-2b + LoRA using Executor / cpp runtime</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../advanced/expert-parallelism.html">Expert Parallelism in TensorRT-LLM</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../advanced/kv-cache-management.html">KV Cache Management: Pools, Blocks, and Events</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../advanced/kv-cache-reuse.html">KV cache reuse</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../advanced/speculative-decoding.html">Speculative Sampling</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../advanced/disaggregated-service.html">Disaggregated-Service (Experimental)</a></li>
</ul>
<p aria-level="2" class="caption" role="heading"><span class="caption-text">Performance</span></p>
<ul class="nav bd-sidenav">
<li class="toctree-l1"><a class="reference internal" href="../../../../performance/perf-overview.html">Overview</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../performance/perf-benchmarking.html">Benchmarking</a></li>
<li class="toctree-l1 has-children"><a class="reference internal" href="../../../../performance/performance-tuning-guide/index.html">Performance Tuning Guide</a><details><summary><span class="toctree-toggle" role="presentation"><i class="fa-solid fa-chevron-down"></i></span></summary><ul>
<li class="toctree-l2"><a class="reference internal" href="../../../../performance/performance-tuning-guide/benchmarking-default-performance.html">Benchmarking Default Performance</a></li>
<li class="toctree-l2"><a class="reference internal" href="../../../../performance/performance-tuning-guide/useful-build-time-flags.html">Useful Build-Time Flags</a></li>
<li class="toctree-l2"><a class="reference internal" href="../../../../performance/performance-tuning-guide/tuning-max-batch-size-and-max-num-tokens.html">Tuning Max Batch Size and Max Num Tokens</a></li>
<li class="toctree-l2"><a class="reference internal" href="../../../../performance/performance-tuning-guide/deciding-model-sharding-strategy.html">Deciding Model Sharding Strategy</a></li>
<li class="toctree-l2"><a class="reference internal" href="../../../../performance/performance-tuning-guide/fp8-quantization.html">FP8 Quantization</a></li>
<li class="toctree-l2"><a class="reference internal" href="../../../../performance/performance-tuning-guide/useful-runtime-flags.html">Useful Runtime Options</a></li>
</ul>
</details></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../performance/perf-analysis.html">Performance Analysis</a></li>
</ul>
<p aria-level="2" class="caption" role="heading"><span class="caption-text">Reference</span></p>
<ul class="nav bd-sidenav">
<li class="toctree-l1"><a class="reference internal" href="../../../../reference/troubleshooting.html">Troubleshooting</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../reference/support-matrix.html">Support Matrix</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../reference/precision.html">Numerical Precision</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../reference/memory.html">Memory Usage of TensorRT-LLM</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../reference/ci-overview.html">Continuous Integration Overview</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../reference/dev-containers.html">Using Dev Containers</a></li>
</ul>
<p aria-level="2" class="caption" role="heading"><span class="caption-text">Blogs</span></p>
<ul class="nav bd-sidenav">
<li class="toctree-l1"><a class="reference internal" href="../../../../blogs/H100vsA100.html">H100 has 4.6x A100 Performance in TensorRT-LLM, achieving 10,000 tok/s at 100ms to first token</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../blogs/H200launch.html">H200 achieves nearly 12,000 tokens/sec on Llama2-13B with TensorRT-LLM</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../blogs/Falcon180B-H200.html">Falcon-180B on a single H200 GPU with INT4 AWQ, and 6.7x faster Llama-70B over A100</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../blogs/quantization-in-TRT-LLM.html">Speed up inference with SOTA quantization techniques in TRT-LLM</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../blogs/XQA-kernel.html">New XQA-kernel provides 2.4x more Llama-70B throughput within the same latency budget</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../blogs/tech_blog/blog1_Pushing_Latency_Boundaries_Optimizing_DeepSeek-R1_Performance_on_NVIDIA_B200_GPUs.html">Pushing Latency Boundaries: Optimizing DeepSeek-R1 Performance on NVIDIA B200 GPUs</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../blogs/tech_blog/blog2_DeepSeek_R1_MTP_Implementation_and_Optimization.html">DeepSeek R1 MTP Implementation and Optimization</a></li>
</ul>
</div>
</nav></div>
</div>
<div class="sidebar-primary-items__end sidebar-primary__section">
</div>
</div>
<main id="main-content" class="bd-main" role="main">
<div class="bd-content">
<div class="bd-article-container">
<div class="bd-header-article d-print-none">
<div class="header-article-items header-article__inner">
<div class="header-article-items__start">
<div class="header-article-item">
<nav aria-label="Breadcrumb" class="d-print-none">
<ul class="bd-breadcrumbs">
<li class="breadcrumb-item breadcrumb-home">
<a href="../../../../index.html" class="nav-link" aria-label="Home">
<i class="fa-solid fa-home"></i>
</a>
</li>
<li class="breadcrumb-item"><a href="../../../index.html" class="nav-link">Module code</a></li>
<li class="breadcrumb-item active" aria-current="page"><span class="ellipsis">tensorrt_llm.models.mllama.model</span></li>
</ul>
</nav>
</div>
</div>
</div>
</div>
<div id="searchbox"></div>
<article class="bd-article">
<h1>Source code for tensorrt_llm.models.mllama.model</h1><div class="highlight"><pre>
<span></span><span class="c1"># SPDX-FileCopyrightText: Copyright (c) 2022-2024 NVIDIA CORPORATION &amp; AFFILIATES. All rights reserved.</span>
<span class="c1"># SPDX-License-Identifier: Apache-2.0</span>
<span class="c1">#</span>
<span class="c1"># Licensed under the Apache License, Version 2.0 (the &quot;License&quot;);</span>
<span class="c1"># you may not use this file except in compliance with the License.</span>
<span class="c1"># You may obtain a copy of the License at</span>
<span class="c1">#</span>
<span class="c1"># http://www.apache.org/licenses/LICENSE-2.0</span>
<span class="c1">#</span>
<span class="c1"># Unless required by applicable law or agreed to in writing, software</span>
<span class="c1"># distributed under the License is distributed on an &quot;AS IS&quot; BASIS,</span>
<span class="c1"># WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.</span>
<span class="c1"># See the License for the specific language governing permissions and</span>
<span class="c1"># limitations under the License.</span>
<span class="kn">import</span><span class="w"> </span><span class="nn">math</span>
<span class="kn">from</span><span class="w"> </span><span class="nn">collections</span><span class="w"> </span><span class="kn">import</span> <span class="n">OrderedDict</span>
<span class="kn">from</span><span class="w"> </span><span class="nn">typing</span><span class="w"> </span><span class="kn">import</span> <span class="n">List</span><span class="p">,</span> <span class="n">Optional</span><span class="p">,</span> <span class="n">Union</span>
<span class="kn">import</span><span class="w"> </span><span class="nn">tensorrt</span><span class="w"> </span><span class="k">as</span><span class="w"> </span><span class="nn">trt</span>
<span class="kn">import</span><span class="w"> </span><span class="nn">torch</span>
<span class="kn">from</span><span class="w"> </span><span class="nn">tensorrt_llm._common</span><span class="w"> </span><span class="kn">import</span> <span class="n">default_net</span>
<span class="kn">from</span><span class="w"> </span><span class="nn">tensorrt_llm._utils</span><span class="w"> </span><span class="kn">import</span> <span class="n">numpy_to_torch</span><span class="p">,</span> <span class="n">str_dtype_to_torch</span>
<span class="kn">from</span><span class="w"> </span><span class="nn">tensorrt_llm.bindings</span><span class="w"> </span><span class="kn">import</span> <span class="n">KVCacheType</span>
<span class="kn">from</span><span class="w"> </span><span class="nn">tensorrt_llm.functional</span><span class="w"> </span><span class="kn">import</span> <span class="p">(</span><span class="n">Conditional</span><span class="p">,</span> <span class="n">LayerNormPositionType</span><span class="p">,</span>
<span class="n">LayerNormType</span><span class="p">,</span> <span class="n">MLPType</span><span class="p">,</span>
<span class="n">PositionEmbeddingType</span><span class="p">,</span> <span class="n">Tensor</span><span class="p">,</span> <span class="n">assertion</span><span class="p">,</span>
<span class="n">gather_last_token_logits</span><span class="p">,</span> <span class="n">maximum</span><span class="p">,</span> <span class="n">minimum</span><span class="p">,</span>
<span class="n">recv</span><span class="p">,</span> <span class="n">reduce</span><span class="p">,</span> <span class="n">send</span><span class="p">,</span> <span class="n">shape</span><span class="p">,</span> <span class="n">tanh</span><span class="p">)</span>
<span class="kn">from</span><span class="w"> </span><span class="nn">tensorrt_llm.layers</span><span class="w"> </span><span class="kn">import</span> <span class="p">(</span><span class="n">MLP</span><span class="p">,</span> <span class="n">Attention</span><span class="p">,</span> <span class="n">AttentionMaskParams</span><span class="p">,</span>
<span class="n">AttentionMaskType</span><span class="p">,</span> <span class="n">AttentionParams</span><span class="p">,</span>
<span class="n">ColumnLinear</span><span class="p">,</span> <span class="n">Embedding</span><span class="p">,</span> <span class="n">FusedGatedMLP</span><span class="p">,</span>
<span class="n">GatedMLP</span><span class="p">,</span> <span class="n">GroupNorm</span><span class="p">,</span> <span class="n">KeyValueCacheParams</span><span class="p">,</span>
<span class="n">LayerNorm</span><span class="p">,</span> <span class="n">LoraParams</span><span class="p">,</span> <span class="n">RmsNorm</span><span class="p">)</span>
<span class="kn">from</span><span class="w"> </span><span class="nn">tensorrt_llm.lora_manager</span><span class="w"> </span><span class="kn">import</span> <span class="p">(</span><span class="n">LoraConfig</span><span class="p">,</span>
<span class="n">get_default_trtllm_modules_to_hf_modules</span><span class="p">,</span>
<span class="n">use_lora</span><span class="p">)</span>
<span class="kn">from</span><span class="w"> </span><span class="nn">tensorrt_llm.mapping</span><span class="w"> </span><span class="kn">import</span> <span class="n">Mapping</span>
<span class="kn">from</span><span class="w"> </span><span class="nn">tensorrt_llm.models.model_weights_loader</span><span class="w"> </span><span class="kn">import</span> <span class="n">ModelWeightsLoader</span>
<span class="kn">from</span><span class="w"> </span><span class="nn">tensorrt_llm.models.modeling_utils</span><span class="w"> </span><span class="kn">import</span> <span class="n">PretrainedModel</span><span class="p">,</span> <span class="n">QuantConfig</span>
<span class="kn">from</span><span class="w"> </span><span class="nn">tensorrt_llm.module</span><span class="w"> </span><span class="kn">import</span> <span class="n">Module</span><span class="p">,</span> <span class="n">ModuleList</span>
<span class="kn">from</span><span class="w"> </span><span class="nn">tensorrt_llm.parameter</span><span class="w"> </span><span class="kn">import</span> <span class="n">Parameter</span>
<span class="kn">from</span><span class="w"> </span><span class="nn">tensorrt_llm.quantization</span><span class="w"> </span><span class="kn">import</span> <span class="n">QuantMode</span>
<span class="kn">from</span><span class="w"> </span><span class="nn">.config</span><span class="w"> </span><span class="kn">import</span> <span class="n">MLLaMAConfig</span>
<span class="n">layernorm_map</span> <span class="o">=</span> <span class="p">{</span>
<span class="n">LayerNormType</span><span class="o">.</span><span class="n">LayerNorm</span><span class="p">:</span> <span class="n">LayerNorm</span><span class="p">,</span>
<span class="n">LayerNormType</span><span class="o">.</span><span class="n">RmsNorm</span><span class="p">:</span> <span class="n">RmsNorm</span><span class="p">,</span>
<span class="n">LayerNormType</span><span class="o">.</span><span class="n">GroupNorm</span><span class="p">:</span> <span class="n">GroupNorm</span><span class="p">,</span>
<span class="p">}</span>
<span class="n">mlp_map</span> <span class="o">=</span> <span class="p">{</span>
<span class="n">MLPType</span><span class="o">.</span><span class="n">MLP</span><span class="p">:</span> <span class="n">MLP</span><span class="p">,</span>
<span class="n">MLPType</span><span class="o">.</span><span class="n">GatedMLP</span><span class="p">:</span> <span class="n">GatedMLP</span><span class="p">,</span>
<span class="n">MLPType</span><span class="o">.</span><span class="n">FusedGatedMLP</span><span class="p">:</span> <span class="n">FusedGatedMLP</span><span class="p">,</span>
<span class="p">}</span>
<span class="n">ADD_DEBUG_TENSOR</span> <span class="o">=</span> <span class="kc">False</span>
<span class="k">class</span><span class="w"> </span><span class="nc">CrossAttentionTransformerBlock</span><span class="p">(</span><span class="n">Module</span><span class="p">):</span>
<span class="k">def</span><span class="w"> </span><span class="fm">__init__</span><span class="p">(</span>
<span class="bp">self</span><span class="p">,</span>
<span class="o">*</span><span class="p">,</span>
<span class="n">local_layer_idx</span><span class="p">,</span>
<span class="n">hidden_size</span><span class="p">,</span>
<span class="n">ffn_hidden_size</span><span class="p">,</span>
<span class="n">num_attention_heads</span><span class="p">,</span>
<span class="n">num_kv_heads</span><span class="p">,</span>
<span class="n">head_size</span><span class="p">,</span>
<span class="n">max_position_embeddings</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span>
<span class="n">q_scaling</span><span class="o">=</span><span class="mf">1.0</span><span class="p">,</span>
<span class="n">has_attention_qkvo_bias</span><span class="o">=</span><span class="kc">False</span><span class="p">,</span>
<span class="n">has_mlp_bias</span><span class="o">=</span><span class="kc">False</span><span class="p">,</span>
<span class="n">layernorm_position</span><span class="o">=</span><span class="n">LayerNormPositionType</span><span class="o">.</span><span class="n">pre_layernorm</span><span class="p">,</span>
<span class="n">layernorm_type</span><span class="o">=</span><span class="n">LayerNormType</span><span class="o">.</span><span class="n">RmsNorm</span><span class="p">,</span>
<span class="n">layernorm_eps</span><span class="o">=</span><span class="mf">1e-5</span><span class="p">,</span>
<span class="n">hidden_act</span><span class="o">=</span><span class="s2">&quot;gated-silu&quot;</span><span class="p">,</span>
<span class="n">mlp_type</span><span class="o">=</span><span class="n">MLPType</span><span class="o">.</span><span class="n">GatedMLP</span><span class="p">,</span>
<span class="n">mapping</span><span class="o">=</span><span class="n">Mapping</span><span class="p">(),</span>
<span class="n">dtype</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span>
<span class="n">residual_scaling</span><span class="o">=</span><span class="mf">1.0</span><span class="p">,</span>
<span class="n">relative_attention</span><span class="o">=</span><span class="kc">False</span><span class="p">,</span>
<span class="n">max_distance</span><span class="o">=</span><span class="mi">0</span><span class="p">,</span>
<span class="n">num_buckets</span><span class="o">=</span><span class="mi">0</span><span class="p">,</span>
<span class="n">fp16_clamping</span><span class="o">=</span><span class="kc">False</span><span class="p">,</span>
<span class="n">skip_cross_kv</span><span class="o">=</span><span class="kc">False</span><span class="p">,</span>
<span class="n">use_implicit_relative_attention</span><span class="o">=</span><span class="kc">False</span><span class="p">,</span>
<span class="n">rotary_embedding_base</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span>
<span class="n">rotary_embedding_scaling</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span>
<span class="n">quant_mode</span><span class="o">=</span><span class="n">QuantMode</span><span class="p">(</span><span class="mi">0</span><span class="p">),</span>
<span class="p">):</span>
<span class="nb">super</span><span class="p">()</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span>
<span class="bp">self</span><span class="o">.</span><span class="n">local_layer_idx</span> <span class="o">=</span> <span class="n">local_layer_idx</span>
<span class="bp">self</span><span class="o">.</span><span class="n">layernorm_type</span> <span class="o">=</span> <span class="n">layernorm_type</span>
<span class="n">ln_type</span> <span class="o">=</span> <span class="n">layernorm_map</span><span class="p">[</span><span class="n">layernorm_type</span><span class="p">]</span>
<span class="bp">self</span><span class="o">.</span><span class="n">layernorm_position</span> <span class="o">=</span> <span class="n">layernorm_position</span>
<span class="k">assert</span> <span class="bp">self</span><span class="o">.</span><span class="n">layernorm_position</span> <span class="o">==</span> <span class="n">LayerNormPositionType</span><span class="o">.</span><span class="n">pre_layernorm</span>
<span class="bp">self</span><span class="o">.</span><span class="n">cross_attention</span> <span class="o">=</span> <span class="n">Attention</span><span class="p">(</span>
<span class="n">local_layer_idx</span><span class="o">=</span><span class="n">local_layer_idx</span><span class="p">,</span>
<span class="n">hidden_size</span><span class="o">=</span><span class="n">hidden_size</span><span class="p">,</span>
<span class="n">num_attention_heads</span><span class="o">=</span><span class="n">num_attention_heads</span><span class="p">,</span>
<span class="n">attention_head_size</span><span class="o">=</span><span class="n">head_size</span><span class="p">,</span>
<span class="n">num_kv_heads</span><span class="o">=</span><span class="n">num_kv_heads</span><span class="p">,</span>
<span class="n">max_position_embeddings</span><span class="o">=</span><span class="n">max_position_embeddings</span><span class="p">,</span>
<span class="n">q_scaling</span><span class="o">=</span><span class="n">q_scaling</span><span class="p">,</span>
<span class="n">bias</span><span class="o">=</span><span class="n">has_attention_qkvo_bias</span><span class="p">,</span>
<span class="n">attention_mask_type</span><span class="o">=</span><span class="n">AttentionMaskType</span><span class="o">.</span><span class="n">causal</span><span class="p">,</span>
<span class="n">tp_group</span><span class="o">=</span><span class="n">mapping</span><span class="o">.</span><span class="n">tp_group</span><span class="p">,</span>
<span class="n">tp_size</span><span class="o">=</span><span class="n">mapping</span><span class="o">.</span><span class="n">tp_size</span><span class="p">,</span>
<span class="n">tp_rank</span><span class="o">=</span><span class="n">mapping</span><span class="o">.</span><span class="n">tp_rank</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="n">dtype</span><span class="p">,</span>
<span class="n">cross_attention</span><span class="o">=</span><span class="kc">True</span><span class="p">,</span>
<span class="n">relative_attention</span><span class="o">=</span>
<span class="kc">False</span><span class="p">,</span> <span class="c1"># Cross attention has no relative attention bias</span>
<span class="n">max_distance</span><span class="o">=</span><span class="n">max_distance</span><span class="p">,</span>
<span class="n">num_buckets</span><span class="o">=</span><span class="n">num_buckets</span><span class="p">,</span>
<span class="n">position_embedding_type</span><span class="o">=</span><span class="n">PositionEmbeddingType</span><span class="o">.</span>
<span class="n">learned_absolute</span><span class="p">,</span> <span class="c1"># we don&#39;t use rope for cross attn</span>
<span class="n">skip_cross_kv</span><span class="o">=</span><span class="n">skip_cross_kv</span><span class="p">,</span>
<span class="n">qk_layernorm</span><span class="o">=</span><span class="kc">True</span><span class="p">,</span>
<span class="n">layernorm_type</span><span class="o">=</span><span class="n">layernorm_type</span><span class="p">,</span>
<span class="n">quant_mode</span><span class="o">=</span><span class="n">quant_mode</span><span class="p">,</span>
<span class="p">)</span>
<span class="bp">self</span><span class="o">.</span><span class="n">input_layernorm</span> <span class="o">=</span> <span class="n">ln_type</span><span class="p">(</span><span class="n">normalized_shape</span><span class="o">=</span><span class="n">hidden_size</span><span class="p">,</span>
<span class="n">eps</span><span class="o">=</span><span class="n">layernorm_eps</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="n">dtype</span><span class="p">)</span>
<span class="bp">self</span><span class="o">.</span><span class="n">gate_attn</span> <span class="o">=</span> <span class="n">Parameter</span><span class="p">(</span><span class="n">shape</span><span class="o">=</span><span class="nb">tuple</span><span class="p">((</span><span class="mi">1</span><span class="p">,</span> <span class="p">)),</span> <span class="n">dtype</span><span class="o">=</span><span class="n">dtype</span><span class="p">)</span>
<span class="bp">self</span><span class="o">.</span><span class="n">mlp_type</span> <span class="o">=</span> <span class="n">mlp_type</span>
<span class="n">mlp_f</span> <span class="o">=</span> <span class="n">mlp_map</span><span class="p">[</span><span class="n">mlp_type</span><span class="p">]</span>
<span class="bp">self</span><span class="o">.</span><span class="n">mlp</span> <span class="o">=</span> <span class="n">mlp_f</span><span class="p">(</span>
<span class="n">hidden_size</span><span class="o">=</span><span class="n">hidden_size</span><span class="p">,</span>
<span class="n">ffn_hidden_size</span><span class="o">=</span><span class="n">ffn_hidden_size</span><span class="p">,</span>
<span class="n">hidden_act</span><span class="o">=</span><span class="n">hidden_act</span><span class="p">,</span>
<span class="n">bias</span><span class="o">=</span><span class="n">has_mlp_bias</span><span class="p">,</span>
<span class="n">tp_group</span><span class="o">=</span><span class="n">mapping</span><span class="o">.</span><span class="n">tp_group</span><span class="p">,</span>
<span class="n">tp_size</span><span class="o">=</span><span class="n">mapping</span><span class="o">.</span><span class="n">tp_size</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="n">dtype</span><span class="p">,</span>
<span class="p">)</span>
<span class="bp">self</span><span class="o">.</span><span class="n">post_layernorm</span> <span class="o">=</span> <span class="n">ln_type</span><span class="p">(</span><span class="n">normalized_shape</span><span class="o">=</span><span class="n">hidden_size</span><span class="p">,</span>
<span class="n">eps</span><span class="o">=</span><span class="n">layernorm_eps</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="n">dtype</span><span class="p">)</span>
<span class="bp">self</span><span class="o">.</span><span class="n">gate_ffwd</span> <span class="o">=</span> <span class="n">Parameter</span><span class="p">(</span><span class="n">shape</span><span class="o">=</span><span class="nb">tuple</span><span class="p">((</span><span class="mi">1</span><span class="p">,</span> <span class="p">)),</span> <span class="n">dtype</span><span class="o">=</span><span class="n">dtype</span><span class="p">)</span>
<span class="bp">self</span><span class="o">.</span><span class="n">residual_scaling</span> <span class="o">=</span> <span class="n">residual_scaling</span>
<span class="bp">self</span><span class="o">.</span><span class="n">fp16_clamping</span> <span class="o">=</span> <span class="n">fp16_clamping</span>
<span class="bp">self</span><span class="o">.</span><span class="n">no_ffn</span> <span class="o">=</span> <span class="kc">False</span>
<span class="k">def</span><span class="w"> </span><span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span>
<span class="n">hidden_states</span><span class="p">:</span> <span class="n">Tensor</span><span class="p">,</span>
<span class="n">encoder_output</span><span class="p">:</span> <span class="n">Optional</span><span class="p">[</span><span class="n">Tensor</span><span class="p">]</span> <span class="o">=</span> <span class="kc">None</span><span class="p">,</span>
<span class="n">attention_mask_params</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span>
<span class="n">use_cache</span><span class="o">=</span><span class="kc">False</span><span class="p">,</span>
<span class="n">kv_cache_params</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span>
<span class="n">attention_params</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span>
<span class="n">lora_layer_params</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span>
<span class="n">cross_kv_cache_gen</span><span class="p">:</span> <span class="n">Optional</span><span class="p">[</span><span class="n">Tensor</span><span class="p">]</span> <span class="o">=</span> <span class="kc">None</span><span class="p">,</span>
<span class="n">cross_kv_reuse</span><span class="p">:</span> <span class="n">Optional</span><span class="p">[</span><span class="n">Tensor</span><span class="p">]</span> <span class="o">=</span> <span class="kc">None</span><span class="p">,</span>
<span class="n">full_text_row_masked_out_mask</span><span class="p">:</span> <span class="n">Tensor</span> <span class="o">=</span> <span class="kc">None</span><span class="p">,</span>
<span class="n">skip_cross_attn_blocks</span><span class="p">:</span> <span class="n">Tensor</span> <span class="o">=</span> <span class="kc">None</span><span class="p">):</span>
<span class="k">assert</span> <span class="nb">isinstance</span><span class="p">(</span><span class="n">hidden_states</span><span class="p">,</span> <span class="n">Tensor</span><span class="p">)</span>
<span class="k">if</span> <span class="n">encoder_output</span><span class="p">:</span>
<span class="k">assert</span> <span class="nb">isinstance</span><span class="p">(</span><span class="n">encoder_output</span><span class="p">,</span> <span class="n">Tensor</span><span class="p">)</span>
<span class="k">if</span> <span class="n">ADD_DEBUG_TENSOR</span><span class="p">:</span>
<span class="n">hidden_states</span><span class="o">.</span><span class="n">mark_output</span><span class="p">(</span>
<span class="sa">f</span><span class="s1">&#39;</span><span class="si">{</span><span class="bp">self</span><span class="o">.</span><span class="n">local_layer_idx</span><span class="si">:</span><span class="s1">2d</span><span class="si">}</span><span class="s1">/1.0: hidden_states&#39;</span><span class="p">,</span>
<span class="n">hidden_states</span><span class="o">.</span><span class="n">dtype</span><span class="p">)</span>
<span class="c1"># cross attention</span>
<span class="n">residual</span> <span class="o">=</span> <span class="n">hidden_states</span> <span class="o">*</span> <span class="bp">self</span><span class="o">.</span><span class="n">residual_scaling</span>
<span class="c1"># skip input_layernorm</span>
<span class="k">if</span> <span class="n">skip_cross_attn_blocks</span> <span class="ow">is</span> <span class="ow">not</span> <span class="kc">None</span><span class="p">:</span>
<span class="n">input_ln_conditional</span> <span class="o">=</span> <span class="n">Conditional</span><span class="p">(</span><span class="n">skip_cross_attn_blocks</span><span class="p">)</span>
<span class="n">skip_result</span> <span class="o">=</span> <span class="n">input_ln_conditional</span><span class="o">.</span><span class="n">add_input</span><span class="p">(</span><span class="n">hidden_states</span><span class="p">)</span>
<span class="n">hidden_states</span> <span class="o">=</span> <span class="n">input_ln_conditional</span><span class="o">.</span><span class="n">add_input</span><span class="p">(</span><span class="n">hidden_states</span><span class="p">)</span>
<span class="n">hidden_states</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">input_layernorm</span><span class="p">(</span><span class="n">hidden_states</span><span class="p">)</span>
<span class="n">hidden_states</span> <span class="o">=</span> <span class="n">input_ln_conditional</span><span class="o">.</span><span class="n">add_output</span><span class="p">(</span>
<span class="n">skip_result</span><span class="p">,</span> <span class="n">hidden_states</span><span class="p">)</span>
<span class="k">else</span><span class="p">:</span>
<span class="n">hidden_states</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">input_layernorm</span><span class="p">(</span><span class="n">hidden_states</span><span class="p">)</span>
<span class="k">if</span> <span class="n">ADD_DEBUG_TENSOR</span><span class="p">:</span>
<span class="n">hidden_states</span><span class="o">.</span><span class="n">mark_output</span><span class="p">(</span>
<span class="sa">f</span><span class="s1">&#39;</span><span class="si">{</span><span class="bp">self</span><span class="o">.</span><span class="n">local_layer_idx</span><span class="si">:</span><span class="s1">2d</span><span class="si">}</span><span class="s1">/2.1: normed_input&#39;</span><span class="p">,</span>
<span class="n">hidden_states</span><span class="o">.</span><span class="n">dtype</span><span class="p">)</span>
<span class="c1"># pass full_text_row_masked_out_mask and xattn_mask</span>
<span class="n">attention_output</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">cross_attention</span><span class="p">(</span>
<span class="n">hidden_states</span><span class="o">=</span><span class="n">hidden_states</span><span class="p">,</span>
<span class="n">attention_mask</span><span class="o">=</span><span class="n">attention_mask_params</span><span class="o">.</span><span class="n">cross_attention_mask</span><span class="p">,</span>
<span class="n">attention_packed_mask</span><span class="o">=</span><span class="n">attention_mask_params</span><span class="o">.</span>
<span class="n">cross_attention_packed_mask</span><span class="p">,</span>
<span class="n">encoder_output</span><span class="o">=</span><span class="n">encoder_output</span><span class="p">,</span>
<span class="n">use_cache</span><span class="o">=</span><span class="n">use_cache</span><span class="p">,</span>
<span class="n">kv_cache_params</span><span class="o">=</span><span class="n">kv_cache_params</span><span class="p">,</span>
<span class="n">attention_params</span><span class="o">=</span><span class="n">attention_params</span><span class="p">,</span>
<span class="n">lora_layer_params</span><span class="o">=</span><span class="n">lora_layer_params</span><span class="p">,</span>
<span class="n">cross_kv_cache_gen</span><span class="o">=</span><span class="n">cross_kv_cache_gen</span><span class="p">,</span>
<span class="n">cross_kv_reuse</span><span class="o">=</span><span class="n">cross_kv_reuse</span><span class="p">,</span>
<span class="n">skip_attn</span><span class="o">=</span><span class="n">skip_cross_attn_blocks</span><span class="p">,</span>
<span class="p">)</span>
<span class="k">if</span> <span class="n">use_cache</span><span class="p">:</span>
<span class="n">attention_output</span><span class="p">,</span> <span class="n">presents_cross</span> <span class="o">=</span> <span class="n">attention_output</span>
<span class="k">if</span> <span class="n">ADD_DEBUG_TENSOR</span><span class="p">:</span>
<span class="n">attention_output</span><span class="o">.</span><span class="n">mark_output</span><span class="p">(</span>
<span class="sa">f</span><span class="s1">&#39;</span><span class="si">{</span><span class="bp">self</span><span class="o">.</span><span class="n">local_layer_idx</span><span class="si">:</span><span class="s1">2d</span><span class="si">}</span><span class="s1">/3.1: cross_attention_output&#39;</span><span class="p">,</span>
<span class="n">attention_output</span><span class="o">.</span><span class="n">dtype</span><span class="p">)</span>
<span class="n">attn_residual_scale</span> <span class="o">=</span> <span class="n">tanh</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">gate_attn</span><span class="o">.</span><span class="n">value</span><span class="o">.</span><span class="n">cast</span><span class="p">(</span><span class="n">trt</span><span class="o">.</span><span class="n">float32</span><span class="p">))</span><span class="o">.</span><span class="n">cast</span><span class="p">(</span>
<span class="n">attention_output</span><span class="o">.</span><span class="n">dtype</span><span class="p">)</span>
<span class="n">attention_input</span> <span class="o">=</span> <span class="n">hidden_states</span>
<span class="n">hidden_states</span> <span class="o">=</span> <span class="n">residual</span> <span class="o">+</span> <span class="n">attn_residual_scale</span> <span class="o">*</span> <span class="n">attention_output</span>
<span class="c1"># use to skip attention_output with residual</span>
<span class="c1"># Since conditional does not work for gpt_attention_plugin, we replace the</span>
<span class="c1"># attention_output by hidden_states (input of attention) now.</span>
<span class="k">if</span> <span class="n">skip_cross_attn_blocks</span> <span class="ow">is</span> <span class="ow">not</span> <span class="kc">None</span><span class="p">:</span>
<span class="n">attn_conditional</span> <span class="o">=</span> <span class="n">Conditional</span><span class="p">(</span><span class="n">skip_cross_attn_blocks</span><span class="p">)</span>
<span class="n">skip_result</span> <span class="o">=</span> <span class="n">attn_conditional</span><span class="o">.</span><span class="n">add_input</span><span class="p">(</span><span class="n">attention_input</span><span class="p">)</span>
<span class="n">hidden_states</span> <span class="o">=</span> <span class="n">attn_conditional</span><span class="o">.</span><span class="n">add_input</span><span class="p">(</span><span class="n">hidden_states</span><span class="p">)</span>
<span class="n">hidden_states</span> <span class="o">=</span> <span class="n">attn_conditional</span><span class="o">.</span><span class="n">add_output</span><span class="p">(</span><span class="n">skip_result</span><span class="p">,</span>
<span class="n">hidden_states</span><span class="p">)</span>
<span class="k">if</span> <span class="n">ADD_DEBUG_TENSOR</span><span class="p">:</span>
<span class="n">hidden_states</span><span class="o">.</span><span class="n">mark_output</span><span class="p">(</span>
<span class="sa">f</span><span class="s1">&#39;</span><span class="si">{</span><span class="bp">self</span><span class="o">.</span><span class="n">local_layer_idx</span><span class="si">:</span><span class="s1">2d</span><span class="si">}</span><span class="s1">/3.2: cross_attn_output_with_residual&#39;</span><span class="p">,</span>
<span class="n">hidden_states</span><span class="o">.</span><span class="n">dtype</span><span class="p">)</span>
<span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">fp16_clamping</span><span class="p">:</span>
<span class="n">hidden_states</span> <span class="o">=</span> <span class="n">maximum</span><span class="p">(</span><span class="o">-</span><span class="mf">64000.0</span><span class="p">,</span> <span class="n">hidden_states</span><span class="p">)</span>
<span class="n">hidden_states</span> <span class="o">=</span> <span class="n">minimum</span><span class="p">(</span><span class="mf">64000.0</span><span class="p">,</span> <span class="n">hidden_states</span><span class="p">)</span>
<span class="c1"># MLP</span>
<span class="c1"># skip post_layernorm and mlp</span>
<span class="k">if</span> <span class="n">skip_cross_attn_blocks</span> <span class="ow">is</span> <span class="ow">not</span> <span class="kc">None</span><span class="p">:</span>
<span class="n">mlp_conditional</span> <span class="o">=</span> <span class="n">Conditional</span><span class="p">(</span><span class="n">skip_cross_attn_blocks</span><span class="p">)</span>
<span class="n">skip_case</span> <span class="o">=</span> <span class="n">mlp_conditional</span><span class="o">.</span><span class="n">add_input</span><span class="p">(</span><span class="n">hidden_states</span><span class="p">)</span>
<span class="n">hidden_states</span> <span class="o">=</span> <span class="n">mlp_conditional</span><span class="o">.</span><span class="n">add_input</span><span class="p">(</span><span class="n">hidden_states</span><span class="p">)</span>
<span class="n">attention_output</span> <span class="o">=</span> <span class="n">attention_output</span> <span class="o">*</span> <span class="n">full_text_row_masked_out_mask</span> <span class="c1"># TODO should move this mask into attention?</span>
<span class="n">residual</span> <span class="o">=</span> <span class="n">hidden_states</span> <span class="o">*</span> <span class="bp">self</span><span class="o">.</span><span class="n">residual_scaling</span>
<span class="n">hidden_states</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">post_layernorm</span><span class="p">(</span><span class="n">hidden_states</span><span class="p">)</span>
<span class="n">hidden_states</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">mlp</span><span class="p">(</span><span class="n">hidden_states</span><span class="p">,</span>
<span class="n">lora_layer_params</span><span class="o">=</span><span class="n">lora_layer_params</span><span class="p">)</span>
<span class="k">if</span> <span class="n">ADD_DEBUG_TENSOR</span><span class="p">:</span>
<span class="n">hidden_states</span><span class="o">.</span><span class="n">mark_output</span><span class="p">(</span>
<span class="sa">f</span><span class="s1">&#39;</span><span class="si">{</span><span class="bp">self</span><span class="o">.</span><span class="n">local_layer_idx</span><span class="si">:</span><span class="s1">2d</span><span class="si">}</span><span class="s1">/4.1: mlp_output&#39;</span><span class="p">,</span>
<span class="n">hidden_states</span><span class="o">.</span><span class="n">dtype</span><span class="p">)</span>
<span class="n">hidden_states</span> <span class="o">=</span> <span class="n">hidden_states</span> <span class="o">*</span> <span class="n">full_text_row_masked_out_mask</span>
<span class="k">if</span> <span class="n">ADD_DEBUG_TENSOR</span><span class="p">:</span>
<span class="n">hidden_states</span><span class="o">.</span><span class="n">mark_output</span><span class="p">(</span>
<span class="sa">f</span><span class="s1">&#39;</span><span class="si">{</span><span class="bp">self</span><span class="o">.</span><span class="n">local_layer_idx</span><span class="si">:</span><span class="s1">2d</span><span class="si">}</span><span class="s1">/4.2: masked_mlp_output&#39;</span><span class="p">,</span>
<span class="n">hidden_states</span><span class="o">.</span><span class="n">dtype</span><span class="p">)</span>
<span class="n">ffn_residual_scale</span> <span class="o">=</span> <span class="n">tanh</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">gate_ffwd</span><span class="o">.</span><span class="n">value</span><span class="o">.</span><span class="n">cast</span><span class="p">(</span><span class="n">trt</span><span class="o">.</span><span class="n">float32</span><span class="p">))</span><span class="o">.</span><span class="n">cast</span><span class="p">(</span>
<span class="n">hidden_states</span><span class="o">.</span><span class="n">dtype</span><span class="p">)</span>
<span class="n">hidden_states</span> <span class="o">=</span> <span class="n">residual</span> <span class="o">+</span> <span class="n">ffn_residual_scale</span> <span class="o">*</span> <span class="n">hidden_states</span> <span class="o">*</span> <span class="nb">float</span><span class="p">(</span>
<span class="ow">not</span> <span class="bp">self</span><span class="o">.</span><span class="n">no_ffn</span><span class="p">)</span>
<span class="k">if</span> <span class="n">skip_cross_attn_blocks</span> <span class="ow">is</span> <span class="ow">not</span> <span class="kc">None</span><span class="p">:</span>
<span class="n">hidden_states</span> <span class="o">=</span> <span class="n">mlp_conditional</span><span class="o">.</span><span class="n">add_output</span><span class="p">(</span><span class="n">skip_case</span><span class="p">,</span> <span class="n">hidden_states</span><span class="p">)</span>
<span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">fp16_clamping</span><span class="p">:</span>
<span class="n">hidden_states</span> <span class="o">=</span> <span class="n">maximum</span><span class="p">(</span><span class="o">-</span><span class="mf">64000.0</span><span class="p">,</span> <span class="n">hidden_states</span><span class="p">)</span>
<span class="n">hidden_states</span> <span class="o">=</span> <span class="n">minimum</span><span class="p">(</span><span class="mf">64000.0</span><span class="p">,</span> <span class="n">hidden_states</span><span class="p">)</span>
<span class="k">if</span> <span class="n">ADD_DEBUG_TENSOR</span><span class="p">:</span>
<span class="n">hidden_states</span><span class="o">.</span><span class="n">mark_output</span><span class="p">(</span>
<span class="sa">f</span><span class="s1">&#39;</span><span class="si">{</span><span class="bp">self</span><span class="o">.</span><span class="n">local_layer_idx</span><span class="si">:</span><span class="s1">2d</span><span class="si">}</span><span class="s1">/4.4: transformer_out&#39;</span><span class="p">,</span>
<span class="n">hidden_states</span><span class="o">.</span><span class="n">dtype</span><span class="p">)</span>
<span class="k">if</span> <span class="n">use_cache</span><span class="p">:</span>
<span class="k">return</span> <span class="p">(</span><span class="n">hidden_states</span><span class="p">,</span> <span class="n">presents_cross</span><span class="p">)</span>
<span class="k">return</span> <span class="n">hidden_states</span>
<span class="k">class</span><span class="w"> </span><span class="nc">TransformerBlock</span><span class="p">(</span><span class="n">Module</span><span class="p">):</span>
<span class="k">def</span><span class="w"> </span><span class="fm">__init__</span><span class="p">(</span>
<span class="bp">self</span><span class="p">,</span>
<span class="o">*</span><span class="p">,</span>
<span class="n">local_layer_idx</span><span class="p">,</span>
<span class="n">hidden_size</span><span class="p">,</span>
<span class="n">ffn_hidden_size</span><span class="p">,</span>
<span class="n">num_attention_heads</span><span class="p">,</span>
<span class="n">num_kv_heads</span><span class="p">,</span>
<span class="n">head_size</span><span class="p">,</span>
<span class="n">max_position_embeddings</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span>
<span class="n">q_scaling</span><span class="o">=</span><span class="mf">1.0</span><span class="p">,</span>
<span class="n">has_attention_qkvo_bias</span><span class="o">=</span><span class="kc">False</span><span class="p">,</span>
<span class="n">has_mlp_bias</span><span class="o">=</span><span class="kc">False</span><span class="p">,</span>
<span class="n">layernorm_position</span><span class="o">=</span><span class="n">LayerNormPositionType</span><span class="o">.</span><span class="n">pre_layernorm</span><span class="p">,</span>
<span class="n">layernorm_type</span><span class="o">=</span><span class="n">LayerNormType</span><span class="o">.</span><span class="n">RmsNorm</span><span class="p">,</span>
<span class="n">layernorm_eps</span><span class="o">=</span><span class="mf">1e-5</span><span class="p">,</span>
<span class="n">hidden_act</span><span class="o">=</span><span class="s2">&quot;gated-silu&quot;</span><span class="p">,</span>
<span class="n">mlp_type</span><span class="o">=</span><span class="n">MLPType</span><span class="o">.</span><span class="n">GatedMLP</span><span class="p">,</span>
<span class="n">mapping</span><span class="o">=</span><span class="n">Mapping</span><span class="p">(),</span>
<span class="n">dtype</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span>
<span class="n">residual_scaling</span><span class="o">=</span><span class="mf">1.0</span><span class="p">,</span>
<span class="n">relative_attention</span><span class="o">=</span><span class="kc">False</span><span class="p">,</span>
<span class="n">max_distance</span><span class="o">=</span><span class="mi">0</span><span class="p">,</span>
<span class="n">num_buckets</span><span class="o">=</span><span class="mi">0</span><span class="p">,</span>
<span class="n">fp16_clamping</span><span class="o">=</span><span class="kc">False</span><span class="p">,</span>
<span class="n">skip_cross_kv</span><span class="o">=</span><span class="kc">False</span><span class="p">,</span>
<span class="n">use_implicit_relative_attention</span><span class="o">=</span><span class="kc">False</span><span class="p">,</span>
<span class="n">rotary_embedding_base</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span>
<span class="n">rotary_embedding_scaling</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span>
<span class="n">quant_mode</span><span class="o">=</span><span class="n">QuantMode</span><span class="p">(</span><span class="mi">0</span><span class="p">),</span>
<span class="p">):</span>
<span class="nb">super</span><span class="p">()</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span>
<span class="bp">self</span><span class="o">.</span><span class="n">local_layer_idx</span> <span class="o">=</span> <span class="n">local_layer_idx</span>
<span class="bp">self</span><span class="o">.</span><span class="n">layernorm_type</span> <span class="o">=</span> <span class="n">layernorm_type</span>
<span class="n">ln_type</span> <span class="o">=</span> <span class="n">layernorm_map</span><span class="p">[</span><span class="n">layernorm_type</span><span class="p">]</span>
<span class="bp">self</span><span class="o">.</span><span class="n">layernorm_position</span> <span class="o">=</span> <span class="n">layernorm_position</span>
<span class="k">assert</span> <span class="bp">self</span><span class="o">.</span><span class="n">layernorm_position</span> <span class="o">==</span> <span class="n">LayerNormPositionType</span><span class="o">.</span><span class="n">pre_layernorm</span>
<span class="bp">self</span><span class="o">.</span><span class="n">self_attention</span> <span class="o">=</span> <span class="n">Attention</span><span class="p">(</span>
<span class="n">local_layer_idx</span><span class="o">=</span><span class="n">local_layer_idx</span><span class="p">,</span>
<span class="n">hidden_size</span><span class="o">=</span><span class="n">hidden_size</span><span class="p">,</span>
<span class="n">num_attention_heads</span><span class="o">=</span><span class="n">num_attention_heads</span><span class="p">,</span>
<span class="n">attention_head_size</span><span class="o">=</span><span class="n">head_size</span><span class="p">,</span>
<span class="n">num_kv_heads</span><span class="o">=</span><span class="n">num_kv_heads</span><span class="p">,</span>
<span class="n">max_position_embeddings</span><span class="o">=</span><span class="n">max_position_embeddings</span><span class="p">,</span>
<span class="n">q_scaling</span><span class="o">=</span><span class="n">q_scaling</span><span class="p">,</span>
<span class="n">bias</span><span class="o">=</span><span class="n">has_attention_qkvo_bias</span><span class="p">,</span>
<span class="n">attention_mask_type</span><span class="o">=</span><span class="n">AttentionMaskType</span><span class="o">.</span><span class="n">causal</span><span class="p">,</span>
<span class="n">tp_group</span><span class="o">=</span><span class="n">mapping</span><span class="o">.</span><span class="n">tp_group</span><span class="p">,</span>
<span class="n">tp_size</span><span class="o">=</span><span class="n">mapping</span><span class="o">.</span><span class="n">tp_size</span><span class="p">,</span>
<span class="n">tp_rank</span><span class="o">=</span><span class="n">mapping</span><span class="o">.</span><span class="n">tp_rank</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="n">dtype</span><span class="p">,</span>
<span class="n">cross_attention</span><span class="o">=</span><span class="kc">False</span><span class="p">,</span>
<span class="n">relative_attention</span><span class="o">=</span><span class="n">relative_attention</span><span class="p">,</span>
<span class="n">max_distance</span><span class="o">=</span><span class="n">max_distance</span> <span class="k">if</span> <span class="n">use_implicit_relative_attention</span> <span class="k">else</span> <span class="mi">0</span><span class="p">,</span>
<span class="n">num_buckets</span><span class="o">=</span><span class="n">num_buckets</span><span class="p">,</span>
<span class="n">position_embedding_type</span><span class="o">=</span><span class="n">PositionEmbeddingType</span><span class="o">.</span><span class="n">relative</span>
<span class="k">if</span> <span class="n">relative_attention</span> <span class="k">else</span> <span class="n">PositionEmbeddingType</span><span class="o">.</span><span class="n">rope_gpt_neox</span><span class="p">,</span>
<span class="n">use_implicit_relative_attention</span><span class="o">=</span><span class="n">use_implicit_relative_attention</span><span class="p">,</span>
<span class="n">rotary_embedding_base</span><span class="o">=</span><span class="n">rotary_embedding_base</span><span class="p">,</span>
<span class="n">rotary_embedding_scaling</span><span class="o">=</span><span class="n">rotary_embedding_scaling</span><span class="p">,</span>
<span class="n">quant_mode</span><span class="o">=</span><span class="n">quant_mode</span><span class="p">,</span>
<span class="p">)</span>
<span class="bp">self</span><span class="o">.</span><span class="n">input_layernorm</span> <span class="o">=</span> <span class="n">ln_type</span><span class="p">(</span><span class="n">normalized_shape</span><span class="o">=</span><span class="n">hidden_size</span><span class="p">,</span>
<span class="n">eps</span><span class="o">=</span><span class="n">layernorm_eps</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="n">dtype</span><span class="p">)</span>
<span class="bp">self</span><span class="o">.</span><span class="n">mlp_type</span> <span class="o">=</span> <span class="n">mlp_type</span>
<span class="n">mlp_f</span> <span class="o">=</span> <span class="n">mlp_map</span><span class="p">[</span><span class="n">mlp_type</span><span class="p">]</span>
<span class="bp">self</span><span class="o">.</span><span class="n">mlp</span> <span class="o">=</span> <span class="n">mlp_f</span><span class="p">(</span>
<span class="n">hidden_size</span><span class="o">=</span><span class="n">hidden_size</span><span class="p">,</span>
<span class="n">ffn_hidden_size</span><span class="o">=</span><span class="n">ffn_hidden_size</span><span class="p">,</span>
<span class="n">hidden_act</span><span class="o">=</span><span class="n">hidden_act</span><span class="p">,</span>
<span class="n">bias</span><span class="o">=</span><span class="n">has_mlp_bias</span><span class="p">,</span>
<span class="n">tp_group</span><span class="o">=</span><span class="n">mapping</span><span class="o">.</span><span class="n">tp_group</span><span class="p">,</span>
<span class="n">tp_size</span><span class="o">=</span><span class="n">mapping</span><span class="o">.</span><span class="n">tp_size</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="n">dtype</span><span class="p">,</span>
<span class="p">)</span>
<span class="bp">self</span><span class="o">.</span><span class="n">post_layernorm</span> <span class="o">=</span> <span class="n">ln_type</span><span class="p">(</span><span class="n">normalized_shape</span><span class="o">=</span><span class="n">hidden_size</span><span class="p">,</span>
<span class="n">eps</span><span class="o">=</span><span class="n">layernorm_eps</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="n">dtype</span><span class="p">)</span>
<span class="bp">self</span><span class="o">.</span><span class="n">residual_scaling</span> <span class="o">=</span> <span class="n">residual_scaling</span>
<span class="bp">self</span><span class="o">.</span><span class="n">fp16_clamping</span> <span class="o">=</span> <span class="n">fp16_clamping</span>
<span class="k">def</span><span class="w"> </span><span class="nf">forward</span><span class="p">(</span>
<span class="bp">self</span><span class="p">,</span>
<span class="n">hidden_states</span><span class="p">:</span> <span class="n">Tensor</span><span class="p">,</span>
<span class="n">encoder_output</span><span class="p">:</span> <span class="n">Optional</span><span class="p">[</span><span class="n">Tensor</span><span class="p">]</span> <span class="o">=</span> <span class="kc">None</span><span class="p">,</span> <span class="c1"># not used</span>
<span class="n">attention_mask_params</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span>
<span class="n">use_cache</span><span class="o">=</span><span class="kc">False</span><span class="p">,</span>
<span class="n">kv_cache_params</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span>
<span class="n">attention_params</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span>
<span class="n">lora_layer_params</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span>
<span class="n">cross_kv_cache_gen</span><span class="p">:</span> <span class="n">Optional</span><span class="p">[</span><span class="n">Tensor</span><span class="p">]</span> <span class="o">=</span> <span class="kc">None</span><span class="p">,</span>
<span class="n">cross_kv_reuse</span><span class="p">:</span> <span class="n">Optional</span><span class="p">[</span><span class="n">Tensor</span><span class="p">]</span> <span class="o">=</span> <span class="kc">None</span><span class="p">,</span>
<span class="n">full_text_row_masked_out_mask</span><span class="p">:</span> <span class="n">Tensor</span> <span class="o">=</span> <span class="kc">None</span><span class="p">,</span> <span class="c1"># not used</span>
<span class="n">skip_cross_attn_blocks</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span>
<span class="p">):</span>
<span class="k">assert</span> <span class="nb">isinstance</span><span class="p">(</span><span class="n">hidden_states</span><span class="p">,</span> <span class="n">Tensor</span><span class="p">)</span>
<span class="c1"># self-attention</span>
<span class="n">residual</span> <span class="o">=</span> <span class="n">hidden_states</span> <span class="o">*</span> <span class="bp">self</span><span class="o">.</span><span class="n">residual_scaling</span>
<span class="k">if</span> <span class="n">ADD_DEBUG_TENSOR</span><span class="p">:</span>
<span class="n">hidden_states</span><span class="o">.</span><span class="n">mark_output</span><span class="p">(</span>
<span class="sa">f</span><span class="s1">&#39;</span><span class="si">{</span><span class="bp">self</span><span class="o">.</span><span class="n">local_layer_idx</span><span class="si">:</span><span class="s1">2d</span><span class="si">}</span><span class="s1">/1.0: hidden_states&#39;</span><span class="p">,</span>
<span class="n">hidden_states</span><span class="o">.</span><span class="n">dtype</span><span class="p">)</span>
<span class="n">hidden_states</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">input_layernorm</span><span class="p">(</span><span class="n">hidden_states</span><span class="p">)</span>
<span class="k">if</span> <span class="n">ADD_DEBUG_TENSOR</span><span class="p">:</span>
<span class="n">hidden_states</span><span class="o">.</span><span class="n">mark_output</span><span class="p">(</span>
<span class="sa">f</span><span class="s1">&#39;</span><span class="si">{</span><span class="bp">self</span><span class="o">.</span><span class="n">local_layer_idx</span><span class="si">:</span><span class="s1">2d</span><span class="si">}</span><span class="s1">/2.1: normed attn_input&#39;</span><span class="p">,</span>
<span class="n">hidden_states</span><span class="o">.</span><span class="n">dtype</span><span class="p">)</span>
<span class="n">attention_output</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">self_attention</span><span class="p">(</span>
<span class="n">hidden_states</span><span class="o">=</span><span class="n">hidden_states</span><span class="p">,</span>
<span class="n">attention_mask</span><span class="o">=</span><span class="n">attention_mask_params</span><span class="o">.</span><span class="n">self_attention_mask</span><span class="p">,</span>
<span class="n">use_cache</span><span class="o">=</span><span class="n">use_cache</span><span class="p">,</span>
<span class="n">kv_cache_params</span><span class="o">=</span><span class="n">kv_cache_params</span><span class="p">,</span>
<span class="n">attention_params</span><span class="o">=</span><span class="n">attention_params</span><span class="p">,</span>
<span class="n">lora_layer_params</span><span class="o">=</span><span class="n">lora_layer_params</span><span class="p">)</span>
<span class="k">if</span> <span class="n">use_cache</span><span class="p">:</span>
<span class="n">attention_output</span><span class="p">,</span> <span class="n">presents_self</span> <span class="o">=</span> <span class="n">attention_output</span>
<span class="k">if</span> <span class="n">ADD_DEBUG_TENSOR</span><span class="p">:</span>
<span class="n">attention_output</span><span class="o">.</span><span class="n">mark_output</span><span class="p">(</span>
<span class="sa">f</span><span class="s1">&#39;</span><span class="si">{</span><span class="bp">self</span><span class="o">.</span><span class="n">local_layer_idx</span><span class="si">:</span><span class="s1">2d</span><span class="si">}</span><span class="s1">/3.1: self_attention_output&#39;</span><span class="p">,</span>
<span class="n">attention_output</span><span class="o">.</span><span class="n">dtype</span><span class="p">)</span>
<span class="n">hidden_states</span> <span class="o">=</span> <span class="n">residual</span> <span class="o">+</span> <span class="n">attention_output</span>
<span class="k">if</span> <span class="n">ADD_DEBUG_TENSOR</span><span class="p">:</span>
<span class="n">hidden_states</span><span class="o">.</span><span class="n">mark_output</span><span class="p">(</span>
<span class="sa">f</span><span class="s1">&#39;</span><span class="si">{</span><span class="bp">self</span><span class="o">.</span><span class="n">local_layer_idx</span><span class="si">:</span><span class="s1">2d</span><span class="si">}</span><span class="s1">/3.1: attention_output_with_residual&#39;</span><span class="p">,</span>
<span class="n">hidden_states</span><span class="o">.</span><span class="n">dtype</span><span class="p">)</span>
<span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">fp16_clamping</span><span class="p">:</span>
<span class="n">hidden_states</span> <span class="o">=</span> <span class="n">maximum</span><span class="p">(</span><span class="o">-</span><span class="mf">64000.0</span><span class="p">,</span> <span class="n">hidden_states</span><span class="p">)</span>
<span class="n">hidden_states</span> <span class="o">=</span> <span class="n">minimum</span><span class="p">(</span><span class="mf">64000.0</span><span class="p">,</span> <span class="n">hidden_states</span><span class="p">)</span>
<span class="c1"># MLP</span>
<span class="n">residual</span> <span class="o">=</span> <span class="n">hidden_states</span> <span class="o">*</span> <span class="bp">self</span><span class="o">.</span><span class="n">residual_scaling</span>
<span class="n">hidden_states</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">post_layernorm</span><span class="p">(</span><span class="n">hidden_states</span><span class="p">)</span>
<span class="k">if</span> <span class="n">ADD_DEBUG_TENSOR</span><span class="p">:</span>
<span class="n">hidden_states</span><span class="o">.</span><span class="n">mark_output</span><span class="p">(</span>
<span class="sa">f</span><span class="s1">&#39;</span><span class="si">{</span><span class="bp">self</span><span class="o">.</span><span class="n">local_layer_idx</span><span class="si">:</span><span class="s1">2d</span><span class="si">}</span><span class="s1">/3.2: normed_mlp_input&#39;</span><span class="p">,</span>
<span class="n">hidden_states</span><span class="o">.</span><span class="n">dtype</span><span class="p">)</span>
<span class="n">hidden_states</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">mlp</span><span class="p">(</span><span class="n">hidden_states</span><span class="p">,</span>
<span class="n">lora_layer_params</span><span class="o">=</span><span class="n">lora_layer_params</span><span class="p">)</span>
<span class="k">if</span> <span class="n">ADD_DEBUG_TENSOR</span><span class="p">:</span>
<span class="n">hidden_states</span><span class="o">.</span><span class="n">mark_output</span><span class="p">(</span>
<span class="sa">f</span><span class="s1">&#39;</span><span class="si">{</span><span class="bp">self</span><span class="o">.</span><span class="n">local_layer_idx</span><span class="si">:</span><span class="s1">2d</span><span class="si">}</span><span class="s1">/4.1: mlp_output&#39;</span><span class="p">,</span>
<span class="n">hidden_states</span><span class="o">.</span><span class="n">dtype</span><span class="p">)</span>
<span class="n">hidden_states</span> <span class="o">=</span> <span class="n">residual</span> <span class="o">+</span> <span class="n">hidden_states</span>
<span class="k">if</span> <span class="n">ADD_DEBUG_TENSOR</span><span class="p">:</span>
<span class="n">hidden_states</span><span class="o">.</span><span class="n">mark_output</span><span class="p">(</span>
<span class="sa">f</span><span class="s1">&#39;</span><span class="si">{</span><span class="bp">self</span><span class="o">.</span><span class="n">local_layer_idx</span><span class="si">:</span><span class="s1">2d</span><span class="si">}</span><span class="s1">/4.2: mlp_output_residual&#39;</span><span class="p">,</span>
<span class="n">hidden_states</span><span class="o">.</span><span class="n">dtype</span><span class="p">)</span>
<span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">fp16_clamping</span><span class="p">:</span>
<span class="n">hidden_states</span> <span class="o">=</span> <span class="n">maximum</span><span class="p">(</span><span class="o">-</span><span class="mf">64000.0</span><span class="p">,</span> <span class="n">hidden_states</span><span class="p">)</span>
<span class="n">hidden_states</span> <span class="o">=</span> <span class="n">minimum</span><span class="p">(</span><span class="mf">64000.0</span><span class="p">,</span> <span class="n">hidden_states</span><span class="p">)</span>
<span class="k">if</span> <span class="n">use_cache</span><span class="p">:</span>
<span class="k">return</span> <span class="p">(</span><span class="n">hidden_states</span><span class="p">,</span> <span class="n">presents_self</span><span class="p">)</span>
<span class="k">return</span> <span class="n">hidden_states</span>
<span class="k">class</span><span class="w"> </span><span class="nc">MLLaMAModel</span><span class="p">(</span><span class="n">Module</span><span class="p">):</span>
<span class="k">def</span><span class="w"> </span><span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">config</span><span class="p">:</span> <span class="n">MLLaMAConfig</span><span class="p">)</span> <span class="o">-&gt;</span> <span class="kc">None</span><span class="p">:</span>
<span class="nb">super</span><span class="p">()</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span>
<span class="bp">self</span><span class="o">.</span><span class="n">config</span> <span class="o">=</span> <span class="n">config</span>
<span class="bp">self</span><span class="o">.</span><span class="n">position_embedding_type</span> <span class="o">=</span> <span class="n">config</span><span class="o">.</span><span class="n">position_embedding_type</span>
<span class="bp">self</span><span class="o">.</span><span class="n">mapping</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">mapping</span>
<span class="bp">self</span><span class="o">.</span><span class="n">layernorm_type</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">layernorm_type</span>
<span class="n">ln_type</span> <span class="o">=</span> <span class="n">layernorm_map</span><span class="p">[</span><span class="bp">self</span><span class="o">.</span><span class="n">layernorm_type</span><span class="p">]</span>
<span class="bp">self</span><span class="o">.</span><span class="n">has_attention_qkvo_bias</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">has_attention_qkvo_bias</span>
<span class="bp">self</span><span class="o">.</span><span class="n">has_mlp_bias</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">has_mlp_bias</span>
<span class="bp">self</span><span class="o">.</span><span class="n">has_model_final_layernorm</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">has_model_final_layernorm</span>
<span class="bp">self</span><span class="o">.</span><span class="n">_dtype</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">dtype</span>
<span class="c1"># no quantization considered for now</span>
<span class="bp">self</span><span class="o">.</span><span class="n">_kv_dtype</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">_dtype</span>
<span class="bp">self</span><span class="o">.</span><span class="n">_logits_dtype</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">logits_dtype</span>
<span class="bp">self</span><span class="o">.</span><span class="n">total_num_layers</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">num_hidden_layers</span>
<span class="bp">self</span><span class="o">.</span><span class="n">num_layers</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">num_hidden_layers</span> <span class="o">//</span> <span class="bp">self</span><span class="o">.</span><span class="n">mapping</span><span class="o">.</span><span class="n">pp_size</span>
<span class="bp">self</span><span class="o">.</span><span class="n">hidden_size</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">hidden_size</span>
<span class="bp">self</span><span class="o">.</span><span class="n">encoder_hidden_size</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">hidden_size</span>
<span class="bp">self</span><span class="o">.</span><span class="n">num_heads</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">num_attention_heads</span>
<span class="c1"># num_kv_heads = self.num_heads</span>
<span class="n">num_kv_heads</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">num_key_value_heads</span>
<span class="k">if</span> <span class="n">num_kv_heads</span> <span class="ow">is</span> <span class="kc">None</span> <span class="ow">or</span> <span class="n">num_kv_heads</span> <span class="o">&lt;=</span> <span class="mi">0</span><span class="p">:</span>
<span class="n">num_kv_heads</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">num_heads</span>
<span class="bp">self</span><span class="o">.</span><span class="n">num_kv_heads</span> <span class="o">=</span> <span class="n">num_kv_heads</span>
<span class="bp">self</span><span class="o">.</span><span class="n">head_size</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">hidden_size</span> <span class="o">//</span> <span class="bp">self</span><span class="o">.</span><span class="n">num_heads</span> <span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">head_size</span> <span class="ow">is</span> <span class="kc">None</span> <span class="k">else</span> <span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">head_size</span>
<span class="bp">self</span><span class="o">.</span><span class="n">fp16_clamping</span> <span class="o">=</span> <span class="kc">False</span>
<span class="bp">self</span><span class="o">.</span><span class="n">skip_cross_kv</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">skip_cross_kv</span>
<span class="bp">self</span><span class="o">.</span><span class="n">mlp_type</span> <span class="o">=</span> <span class="n">MLPType</span><span class="o">.</span><span class="n">MLP</span> <span class="k">if</span> <span class="ow">not</span> <span class="nb">hasattr</span><span class="p">(</span>
<span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="p">,</span> <span class="s2">&quot;mlp_type&quot;</span><span class="p">)</span> <span class="k">else</span> <span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">mlp_type</span>
<span class="bp">self</span><span class="o">.</span><span class="n">use_implicit_relative_attention</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">use_implicit_relative_attention</span> <span class="k">if</span> <span class="nb">hasattr</span><span class="p">(</span>
<span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="p">,</span> <span class="s2">&quot;use_implicit_relative_attention&quot;</span><span class="p">)</span> <span class="k">else</span> <span class="kc">False</span>
<span class="bp">self</span><span class="o">.</span><span class="n">cross_attention_layers</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">cross_attention_layers</span>
<span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">mapping</span><span class="o">.</span><span class="n">is_first_pp_rank</span><span class="p">():</span>
<span class="bp">self</span><span class="o">.</span><span class="n">vocab_embedding</span> <span class="o">=</span> <span class="n">Embedding</span><span class="p">(</span>
<span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">embed_vocab_size</span><span class="p">,</span>
<span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">hidden_size</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">_dtype</span><span class="p">,</span>
<span class="n">tp_size</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">mapping</span><span class="o">.</span><span class="n">tp_size</span>
<span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">use_parallel_embedding</span> <span class="k">else</span> <span class="mi">1</span><span class="p">,</span>
<span class="n">tp_group</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">mapping</span><span class="o">.</span><span class="n">tp_group</span>
<span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">use_parallel_embedding</span> <span class="k">else</span> <span class="kc">None</span><span class="p">,</span>
<span class="n">sharding_dim</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">embedding_sharding_dim</span><span class="p">,</span>
<span class="n">tp_rank</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">mapping</span><span class="o">.</span><span class="n">tp_rank</span><span class="p">)</span>
<span class="n">layers_range</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">mapping</span><span class="o">.</span><span class="n">pp_layers</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">total_num_layers</span><span class="p">)</span>
<span class="n">_layers</span> <span class="o">=</span> <span class="p">[]</span>
<span class="k">for</span> <span class="n">layer_idx</span> <span class="ow">in</span> <span class="n">layers_range</span><span class="p">:</span>
<span class="n">local_layer_idx</span> <span class="o">=</span> <span class="n">layer_idx</span> <span class="o">-</span> <span class="n">layers_range</span><span class="p">[</span><span class="mi">0</span><span class="p">]</span>
<span class="n">args</span> <span class="o">=</span> <span class="p">{</span>
<span class="s2">&quot;local_layer_idx&quot;</span><span class="p">:</span> <span class="n">local_layer_idx</span><span class="p">,</span>
<span class="s2">&quot;hidden_size&quot;</span><span class="p">:</span> <span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">hidden_size</span><span class="p">,</span>
<span class="s2">&quot;ffn_hidden_size&quot;</span><span class="p">:</span> <span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">intermediate_size</span><span class="p">,</span>
<span class="s2">&quot;num_attention_heads&quot;</span><span class="p">:</span> <span class="bp">self</span><span class="o">.</span><span class="n">num_heads</span><span class="p">,</span>
<span class="s2">&quot;num_kv_heads&quot;</span><span class="p">:</span> <span class="bp">self</span><span class="o">.</span><span class="n">num_kv_heads</span><span class="p">,</span>
<span class="s2">&quot;head_size&quot;</span><span class="p">:</span> <span class="bp">self</span><span class="o">.</span><span class="n">head_size</span><span class="p">,</span>
<span class="s2">&quot;max_position_embeddings&quot;</span><span class="p">:</span> <span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">max_position_embeddings</span><span class="p">,</span>
<span class="s2">&quot;layernorm_position&quot;</span><span class="p">:</span> <span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">layernorm_position</span><span class="p">,</span>
<span class="s2">&quot;layernorm_eps&quot;</span><span class="p">:</span> <span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">norm_epsilon</span><span class="p">,</span>
<span class="s2">&quot;layernorm_type&quot;</span><span class="p">:</span> <span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">layernorm_type</span><span class="p">,</span>
<span class="s2">&quot;hidden_act&quot;</span><span class="p">:</span> <span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">hidden_act</span><span class="p">,</span>
<span class="s2">&quot;mlp_type&quot;</span><span class="p">:</span> <span class="bp">self</span><span class="o">.</span><span class="n">mlp_type</span><span class="p">,</span>
<span class="s2">&quot;mapping&quot;</span><span class="p">:</span> <span class="bp">self</span><span class="o">.</span><span class="n">mapping</span><span class="p">,</span>
<span class="s2">&quot;dtype&quot;</span><span class="p">:</span> <span class="bp">self</span><span class="o">.</span><span class="n">_dtype</span><span class="p">,</span>
<span class="s2">&quot;residual_scaling&quot;</span><span class="p">:</span> <span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">residual_scaling</span><span class="p">,</span>
<span class="s2">&quot;max_distance&quot;</span><span class="p">:</span> <span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">max_distance</span><span class="p">,</span>
<span class="s2">&quot;num_buckets&quot;</span><span class="p">:</span> <span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">num_buckets</span><span class="p">,</span>
<span class="s2">&quot;fp16_clamping&quot;</span><span class="p">:</span> <span class="bp">self</span><span class="o">.</span><span class="n">fp16_clamping</span><span class="p">,</span>
<span class="s2">&quot;skip_cross_kv&quot;</span><span class="p">:</span> <span class="bp">self</span><span class="o">.</span><span class="n">skip_cross_kv</span><span class="p">,</span>
<span class="s2">&quot;rotary_embedding_base&quot;</span><span class="p">:</span> <span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">rotary_base</span><span class="p">,</span>
<span class="s2">&quot;rotary_embedding_scaling&quot;</span><span class="p">:</span> <span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">rotary_scaling</span><span class="p">,</span>
<span class="s2">&quot;quant_mode&quot;</span><span class="p">:</span> <span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">quant_mode</span><span class="p">,</span>
<span class="p">}</span>
<span class="k">if</span> <span class="n">layer_idx</span> <span class="ow">in</span> <span class="bp">self</span><span class="o">.</span><span class="n">cross_attention_layers</span><span class="p">:</span>
<span class="k">assert</span> <span class="n">layers_range</span><span class="p">[</span><span class="mi">0</span><span class="p">]</span> <span class="o">==</span> <span class="mi">0</span><span class="p">,</span> <span class="s2">&quot;not support PP now&quot;</span>
<span class="n">_layers</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">CrossAttentionTransformerBlock</span><span class="p">(</span><span class="o">**</span><span class="n">args</span><span class="p">))</span>
<span class="k">else</span><span class="p">:</span>
<span class="n">_layers</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">TransformerBlock</span><span class="p">(</span><span class="o">**</span><span class="n">args</span><span class="p">))</span>
<span class="bp">self</span><span class="o">.</span><span class="n">layers</span> <span class="o">=</span> <span class="n">ModuleList</span><span class="p">(</span><span class="n">_layers</span><span class="p">)</span>
<span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">mapping</span><span class="o">.</span><span class="n">is_last_pp_rank</span><span class="p">():</span>
<span class="bp">self</span><span class="o">.</span><span class="n">ln_f</span> <span class="o">=</span> <span class="kc">None</span>
<span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">has_model_final_layernorm</span><span class="p">:</span>
<span class="bp">self</span><span class="o">.</span><span class="n">ln_f</span> <span class="o">=</span> <span class="n">ln_type</span><span class="p">(</span><span class="n">normalized_shape</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">hidden_size</span><span class="p">,</span>
<span class="n">eps</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">norm_epsilon</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">dtype</span><span class="p">)</span>
<span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">relative_attention</span> <span class="ow">and</span> <span class="ow">not</span> <span class="bp">self</span><span class="o">.</span><span class="n">use_implicit_relative_attention</span><span class="p">:</span>
<span class="bp">self</span><span class="o">.</span><span class="n">rel_attn_table</span> <span class="o">=</span> <span class="n">Parameter</span><span class="p">(</span>
<span class="n">shape</span><span class="o">=</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">num_attention_heads</span> <span class="o">//</span> <span class="bp">self</span><span class="o">.</span><span class="n">mapping</span><span class="o">.</span><span class="n">tp_size</span><span class="p">,</span>
<span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">num_buckets</span><span class="p">),</span>
<span class="n">dtype</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">_dtype</span><span class="p">)</span>
<span class="k">def</span><span class="w"> </span><span class="nf">forward</span><span class="p">(</span>
<span class="bp">self</span><span class="p">,</span>
<span class="n">decoder_input_ids</span><span class="p">:</span> <span class="n">Tensor</span><span class="p">,</span>
<span class="n">encoder_output</span><span class="p">:</span> <span class="n">Tensor</span><span class="p">,</span>
<span class="n">use_cache</span><span class="o">=</span><span class="kc">False</span><span class="p">,</span>
<span class="n">attention_mask_params</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span>
<span class="n">last_token_ids</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span>
<span class="n">kv_cache_params</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span>
<span class="n">attention_params</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span>
<span class="n">hidden_states</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span>
<span class="n">lora_params</span><span class="p">:</span> <span class="n">LoraParams</span> <span class="o">=</span> <span class="kc">None</span><span class="p">,</span>
<span class="n">cross_kv_cache_gen</span><span class="p">:</span> <span class="n">Optional</span><span class="p">[</span><span class="n">Tensor</span><span class="p">]</span> <span class="o">=</span> <span class="kc">None</span><span class="p">,</span>
<span class="n">cross_kv_reuse</span><span class="p">:</span> <span class="n">Optional</span><span class="p">[</span><span class="n">Tensor</span><span class="p">]</span> <span class="o">=</span> <span class="kc">None</span><span class="p">,</span>
<span class="n">prompt_embedding_table</span><span class="p">:</span> <span class="n">Optional</span><span class="p">[</span><span class="n">Tensor</span><span class="p">]</span> <span class="o">=</span> <span class="kc">None</span><span class="p">,</span>
<span class="n">prompt_tasks</span><span class="p">:</span> <span class="n">Optional</span><span class="p">[</span><span class="n">Tensor</span><span class="p">]</span> <span class="o">=</span> <span class="kc">None</span><span class="p">,</span>
<span class="n">prompt_vocab_size</span><span class="p">:</span> <span class="n">Optional</span><span class="p">[</span><span class="n">Tensor</span><span class="p">]</span> <span class="o">=</span> <span class="kc">None</span><span class="p">,</span>
<span class="n">skip_cross_attn_blocks</span><span class="p">:</span> <span class="n">Optional</span><span class="p">[</span><span class="n">Tensor</span><span class="p">]</span> <span class="o">=</span> <span class="kc">None</span><span class="p">,</span>
<span class="p">):</span>
<span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">mapping</span><span class="o">.</span><span class="n">is_first_pp_rank</span><span class="p">():</span>
<span class="k">assert</span> <span class="nb">isinstance</span><span class="p">(</span><span class="n">decoder_input_ids</span><span class="p">,</span> <span class="n">Tensor</span><span class="p">)</span>
<span class="k">else</span><span class="p">:</span>
<span class="k">assert</span> <span class="nb">isinstance</span><span class="p">(</span><span class="n">hidden_states</span><span class="p">,</span> <span class="n">Tensor</span><span class="p">)</span>
<span class="c1"># In PP, layer 0 has ids as inputs, all other layers have hidden_states as inputs</span>
<span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">mapping</span><span class="o">.</span><span class="n">is_first_pp_rank</span><span class="p">():</span>
<span class="n">hidden_states</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">vocab_embedding</span><span class="p">(</span><span class="n">decoder_input_ids</span><span class="p">)</span>
<span class="bp">self</span><span class="o">.</span><span class="n">register_network_output</span><span class="p">(</span><span class="s1">&#39;embedding_layer_output&#39;</span><span class="p">,</span>
<span class="n">hidden_states</span><span class="p">)</span>
<span class="k">else</span><span class="p">:</span>
<span class="n">hidden_states</span> <span class="o">=</span> <span class="n">recv</span><span class="p">(</span><span class="n">hidden_states</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">mapping</span><span class="o">.</span><span class="n">prev_pp_rank</span><span class="p">())</span>
<span class="n">kv_cache_params</span><span class="o">.</span><span class="n">fill_none_tensor_list</span><span class="p">(</span><span class="nb">len</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">layers</span><span class="p">))</span>
<span class="n">full_text_row_masked_out_mask</span> <span class="o">=</span> <span class="n">reduce</span><span class="p">(</span>
<span class="p">(</span><span class="n">attention_mask_params</span><span class="o">.</span><span class="n">cross_attention_mask</span><span class="p">)</span><span class="o">.</span><span class="n">cast</span><span class="p">(</span>
<span class="n">hidden_states</span><span class="o">.</span><span class="n">dtype</span><span class="p">),</span>
<span class="n">trt</span><span class="o">.</span><span class="n">ReduceOperation</span><span class="o">.</span><span class="n">MAX</span><span class="p">,</span>
<span class="n">dim</span><span class="o">=-</span><span class="mi">1</span><span class="p">,</span>
<span class="n">keepdim</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span>
<span class="k">if</span> <span class="n">ADD_DEBUG_TENSOR</span><span class="p">:</span>
<span class="n">full_text_row_masked_out_mask</span><span class="o">.</span><span class="n">mark_output</span><span class="p">(</span>
<span class="s2">&quot;full_text_row_masked_out_mask&quot;</span><span class="p">,</span>
<span class="n">full_text_row_masked_out_mask</span><span class="o">.</span><span class="n">dtype</span><span class="p">)</span>
<span class="n">cross_attention_mask_type</span> <span class="o">=</span> <span class="n">attention_mask_params</span><span class="o">.</span><span class="n">cross_attention_mask</span><span class="o">.</span><span class="n">dtype</span>
<span class="n">attention_mask_params</span><span class="o">.</span><span class="n">cross_attention_mask</span> <span class="o">=</span> <span class="p">(</span>
<span class="n">attention_mask_params</span><span class="o">.</span><span class="n">cross_attention_mask</span><span class="o">.</span><span class="n">cast</span><span class="p">(</span>
<span class="n">full_text_row_masked_out_mask</span><span class="o">.</span><span class="n">dtype</span><span class="p">)</span> <span class="o">*</span>
<span class="n">full_text_row_masked_out_mask</span><span class="p">)</span><span class="o">.</span><span class="n">cast</span><span class="p">(</span><span class="n">cross_attention_mask_type</span><span class="p">)</span>
<span class="n">invert_mask</span> <span class="o">=</span> <span class="mf">1.0</span> <span class="o">-</span> <span class="n">attention_mask_params</span><span class="o">.</span><span class="n">cross_attention_mask</span><span class="o">.</span><span class="n">cast</span><span class="p">(</span>
<span class="n">hidden_states</span><span class="o">.</span><span class="n">dtype</span><span class="p">)</span>
<span class="n">invert_full_text_row_masked_out_mask</span> <span class="o">=</span> <span class="mf">1.0</span> <span class="o">-</span> <span class="n">full_text_row_masked_out_mask</span>
<span class="n">final_mask</span> <span class="o">=</span> <span class="n">invert_mask</span> <span class="o">-</span> <span class="n">invert_full_text_row_masked_out_mask</span>
<span class="n">attention_mask_params</span><span class="o">.</span><span class="n">cross_attention_mask</span> <span class="o">=</span> <span class="n">final_mask</span><span class="o">.</span><span class="n">cast</span><span class="p">(</span>
<span class="n">cross_attention_mask_type</span><span class="p">)</span>
<span class="k">if</span> <span class="n">ADD_DEBUG_TENSOR</span><span class="p">:</span>
<span class="n">attention_mask_params</span><span class="o">.</span><span class="n">cross_attention_mask</span><span class="o">.</span><span class="n">mark_output</span><span class="p">(</span>
<span class="s2">&quot;attention_mask_params.cross_attention_mask&quot;</span><span class="p">,</span>
<span class="n">attention_mask_params</span><span class="o">.</span><span class="n">cross_attention_mask</span><span class="o">.</span><span class="n">dtype</span><span class="p">)</span>
<span class="k">if</span> <span class="n">use_cache</span><span class="p">:</span>
<span class="n">presents</span> <span class="o">=</span> <span class="p">[]</span>
<span class="k">for</span> <span class="n">i</span><span class="p">,</span> <span class="p">(</span><span class="n">decoder_layer</span><span class="p">,</span> <span class="n">past</span><span class="p">)</span> <span class="ow">in</span> <span class="nb">enumerate</span><span class="p">(</span>
<span class="nb">zip</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">layers</span><span class="p">,</span> <span class="n">kv_cache_params</span><span class="o">.</span><span class="n">past_key_value</span><span class="p">)):</span>
<span class="n">lora_layer_params</span> <span class="o">=</span> <span class="kc">None</span>
<span class="k">if</span> <span class="n">lora_params</span> <span class="ow">is</span> <span class="ow">not</span> <span class="kc">None</span> <span class="ow">and</span> <span class="n">lora_params</span><span class="o">.</span><span class="n">lora_ranks</span> <span class="ow">is</span> <span class="ow">not</span> <span class="kc">None</span><span class="p">:</span>
<span class="n">lora_layer_params</span> <span class="o">=</span> <span class="n">lora_params</span><span class="o">.</span><span class="n">get_layer_params</span><span class="p">(</span><span class="n">i</span><span class="p">)</span>
<span class="n">hidden_states</span> <span class="o">=</span> <span class="n">decoder_layer</span><span class="p">(</span>
<span class="n">hidden_states</span><span class="p">,</span>
<span class="n">encoder_output</span><span class="o">=</span><span class="n">encoder_output</span><span class="p">,</span>
<span class="n">attention_mask_params</span><span class="o">=</span><span class="n">attention_mask_params</span><span class="p">,</span>
<span class="n">use_cache</span><span class="o">=</span><span class="n">use_cache</span><span class="p">,</span>
<span class="n">kv_cache_params</span><span class="o">=</span><span class="n">KeyValueCacheParams</span><span class="p">(</span>
<span class="n">past_key_value</span><span class="o">=</span><span class="n">past</span><span class="p">,</span>
<span class="n">host_past_key_value_lengths</span><span class="o">=</span><span class="n">kv_cache_params</span><span class="o">.</span>
<span class="n">host_past_key_value_lengths</span><span class="p">,</span>
<span class="n">host_max_attention_window_sizes</span><span class="o">=</span><span class="n">kv_cache_params</span><span class="o">.</span>
<span class="n">host_max_attention_window_sizes</span><span class="p">,</span>
<span class="n">host_sink_token_length</span><span class="o">=</span><span class="n">kv_cache_params</span><span class="o">.</span>
<span class="n">host_sink_token_length</span><span class="p">,</span>
<span class="n">cache_indirection</span><span class="o">=</span><span class="n">kv_cache_params</span><span class="o">.</span><span class="n">cache_indirection</span><span class="p">,</span>
<span class="n">kv_cache_block_offsets</span><span class="o">=</span><span class="n">kv_cache_params</span><span class="o">.</span>
<span class="n">kv_cache_block_offsets</span><span class="p">,</span>
<span class="n">host_kv_cache_block_offsets</span><span class="o">=</span><span class="n">kv_cache_params</span><span class="o">.</span>
<span class="n">host_cross_kv_cache_block_offsets</span><span class="p">,</span>
<span class="n">host_kv_cache_pool_pointers</span><span class="o">=</span><span class="n">kv_cache_params</span><span class="o">.</span>
<span class="n">host_kv_cache_pool_pointers</span><span class="p">,</span>
<span class="n">host_kv_cache_pool_mapping</span><span class="o">=</span><span class="n">kv_cache_params</span><span class="o">.</span>
<span class="n">host_kv_cache_pool_mapping</span><span class="p">,</span>
<span class="n">cross_kv_cache_block_offsets</span><span class="o">=</span><span class="n">kv_cache_params</span><span class="o">.</span>
<span class="n">cross_kv_cache_block_offsets</span><span class="p">,</span>
<span class="n">host_cross_kv_cache_block_offsets</span><span class="o">=</span><span class="n">kv_cache_params</span><span class="o">.</span>
<span class="n">host_cross_kv_cache_block_offsets</span><span class="p">,</span>
<span class="n">host_cross_kv_cache_pool_pointers</span><span class="o">=</span><span class="n">kv_cache_params</span><span class="o">.</span>
<span class="n">host_cross_kv_cache_pool_pointers</span><span class="p">,</span>
<span class="n">host_cross_kv_cache_pool_mapping</span><span class="o">=</span><span class="n">kv_cache_params</span><span class="o">.</span>
<span class="n">host_cross_kv_cache_pool_mapping</span><span class="p">,</span>
<span class="p">),</span>
<span class="n">skip_cross_attn_blocks</span><span class="o">=</span><span class="n">skip_cross_attn_blocks</span> <span class="k">if</span> <span class="nb">isinstance</span><span class="p">(</span>
<span class="n">decoder_layer</span><span class="p">,</span> <span class="n">CrossAttentionTransformerBlock</span><span class="p">)</span> <span class="k">else</span> <span class="kc">None</span><span class="p">,</span>
<span class="n">attention_params</span><span class="o">=</span><span class="n">attention_params</span><span class="p">,</span>
<span class="n">lora_layer_params</span><span class="o">=</span><span class="n">lora_layer_params</span><span class="p">,</span>
<span class="n">cross_kv_cache_gen</span><span class="o">=</span><span class="n">cross_kv_cache_gen</span><span class="p">,</span>
<span class="n">cross_kv_reuse</span><span class="o">=</span><span class="n">cross_kv_reuse</span><span class="p">,</span>
<span class="n">full_text_row_masked_out_mask</span><span class="o">=</span><span class="n">full_text_row_masked_out_mask</span><span class="p">,</span>
<span class="p">)</span>
<span class="k">if</span> <span class="n">use_cache</span><span class="p">:</span>
<span class="n">present</span> <span class="o">=</span> <span class="n">hidden_states</span><span class="p">[</span><span class="mi">1</span><span class="p">]</span>
<span class="n">presents</span><span class="o">.</span><span class="n">append</span><span class="p">((</span><span class="n">present</span><span class="p">))</span>
<span class="n">hidden_states</span> <span class="o">=</span> <span class="n">hidden_states</span><span class="p">[</span><span class="mi">0</span><span class="p">]</span>
<span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">mapping</span><span class="o">.</span><span class="n">is_last_pp_rank</span><span class="p">():</span>
<span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">ln_f</span><span class="p">:</span>
<span class="n">hidden_states</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">ln_f</span><span class="p">(</span><span class="n">hidden_states</span><span class="p">)</span>
<span class="k">else</span><span class="p">:</span>
<span class="n">hidden_states</span> <span class="o">=</span> <span class="n">send</span><span class="p">(</span><span class="n">hidden_states</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">mapping</span><span class="o">.</span><span class="n">next_pp_rank</span><span class="p">())</span>
<span class="k">if</span> <span class="n">use_cache</span><span class="p">:</span>
<span class="k">return</span> <span class="p">(</span><span class="n">hidden_states</span><span class="p">,</span> <span class="nb">tuple</span><span class="p">(</span><span class="n">presents</span><span class="p">))</span>
<span class="k">return</span> <span class="n">hidden_states</span>
<span class="k">def</span><span class="w"> </span><span class="nf">precompute_relative_attention_bias</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">build_config</span><span class="p">):</span>
<span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">relative_attention</span> <span class="ow">and</span> <span class="ow">not</span> <span class="bp">self</span><span class="o">.</span><span class="n">use_implicit_relative_attention</span><span class="p">:</span>
<span class="n">relative_attention_bias_builder</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">ops</span><span class="o">.</span><span class="n">tensorrt_llm</span><span class="o">.</span><span class="n">relative_attention_bias</span>
<span class="n">rel_attn_precomputed</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">zeros</span><span class="p">(</span>
<span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">num_attention_heads</span> <span class="o">//</span> <span class="bp">self</span><span class="o">.</span><span class="n">mapping</span><span class="o">.</span><span class="n">tp_size</span><span class="p">,</span>
<span class="n">build_config</span><span class="o">.</span><span class="n">max_seq_len</span> <span class="o">+</span> <span class="mi">1</span><span class="p">,</span> <span class="n">build_config</span><span class="o">.</span><span class="n">max_seq_len</span> <span class="o">+</span> <span class="mi">1</span><span class="p">),</span>
<span class="n">dtype</span><span class="o">=</span><span class="n">str_dtype_to_torch</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">dtype</span><span class="p">),</span>
<span class="n">device</span><span class="o">=</span><span class="s1">&#39;cuda&#39;</span><span class="p">)</span>
<span class="n">rel_attn_table</span> <span class="o">=</span> <span class="n">numpy_to_torch</span><span class="p">(</span>
<span class="bp">self</span><span class="o">.</span><span class="n">rel_attn_table</span><span class="o">.</span><span class="n">raw_value</span><span class="p">)</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="s1">&#39;cuda&#39;</span><span class="p">)</span>
<span class="n">relative_attention_bias_builder</span><span class="p">(</span>
<span class="n">rel_attn_precomputed</span><span class="p">,</span>
<span class="n">rel_attn_table</span><span class="p">,</span>
<span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">num_attention_heads</span> <span class="o">//</span> <span class="bp">self</span><span class="o">.</span><span class="n">mapping</span><span class="o">.</span><span class="n">tp_size</span><span class="p">,</span>
<span class="n">build_config</span><span class="o">.</span><span class="n">max_seq_len</span><span class="p">,</span>
<span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">num_buckets</span><span class="p">,</span>
<span class="kc">False</span><span class="p">,</span>
<span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">max_distance</span><span class="p">,</span>
<span class="p">)</span>
<span class="k">for</span> <span class="n">layer_idx</span> <span class="ow">in</span> <span class="nb">range</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">num_layers</span><span class="p">):</span>
<span class="bp">self</span><span class="o">.</span><span class="n">layers</span><span class="p">[</span><span class="n">layer_idx</span><span class="p">]</span><span class="o">.</span><span class="n">self_attention</span><span class="o">.</span><span class="n">set_rel_attn_table</span><span class="p">(</span>
<span class="n">build_config</span><span class="o">.</span><span class="n">max_seq_len</span><span class="p">,</span> <span class="n">rel_attn_precomputed</span><span class="p">)</span>
<span class="c1"># TODO try to inherit the DecoderModelForCausalLM</span>
<div class="viewcode-block" id="MLLaMAForCausalLM">
<a class="viewcode-back" href="../../../../python-api/tensorrt_llm.models.html#tensorrt_llm.models.MLLaMAForCausalLM">[docs]</a>
<span class="k">class</span><span class="w"> </span><span class="nc">MLLaMAForCausalLM</span><span class="p">(</span><span class="n">PretrainedModel</span><span class="p">):</span>
<span class="n">config_class</span> <span class="o">=</span> <span class="n">MLLaMAConfig</span>
<span class="k">def</span><span class="w"> </span><span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">config</span><span class="p">:</span> <span class="n">MLLaMAConfig</span><span class="p">):</span>
<span class="nb">super</span><span class="p">()</span><span class="o">.</span><span class="fm">__init__</span><span class="p">(</span><span class="n">config</span><span class="p">)</span>
<span class="n">Attention</span><span class="o">.</span><span class="n">create_attention_const_params</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">config</span><span class="p">)</span>
<span class="bp">self</span><span class="o">.</span><span class="n">position_embedding_type</span> <span class="o">=</span> <span class="n">config</span><span class="o">.</span><span class="n">position_embedding_type</span>
<span class="bp">self</span><span class="o">.</span><span class="n">transformer</span> <span class="o">=</span> <span class="n">MLLaMAModel</span><span class="p">(</span><span class="n">config</span><span class="p">)</span>
<span class="bp">self</span><span class="o">.</span><span class="n">mapping</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">mapping</span>
<span class="bp">self</span><span class="o">.</span><span class="n">has_model_final_layernorm</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">has_model_final_layernorm</span>
<span class="bp">self</span><span class="o">.</span><span class="n">_dtype</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">dtype</span>
<span class="bp">self</span><span class="o">.</span><span class="n">_kv_dtype</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">_dtype</span>
<span class="bp">self</span><span class="o">.</span><span class="n">_logits_dtype</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">logits_dtype</span>
<span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">mapping</span><span class="o">.</span><span class="n">is_last_pp_rank</span><span class="p">():</span>
<span class="bp">self</span><span class="o">.</span><span class="n">lm_head</span> <span class="o">=</span> <span class="n">ColumnLinear</span><span class="p">(</span>
<span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">hidden_size</span><span class="p">,</span>
<span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">vocab_size</span><span class="p">,</span>
<span class="n">bias</span><span class="o">=</span><span class="kc">False</span> <span class="k">if</span> <span class="ow">not</span> <span class="nb">hasattr</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="p">,</span> <span class="s2">&quot;has_lm_head_bias&quot;</span><span class="p">)</span> <span class="k">else</span>
<span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">has_lm_head_bias</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">dtype</span><span class="p">,</span>
<span class="n">tp_group</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">mapping</span><span class="o">.</span><span class="n">tp_group</span><span class="p">,</span>
<span class="n">tp_size</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">mapping</span><span class="o">.</span><span class="n">tp_size</span><span class="p">,</span>
<span class="n">gather_output</span><span class="o">=</span><span class="kc">True</span><span class="p">,</span>
<span class="p">)</span>
<span class="bp">self</span><span class="o">.</span><span class="n">trtllm_modules_to_hf_modules</span> <span class="o">=</span> <span class="p">{</span>
<span class="o">**</span><span class="n">get_default_trtllm_modules_to_hf_modules</span><span class="p">(),</span>
<span class="s2">&quot;attn_q&quot;</span><span class="p">:</span> <span class="s2">&quot;self_attn.q_proj&quot;</span><span class="p">,</span>
<span class="s2">&quot;attn_k&quot;</span><span class="p">:</span> <span class="s2">&quot;self_attn.k_proj&quot;</span><span class="p">,</span>
<span class="s2">&quot;attn_v&quot;</span><span class="p">:</span> <span class="s2">&quot;self_attn.v_proj&quot;</span><span class="p">,</span>
<span class="s2">&quot;attn_dense&quot;</span><span class="p">:</span> <span class="s2">&quot;self_attn.o_proj&quot;</span><span class="p">,</span>
<span class="s2">&quot;cross_attn_q&quot;</span><span class="p">:</span> <span class="s2">&quot;encoder_attn.q_proj&quot;</span><span class="p">,</span>
<span class="s2">&quot;cross_attn_k&quot;</span><span class="p">:</span> <span class="s2">&quot;encoder_attn.k_proj&quot;</span><span class="p">,</span>
<span class="s2">&quot;cross_attn_v&quot;</span><span class="p">:</span> <span class="s2">&quot;encoder_attn.v_proj&quot;</span><span class="p">,</span>
<span class="s2">&quot;cross_attn_dense&quot;</span><span class="p">:</span> <span class="s2">&quot;encoder_attn.o_proj&quot;</span><span class="p">,</span>
<span class="p">}</span>
<div class="viewcode-block" id="MLLaMAForCausalLM.forward">
<a class="viewcode-back" href="../../../../python-api/tensorrt_llm.models.html#tensorrt_llm.models.MLLaMAForCausalLM.forward">[docs]</a>
<span class="k">def</span><span class="w"> </span><span class="nf">forward</span><span class="p">(</span>
<span class="bp">self</span><span class="p">,</span>
<span class="n">decoder_input_ids</span><span class="p">:</span> <span class="n">Tensor</span><span class="p">,</span>
<span class="n">encoder_output</span><span class="p">:</span> <span class="n">Tensor</span><span class="p">,</span>
<span class="n">use_cache</span><span class="o">=</span><span class="kc">False</span><span class="p">,</span>
<span class="n">attention_mask_params</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span>
<span class="n">last_token_ids</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span>
<span class="n">kv_cache_params</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span>
<span class="n">attention_params</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span>
<span class="n">hidden_states</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span>
<span class="n">lora_params</span><span class="p">:</span> <span class="n">LoraParams</span> <span class="o">=</span> <span class="kc">None</span><span class="p">,</span>
<span class="n">cross_kv_cache_gen</span><span class="p">:</span> <span class="n">Optional</span><span class="p">[</span><span class="n">Tensor</span><span class="p">]</span> <span class="o">=</span> <span class="kc">None</span><span class="p">,</span>
<span class="n">cross_kv_reuse</span><span class="p">:</span> <span class="n">Optional</span><span class="p">[</span><span class="n">Tensor</span><span class="p">]</span> <span class="o">=</span> <span class="kc">None</span><span class="p">,</span>
<span class="n">prompt_embedding_table</span><span class="p">:</span> <span class="n">Optional</span><span class="p">[</span><span class="n">Tensor</span><span class="p">]</span> <span class="o">=</span> <span class="kc">None</span><span class="p">,</span>
<span class="n">prompt_tasks</span><span class="p">:</span> <span class="n">Optional</span><span class="p">[</span><span class="n">Tensor</span><span class="p">]</span> <span class="o">=</span> <span class="kc">None</span><span class="p">,</span>
<span class="n">prompt_vocab_size</span><span class="p">:</span> <span class="n">Optional</span><span class="p">[</span><span class="n">Tensor</span><span class="p">]</span> <span class="o">=</span> <span class="kc">None</span><span class="p">,</span>
<span class="n">skip_cross_attn_blocks</span><span class="p">:</span> <span class="n">Optional</span><span class="p">[</span><span class="n">Tensor</span><span class="p">]</span> <span class="o">=</span> <span class="kc">None</span><span class="p">,</span>
<span class="p">):</span>
<span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">mapping</span><span class="o">.</span><span class="n">is_first_pp_rank</span><span class="p">():</span>
<span class="k">assert</span> <span class="nb">isinstance</span><span class="p">(</span><span class="n">decoder_input_ids</span><span class="p">,</span> <span class="n">Tensor</span><span class="p">)</span>
<span class="k">else</span><span class="p">:</span>
<span class="k">assert</span> <span class="nb">isinstance</span><span class="p">(</span><span class="n">hidden_states</span><span class="p">,</span> <span class="n">Tensor</span><span class="p">)</span>
<span class="n">attention_params</span> <span class="o">=</span> <span class="n">Attention</span><span class="o">.</span><span class="n">fill_attention_params</span><span class="p">(</span>
<span class="bp">self</span><span class="p">,</span> <span class="n">attention_params</span><span class="p">)</span>
<span class="n">hidden_states</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">transformer</span><span class="p">(</span>
<span class="n">decoder_input_ids</span><span class="o">=</span><span class="n">decoder_input_ids</span><span class="p">,</span>
<span class="n">encoder_output</span><span class="o">=</span><span class="n">encoder_output</span><span class="p">,</span>
<span class="n">use_cache</span><span class="o">=</span><span class="n">use_cache</span><span class="p">,</span>
<span class="n">attention_mask_params</span><span class="o">=</span><span class="n">attention_mask_params</span><span class="p">,</span>
<span class="n">last_token_ids</span><span class="o">=</span><span class="n">last_token_ids</span><span class="p">,</span>
<span class="n">kv_cache_params</span><span class="o">=</span><span class="n">kv_cache_params</span><span class="p">,</span>
<span class="n">attention_params</span><span class="o">=</span><span class="n">attention_params</span><span class="p">,</span>
<span class="n">hidden_states</span><span class="o">=</span><span class="n">hidden_states</span><span class="p">,</span>
<span class="n">lora_params</span><span class="o">=</span><span class="n">lora_params</span><span class="p">,</span>
<span class="n">cross_kv_cache_gen</span><span class="o">=</span><span class="n">cross_kv_cache_gen</span><span class="p">,</span>
<span class="n">cross_kv_reuse</span><span class="o">=</span><span class="n">cross_kv_reuse</span><span class="p">,</span>
<span class="n">prompt_embedding_table</span><span class="o">=</span><span class="n">prompt_embedding_table</span><span class="p">,</span>
<span class="n">prompt_tasks</span><span class="o">=</span><span class="n">prompt_tasks</span><span class="p">,</span>
<span class="n">prompt_vocab_size</span><span class="o">=</span><span class="n">prompt_vocab_size</span><span class="p">,</span>
<span class="n">skip_cross_attn_blocks</span><span class="o">=</span><span class="n">skip_cross_attn_blocks</span><span class="p">,</span>
<span class="p">)</span>
<span class="k">if</span> <span class="n">use_cache</span><span class="p">:</span>
<span class="n">hidden_states</span><span class="p">,</span> <span class="n">presents</span> <span class="o">=</span> <span class="n">hidden_states</span>
<span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">mapping</span><span class="o">.</span><span class="n">is_last_pp_rank</span><span class="p">():</span>
<span class="c1"># [bs, seq, hidden_size] or [num_tokens, hidden_size] -&gt; [bs, hidden_size]</span>
<span class="n">hidden_states</span> <span class="o">=</span> <span class="n">gather_last_token_logits</span><span class="p">(</span>
<span class="n">hidden_states</span><span class="p">,</span> <span class="n">last_token_ids</span><span class="p">,</span>
<span class="n">default_net</span><span class="p">()</span><span class="o">.</span><span class="n">plugin_config</span><span class="o">.</span><span class="n">remove_input_padding</span><span class="p">)</span>
<span class="c1"># [bs, hidden_size] -&gt; [bs, vocab_size]</span>
<span class="n">lm_logits</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">lm_head</span><span class="p">(</span><span class="n">hidden_states</span><span class="p">)</span>
<span class="n">lm_logits</span><span class="o">.</span><span class="n">mark_output</span><span class="p">(</span><span class="sa">f</span><span class="s1">&#39;logits&#39;</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">_logits_dtype</span><span class="p">)</span>
<span class="k">else</span><span class="p">:</span>
<span class="n">hidden_states</span> <span class="o">=</span> <span class="n">send</span><span class="p">(</span><span class="n">hidden_states</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">mapping</span><span class="o">.</span><span class="n">next_pp_rank</span><span class="p">())</span>
<span class="n">hidden_states</span><span class="o">.</span><span class="n">mark_output</span><span class="p">(</span><span class="sa">f</span><span class="s1">&#39;hidden_states_output&#39;</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">_dtype</span><span class="p">)</span>
<span class="k">if</span> <span class="n">use_cache</span> <span class="ow">and</span> <span class="n">default_net</span><span class="p">()</span><span class="o">.</span><span class="n">plugin_config</span><span class="o">.</span><span class="n">paged_kv_cache</span> <span class="o">==</span> <span class="kc">False</span><span class="p">:</span>
<span class="k">for</span> <span class="n">i</span><span class="p">,</span> <span class="n">present</span> <span class="ow">in</span> <span class="nb">zip</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">mapping</span><span class="o">.</span><span class="n">pp_layers</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">total_num_layers</span><span class="p">),</span>
<span class="n">presents</span><span class="p">):</span>
<span class="n">present</span><span class="p">[</span><span class="mi">0</span><span class="p">]</span><span class="o">.</span><span class="n">mark_output</span><span class="p">(</span><span class="sa">f</span><span class="s1">&#39;present_key_value_</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">&#39;</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">_kv_dtype</span><span class="p">)</span>
<span class="k">if</span> <span class="n">default_net</span><span class="p">()</span><span class="o">.</span><span class="n">plugin_config</span><span class="o">.</span><span class="n">gpt_attention_plugin</span><span class="p">:</span>
<span class="n">present</span><span class="p">[</span><span class="mi">1</span><span class="p">]</span><span class="o">.</span><span class="n">mark_output</span><span class="p">(</span><span class="sa">f</span><span class="s1">&#39;cross_present_key_value_</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">&#39;</span><span class="p">,</span>
<span class="bp">self</span><span class="o">.</span><span class="n">_kv_dtype</span><span class="p">)</span>
<span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">mapping</span><span class="o">.</span><span class="n">is_last_pp_rank</span><span class="p">():</span>
<span class="k">return</span> <span class="p">(</span><span class="n">lm_logits</span><span class="p">,</span> <span class="nb">tuple</span><span class="p">(</span><span class="n">presents</span><span class="p">))</span>
<span class="k">return</span> <span class="p">(</span><span class="n">hidden_states</span><span class="p">,</span> <span class="nb">tuple</span><span class="p">(</span><span class="n">presents</span><span class="p">))</span>
<span class="k">else</span><span class="p">:</span>
<span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">mapping</span><span class="o">.</span><span class="n">is_last_pp_rank</span><span class="p">():</span>
<span class="k">return</span> <span class="n">lm_logits</span>
<span class="k">return</span> <span class="n">hidden_states</span></div>
<div class="viewcode-block" id="MLLaMAForCausalLM.prepare_inputs">
<a class="viewcode-back" href="../../../../python-api/tensorrt_llm.models.html#tensorrt_llm.models.MLLaMAForCausalLM.prepare_inputs">[docs]</a>
<span class="k">def</span><span class="w"> </span><span class="nf">prepare_inputs</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span>
<span class="n">max_batch_size</span><span class="p">,</span>
<span class="n">max_beam_width</span><span class="p">,</span>
<span class="n">max_decoder_input_len</span><span class="p">,</span>
<span class="n">max_seq_len</span><span class="p">,</span>
<span class="n">max_encoder_input_len</span><span class="p">,</span>
<span class="n">gather_context_logits</span><span class="p">:</span> <span class="nb">bool</span> <span class="o">=</span> <span class="kc">False</span><span class="p">,</span>
<span class="n">gather_generation_logits</span><span class="p">:</span> <span class="nb">bool</span> <span class="o">=</span> <span class="kc">False</span><span class="p">,</span>
<span class="n">lora_target_modules</span><span class="p">:</span> <span class="n">List</span><span class="p">[</span><span class="nb">str</span><span class="p">]</span> <span class="o">=</span> <span class="kc">None</span><span class="p">,</span>
<span class="n">prompt_embedding_table_size</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">0</span><span class="p">,</span>
<span class="n">use_cache</span><span class="o">=</span><span class="kc">True</span><span class="p">,</span>
<span class="o">*</span><span class="n">args</span><span class="p">,</span>
<span class="o">**</span><span class="n">kwargs</span><span class="p">):</span>
<span class="w"> </span><span class="sd">&#39;&#39;&#39;@brief: Prepare inputs Tensors for the model, the given sizes are used to determine the</span>
<span class="sd"> ranges of the dimensions of when using TRT dynamic shapes.</span>
<span class="sd"> @return: a list contains values which can be fed into the self.forward()</span>
<span class="sd"> &#39;&#39;&#39;</span>
<span class="c1"># Prepare inputs</span>
<span class="n">max_output_len</span> <span class="o">=</span> <span class="n">max_decoder_input_len</span> <span class="o">+</span> <span class="n">max_seq_len</span>
<span class="n">head_size</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">transformer</span><span class="o">.</span><span class="n">head_size</span>
<span class="n">num_kv_heads</span> <span class="o">=</span> <span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">transformer</span><span class="o">.</span><span class="n">num_kv_heads</span> <span class="o">+</span> <span class="bp">self</span><span class="o">.</span><span class="n">mapping</span><span class="o">.</span><span class="n">tp_size</span> <span class="o">-</span>
<span class="mi">1</span><span class="p">)</span> <span class="o">//</span> <span class="bp">self</span><span class="o">.</span><span class="n">mapping</span><span class="o">.</span><span class="n">tp_size</span>
<span class="n">encoder_head_size</span> <span class="o">=</span> <span class="n">head_size</span>
<span class="n">encoder_num_kv_heads</span> <span class="o">=</span> <span class="n">num_kv_heads</span>
<span class="n">bb_range</span> <span class="o">=</span> <span class="p">[</span>
<span class="mi">1</span><span class="p">,</span> <span class="p">(</span><span class="n">max_batch_size</span> <span class="o">*</span> <span class="n">max_beam_width</span> <span class="o">+</span> <span class="mi">1</span><span class="p">)</span> <span class="o">//</span> <span class="mi">2</span><span class="p">,</span>
<span class="n">max_batch_size</span> <span class="o">*</span> <span class="n">max_beam_width</span>
<span class="p">]</span>
<span class="n">bs_range</span> <span class="o">=</span> <span class="p">[</span><span class="mi">1</span><span class="p">,</span> <span class="p">(</span><span class="n">max_batch_size</span> <span class="o">+</span> <span class="mi">1</span><span class="p">)</span> <span class="o">//</span> <span class="mi">2</span><span class="p">,</span> <span class="n">max_batch_size</span><span class="p">]</span>
<span class="n">beam_width_range</span> <span class="o">=</span> <span class="p">[</span><span class="mi">1</span><span class="p">,</span> <span class="p">(</span><span class="n">max_beam_width</span> <span class="o">+</span> <span class="mi">1</span><span class="p">)</span> <span class="o">//</span> <span class="mi">2</span><span class="p">,</span> <span class="n">max_beam_width</span><span class="p">]</span>
<span class="n">inlen_range</span> <span class="o">=</span> <span class="p">[</span>
<span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="n">max_decoder_input_len</span>
<span class="p">]</span> <span class="c1"># context phase &gt;= 1 (if forced_input_ids), generation phase = 1</span>
<span class="n">encoder_inlen_range</span> <span class="o">=</span> <span class="p">[</span>
<span class="mi">1</span><span class="p">,</span> <span class="p">(</span><span class="n">max_encoder_input_len</span> <span class="o">+</span> <span class="mi">1</span><span class="p">)</span> <span class="o">//</span> <span class="mi">2</span><span class="p">,</span> <span class="n">max_encoder_input_len</span>
<span class="p">]</span>
<span class="n">mask_len_range</span> <span class="o">=</span> <span class="p">[</span><span class="mi">1</span><span class="p">,</span> <span class="p">(</span><span class="n">max_output_len</span> <span class="o">+</span> <span class="mi">1</span><span class="p">)</span> <span class="o">//</span> <span class="mi">2</span> <span class="o">+</span> <span class="mi">1</span><span class="p">,</span> <span class="n">max_output_len</span> <span class="o">+</span> <span class="mi">1</span><span class="p">]</span>
<span class="n">max_output_len_range</span> <span class="o">=</span> <span class="p">[</span><span class="mi">0</span><span class="p">,</span> <span class="p">(</span><span class="n">max_output_len</span> <span class="o">+</span> <span class="mi">1</span><span class="p">)</span> <span class="o">//</span> <span class="mi">2</span><span class="p">,</span> <span class="n">max_output_len</span><span class="p">]</span>
<span class="n">encoder_num_tokens_range</span> <span class="o">=</span> <span class="p">[</span>
<span class="mi">0</span><span class="p">,</span> <span class="c1"># 0 for generation phase, &gt;0 for context phase</span>
<span class="p">(</span><span class="n">max_encoder_input_len</span> <span class="o">*</span> <span class="n">max_batch_size</span> <span class="o">+</span> <span class="mi">1</span><span class="p">)</span> <span class="o">//</span> <span class="mi">2</span><span class="p">,</span>
<span class="n">max_encoder_input_len</span> <span class="o">*</span> <span class="n">max_batch_size</span><span class="p">,</span>
<span class="p">]</span>
<span class="n">decoder_num_tokens_range</span> <span class="o">=</span> <span class="p">[</span>
<span class="mi">1</span><span class="p">,</span>
<span class="n">max_batch_size</span> <span class="o">*</span> <span class="n">max_beam_width</span><span class="p">,</span>
<span class="nb">max</span><span class="p">(</span><span class="n">max_decoder_input_len</span> <span class="o">*</span> <span class="n">max_batch_size</span><span class="p">,</span>
<span class="n">max_beam_width</span> <span class="o">*</span> <span class="n">max_batch_size</span><span class="p">),</span>
<span class="p">]</span>
<span class="c1"># No enable_two_optimization_profiles support yet</span>
<span class="n">encoder_input_len_range</span> <span class="o">=</span> <span class="p">[</span>
<span class="mi">0</span><span class="p">,</span> <span class="c1"># 0 for generation phase, &gt;0 for context phase</span>
<span class="p">(</span><span class="n">max_encoder_input_len</span> <span class="o">+</span> <span class="mi">1</span><span class="p">)</span> <span class="o">//</span> <span class="mi">2</span><span class="p">,</span>
<span class="n">max_encoder_input_len</span>
<span class="p">]</span>
<span class="c1"># pack masks into bits (store as int32).</span>
<span class="n">max_cross_packed_mask_dim0</span> <span class="o">=</span> <span class="n">max_batch_size</span> <span class="o">*</span> <span class="p">(</span>
<span class="p">(</span><span class="n">max_decoder_input_len</span> <span class="o">+</span> <span class="mi">128</span> <span class="o">-</span> <span class="mi">1</span><span class="p">)</span> <span class="o">//</span> <span class="mi">128</span><span class="p">)</span> <span class="o">*</span> <span class="mi">128</span>
<span class="n">max_cross_packed_mask_dim1</span> <span class="o">=</span> <span class="p">(</span>
<span class="p">(</span><span class="n">max_encoder_input_len</span> <span class="o">+</span> <span class="mi">256</span> <span class="o">-</span> <span class="mi">1</span><span class="p">)</span> <span class="o">//</span> <span class="mi">256</span><span class="p">)</span> <span class="o">*</span> <span class="mi">256</span> <span class="o">//</span> <span class="mi">32</span>
<span class="n">cross_packed_mask_dim0_range</span> <span class="o">=</span> <span class="p">[</span>
<span class="mi">1</span><span class="p">,</span> <span class="p">(</span><span class="n">max_cross_packed_mask_dim0</span> <span class="o">+</span> <span class="mi">1</span><span class="p">)</span> <span class="o">//</span> <span class="mi">2</span><span class="p">,</span> <span class="n">max_cross_packed_mask_dim0</span>
<span class="p">]</span>
<span class="n">cross_packed_mask_dim1_range</span> <span class="o">=</span> <span class="p">[</span>
<span class="mi">0</span><span class="p">,</span> <span class="c1"># 0 for generation phase, &gt;0 for context phase</span>
<span class="p">(</span><span class="n">max_cross_packed_mask_dim1</span> <span class="o">+</span> <span class="mi">1</span><span class="p">)</span> <span class="o">//</span> <span class="mi">2</span><span class="p">,</span>
<span class="n">max_cross_packed_mask_dim1</span>
<span class="p">]</span>
<span class="n">past_key_value</span> <span class="o">=</span> <span class="p">[]</span>
<span class="n">sequence_length</span> <span class="o">=</span> <span class="kc">None</span>
<span class="n">host_past_key_value_lengths</span> <span class="o">=</span> <span class="kc">None</span>
<span class="n">attention_mask_params</span> <span class="o">=</span> <span class="n">AttentionMaskParams</span><span class="p">()</span>
<span class="n">use_gpt_attention_plugin</span> <span class="o">=</span> <span class="n">default_net</span><span class="p">(</span>
<span class="p">)</span><span class="o">.</span><span class="n">plugin_config</span><span class="o">.</span><span class="n">gpt_attention_plugin</span>
<span class="n">remove_input_padding</span> <span class="o">=</span> <span class="n">default_net</span><span class="p">()</span><span class="o">.</span><span class="n">plugin_config</span><span class="o">.</span><span class="n">remove_input_padding</span>
<span class="n">paged_kv_cache</span> <span class="o">=</span> <span class="n">default_net</span><span class="p">()</span><span class="o">.</span><span class="n">plugin_config</span><span class="o">.</span><span class="n">paged_kv_cache</span>
<span class="n">tokens_per_block</span> <span class="o">=</span> <span class="n">default_net</span><span class="p">()</span><span class="o">.</span><span class="n">plugin_config</span><span class="o">.</span><span class="n">tokens_per_block</span>
<span class="n">use_lora_plugin</span> <span class="o">=</span> <span class="n">default_net</span><span class="p">()</span><span class="o">.</span><span class="n">plugin_config</span><span class="o">.</span><span class="n">lora_plugin</span>
<span class="n">kv_cache_type</span> <span class="o">=</span> <span class="kc">None</span>
<span class="k">if</span> <span class="ow">not</span> <span class="n">use_cache</span><span class="p">:</span>
<span class="n">kv_cache_type</span> <span class="o">=</span> <span class="n">KVCacheType</span><span class="o">.</span><span class="n">DISABLED</span>
<span class="k">else</span><span class="p">:</span>
<span class="k">if</span> <span class="n">paged_kv_cache</span><span class="p">:</span>
<span class="n">kv_cache_type</span> <span class="o">=</span> <span class="n">KVCacheType</span><span class="o">.</span><span class="n">PAGED</span>
<span class="k">else</span><span class="p">:</span>
<span class="n">kv_cache_type</span> <span class="o">=</span> <span class="n">KVCacheType</span><span class="o">.</span><span class="n">CONTINUOUS</span>
<span class="n">input_ids</span><span class="p">,</span> <span class="n">hidden_states</span> <span class="o">=</span> <span class="kc">None</span><span class="p">,</span> <span class="kc">None</span>
<span class="k">if</span> <span class="n">remove_input_padding</span><span class="p">:</span>
<span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">mapping</span><span class="o">.</span><span class="n">is_first_pp_rank</span><span class="p">():</span>
<span class="n">input_ids</span> <span class="o">=</span> <span class="n">Tensor</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s1">&#39;input_ids&#39;</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="n">trt</span><span class="o">.</span><span class="n">int32</span><span class="p">,</span>
<span class="n">shape</span><span class="o">=</span><span class="p">[</span><span class="o">-</span><span class="mi">1</span><span class="p">],</span>
<span class="n">dim_range</span><span class="o">=</span><span class="n">OrderedDict</span><span class="p">([</span>
<span class="p">(</span><span class="s1">&#39;decoder_num_tokens&#39;</span><span class="p">,</span>
<span class="p">[</span><span class="n">decoder_num_tokens_range</span><span class="p">]),</span>
<span class="p">]))</span>
<span class="k">else</span><span class="p">:</span>
<span class="n">hidden_states</span> <span class="o">=</span> <span class="n">Tensor</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s1">&#39;hidden_states_input&#39;</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">_dtype</span><span class="p">,</span>
<span class="n">shape</span><span class="o">=</span><span class="p">[</span><span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">hidden_size</span><span class="p">],</span>
<span class="n">dim_range</span><span class="o">=</span><span class="n">OrderedDict</span><span class="p">([</span>
<span class="p">(</span><span class="s1">&#39;decoder_num_tokens&#39;</span><span class="p">,</span>
<span class="p">[</span><span class="n">decoder_num_tokens_range</span><span class="p">]),</span>
<span class="p">(</span><span class="s1">&#39;hidden_size&#39;</span><span class="p">,</span> <span class="p">[</span><span class="bp">self</span><span class="o">.</span><span class="n">hidden_size</span><span class="p">]),</span>
<span class="p">]))</span>
<span class="k">else</span><span class="p">:</span>
<span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">mapping</span><span class="o">.</span><span class="n">is_first_pp_rank</span><span class="p">():</span>
<span class="n">input_ids</span> <span class="o">=</span> <span class="n">Tensor</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s1">&#39;input_ids&#39;</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="n">trt</span><span class="o">.</span><span class="n">int32</span><span class="p">,</span>
<span class="n">shape</span><span class="o">=</span><span class="p">[</span><span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="o">-</span><span class="mi">1</span><span class="p">],</span>
<span class="n">dim_range</span><span class="o">=</span><span class="n">OrderedDict</span><span class="p">([</span>
<span class="p">(</span><span class="s1">&#39;batch_size_beam_width&#39;</span><span class="p">,</span> <span class="p">[</span><span class="n">bb_range</span><span class="p">]),</span>
<span class="p">(</span><span class="s1">&#39;input_len&#39;</span><span class="p">,</span> <span class="p">[</span><span class="n">inlen_range</span><span class="p">]),</span>
<span class="p">]))</span>
<span class="k">else</span><span class="p">:</span>
<span class="n">hidden_states</span> <span class="o">=</span> <span class="n">Tensor</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s1">&#39;hidden_states_input&#39;</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">_dtype</span><span class="p">,</span>
<span class="n">shape</span><span class="o">=</span><span class="p">[</span><span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">hidden_size</span><span class="p">],</span>
<span class="n">dim_range</span><span class="o">=</span><span class="n">OrderedDict</span><span class="p">([</span>
<span class="p">(</span><span class="s1">&#39;batch_size_beam_width&#39;</span><span class="p">,</span> <span class="p">[</span><span class="n">bb_range</span>
<span class="p">]),</span>
<span class="p">(</span><span class="s1">&#39;input_len&#39;</span><span class="p">,</span> <span class="p">[</span><span class="n">inlen_range</span><span class="p">]),</span>
<span class="p">(</span><span class="s1">&#39;hidden_size&#39;</span><span class="p">,</span> <span class="p">[</span><span class="bp">self</span><span class="o">.</span><span class="n">hidden_size</span><span class="p">]),</span>
<span class="p">]))</span>
<span class="n">encoder_input_lengths</span> <span class="o">=</span> <span class="n">Tensor</span><span class="p">(</span>
<span class="n">name</span><span class="o">=</span><span class="s2">&quot;encoder_input_lengths&quot;</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="n">trt</span><span class="o">.</span><span class="n">int32</span><span class="p">,</span>
<span class="n">shape</span><span class="o">=</span><span class="p">[</span><span class="o">-</span><span class="mi">1</span><span class="p">],</span>
<span class="n">dim_range</span><span class="o">=</span><span class="n">OrderedDict</span><span class="p">([(</span><span class="s2">&quot;batch_size_beam_width&quot;</span><span class="p">,</span> <span class="p">[</span><span class="n">bb_range</span><span class="p">])]),</span>
<span class="p">)</span>
<span class="n">encoder_max_input_length</span> <span class="o">=</span> <span class="n">Tensor</span><span class="p">(</span>
<span class="n">name</span><span class="o">=</span><span class="s2">&quot;encoder_max_input_length&quot;</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="n">trt</span><span class="o">.</span><span class="n">int32</span><span class="p">,</span>
<span class="n">shape</span><span class="o">=</span><span class="p">[</span><span class="o">-</span><span class="mi">1</span><span class="p">],</span>
<span class="n">dim_range</span><span class="o">=</span><span class="n">OrderedDict</span><span class="p">([(</span><span class="s2">&quot;encoder_max_input_length&quot;</span><span class="p">,</span>
<span class="p">[</span><span class="n">encoder_inlen_range</span><span class="p">])]),</span>
<span class="p">)</span>
<span class="k">if</span> <span class="n">remove_input_padding</span><span class="p">:</span>
<span class="n">encoder_output</span> <span class="o">=</span> <span class="n">Tensor</span><span class="p">(</span>
<span class="n">name</span><span class="o">=</span><span class="s2">&quot;encoder_output&quot;</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">_dtype</span><span class="p">,</span>
<span class="n">shape</span><span class="o">=</span><span class="p">[</span><span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">hidden_size</span><span class="p">],</span>
<span class="n">dim_range</span><span class="o">=</span><span class="n">OrderedDict</span><span class="p">([</span>
<span class="p">(</span><span class="s2">&quot;encoder_num_tokens&quot;</span><span class="p">,</span> <span class="p">[</span><span class="n">encoder_num_tokens_range</span><span class="p">]),</span>
<span class="p">(</span><span class="s2">&quot;hidden_size&quot;</span><span class="p">,</span> <span class="p">[</span><span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">hidden_size</span><span class="p">]),</span>
<span class="p">]),</span>
<span class="p">)</span>
<span class="k">else</span><span class="p">:</span>
<span class="n">encoder_output</span> <span class="o">=</span> <span class="n">Tensor</span><span class="p">(</span>
<span class="n">name</span><span class="o">=</span><span class="s2">&quot;encoder_output&quot;</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">_dtype</span><span class="p">,</span>
<span class="n">shape</span><span class="o">=</span><span class="p">[</span><span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">hidden_size</span><span class="p">],</span>
<span class="n">dim_range</span><span class="o">=</span><span class="n">OrderedDict</span><span class="p">([</span>
<span class="p">(</span><span class="s2">&quot;batch_size_beam_width_encoder&quot;</span><span class="p">,</span> <span class="p">[</span><span class="n">bb_range</span><span class="p">]),</span>
<span class="p">(</span><span class="s2">&quot;encoder_input_len&quot;</span><span class="p">,</span> <span class="p">[</span><span class="n">encoder_input_len_range</span><span class="p">]),</span>
<span class="p">(</span><span class="s2">&quot;hidden_size&quot;</span><span class="p">,</span> <span class="p">[</span><span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">hidden_size</span><span class="p">]),</span>
<span class="p">]),</span>
<span class="p">)</span>
<span class="n">context_lengths</span> <span class="o">=</span> <span class="kc">None</span>
<span class="n">host_context_lengths</span> <span class="o">=</span> <span class="kc">None</span>
<span class="n">host_request_types</span> <span class="o">=</span> <span class="kc">None</span>
<span class="n">host_runtime_perf_knobs</span> <span class="o">=</span> <span class="kc">None</span>
<span class="n">host_context_progress</span> <span class="o">=</span> <span class="kc">None</span>
<span class="k">if</span> <span class="n">use_gpt_attention_plugin</span> <span class="ow">and</span> <span class="n">remove_input_padding</span><span class="p">:</span>
<span class="n">host_context_lengths</span> <span class="o">=</span> <span class="n">Tensor</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s1">&#39;host_context_lengths&#39;</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="n">trt</span><span class="o">.</span><span class="n">int32</span><span class="p">,</span>
<span class="n">shape</span><span class="o">=</span><span class="p">[</span><span class="o">-</span><span class="mi">1</span><span class="p">],</span>
<span class="n">dim_range</span><span class="o">=</span><span class="n">OrderedDict</span><span class="p">([</span>
<span class="p">(</span><span class="s1">&#39;batch_size_beam_width&#39;</span><span class="p">,</span>
<span class="p">[</span><span class="n">bb_range</span><span class="p">])</span>
<span class="p">]))</span>
<span class="k">if</span> <span class="n">use_gpt_attention_plugin</span><span class="p">:</span>
<span class="k">if</span> <span class="n">kv_cache_type</span> <span class="o">!=</span> <span class="n">KVCacheType</span><span class="o">.</span><span class="n">DISABLED</span><span class="p">:</span>
<span class="n">sequence_length</span> <span class="o">=</span> <span class="n">Tensor</span><span class="p">(</span>
<span class="n">name</span><span class="o">=</span><span class="s1">&#39;sequence_length&#39;</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="n">trt</span><span class="o">.</span><span class="n">int32</span><span class="p">,</span>
<span class="n">shape</span><span class="o">=</span><span class="p">[</span><span class="o">-</span><span class="mi">1</span><span class="p">],</span>
<span class="n">dim_range</span><span class="o">=</span><span class="n">OrderedDict</span><span class="p">([(</span><span class="s1">&#39;batch_size_beam_width&#39;</span><span class="p">,</span> <span class="p">[</span><span class="n">bb_range</span><span class="p">])</span>
<span class="p">]),</span>
<span class="p">)</span>
<span class="n">host_past_key_value_lengths</span> <span class="o">=</span> <span class="n">Tensor</span><span class="p">(</span>
<span class="n">name</span><span class="o">=</span><span class="s1">&#39;host_past_key_value_lengths&#39;</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="n">trt</span><span class="o">.</span><span class="n">int32</span><span class="p">,</span>
<span class="n">shape</span><span class="o">=</span><span class="p">[</span><span class="o">-</span><span class="mi">1</span><span class="p">],</span>
<span class="n">dim_range</span><span class="o">=</span><span class="n">OrderedDict</span><span class="p">([(</span><span class="s1">&#39;batch_size_beam_width&#39;</span><span class="p">,</span> <span class="p">[</span><span class="n">bb_range</span><span class="p">])</span>
<span class="p">]),</span>
<span class="p">)</span>
<span class="n">context_lengths</span> <span class="o">=</span> <span class="n">Tensor</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s1">&#39;context_lengths&#39;</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="n">trt</span><span class="o">.</span><span class="n">int32</span><span class="p">,</span>
<span class="n">shape</span><span class="o">=</span><span class="p">[</span><span class="o">-</span><span class="mi">1</span><span class="p">],</span>
<span class="n">dim_range</span><span class="o">=</span><span class="n">OrderedDict</span><span class="p">([</span>
<span class="p">(</span><span class="s1">&#39;batch_size_beam_width&#39;</span><span class="p">,</span> <span class="p">[</span><span class="n">bb_range</span><span class="p">])</span>
<span class="p">]))</span>
<span class="n">host_request_types</span> <span class="o">=</span> <span class="n">Tensor</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s1">&#39;host_request_types&#39;</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="n">trt</span><span class="o">.</span><span class="n">int32</span><span class="p">,</span>
<span class="n">shape</span><span class="o">=</span><span class="p">[</span><span class="o">-</span><span class="mi">1</span><span class="p">],</span>
<span class="n">dim_range</span><span class="o">=</span><span class="n">OrderedDict</span><span class="p">([</span>
<span class="p">(</span><span class="s1">&#39;batch_size_beam_width&#39;</span><span class="p">,</span>
<span class="p">[</span><span class="n">bb_range</span><span class="p">])</span>
<span class="p">]))</span>
<span class="n">host_runtime_perf_knobs</span> <span class="o">=</span> <span class="n">Tensor</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s1">&#39;host_runtime_perf_knobs&#39;</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="n">trt</span><span class="o">.</span><span class="n">int64</span><span class="p">,</span>
<span class="n">shape</span><span class="o">=</span><span class="p">[</span><span class="mi">16</span><span class="p">],</span>
<span class="n">dim_range</span><span class="o">=</span><span class="n">OrderedDict</span><span class="p">([</span>
<span class="p">(</span><span class="s1">&#39;perf_knob_size&#39;</span><span class="p">,</span> <span class="p">[</span><span class="mi">16</span><span class="p">])</span>
<span class="p">]))</span>
<span class="n">host_context_progress</span> <span class="o">=</span> <span class="n">Tensor</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s1">&#39;host_context_progress&#39;</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="n">trt</span><span class="o">.</span><span class="n">int64</span><span class="p">,</span>
<span class="n">shape</span><span class="o">=</span><span class="p">[</span><span class="mi">1</span><span class="p">],</span>
<span class="n">dim_range</span><span class="o">=</span><span class="n">OrderedDict</span><span class="p">([</span>
<span class="p">(</span><span class="s1">&#39;context_progress_size&#39;</span><span class="p">,</span> <span class="p">[</span><span class="mi">1</span><span class="p">])</span>
<span class="p">]))</span>
<span class="n">last_token_ids</span> <span class="o">=</span> <span class="kc">None</span>
<span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">mapping</span><span class="o">.</span><span class="n">is_last_pp_rank</span><span class="p">()</span> <span class="ow">and</span> <span class="ow">not</span> <span class="n">gather_context_logits</span><span class="p">:</span>
<span class="n">last_token_ids</span> <span class="o">=</span> <span class="n">Tensor</span><span class="p">(</span>
<span class="n">name</span><span class="o">=</span><span class="s2">&quot;last_token_ids&quot;</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="n">trt</span><span class="o">.</span><span class="n">int32</span><span class="p">,</span>
<span class="n">shape</span><span class="o">=</span><span class="p">[</span><span class="o">-</span><span class="mi">1</span><span class="p">],</span>
<span class="n">dim_range</span><span class="o">=</span><span class="n">OrderedDict</span><span class="p">([(</span><span class="s2">&quot;batch_size_last_token_ids&quot;</span><span class="p">,</span> <span class="p">[</span><span class="n">bb_range</span><span class="p">])</span>
<span class="p">]),</span>
<span class="p">)</span>
<span class="n">attention_mask</span> <span class="o">=</span> <span class="kc">None</span>
<span class="k">if</span> <span class="ow">not</span> <span class="n">use_gpt_attention_plugin</span><span class="p">:</span>
<span class="n">attention_mask</span> <span class="o">=</span> <span class="n">Tensor</span><span class="p">(</span>
<span class="n">name</span><span class="o">=</span><span class="s1">&#39;attention_mask&#39;</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="n">trt</span><span class="o">.</span><span class="n">int32</span><span class="p">,</span>
<span class="n">shape</span><span class="o">=</span><span class="p">[</span><span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="o">-</span><span class="mi">1</span><span class="p">],</span>
<span class="n">dim_range</span><span class="o">=</span><span class="n">OrderedDict</span><span class="p">([</span>
<span class="p">(</span><span class="s1">&#39;batch_size_beam_width&#39;</span><span class="p">,</span> <span class="p">[</span><span class="n">bb_range</span><span class="p">]),</span>
<span class="p">(</span><span class="s1">&#39;mask_len&#39;</span><span class="p">,</span> <span class="p">[</span><span class="n">mask_len_range</span><span class="p">]),</span>
<span class="p">]),</span>
<span class="p">)</span>
<span class="k">assert</span> <span class="kc">False</span><span class="p">,</span> <span class="s2">&quot;not support non-attention-plugin case now&quot;</span>
<span class="n">cross_attention_mask</span> <span class="o">=</span> <span class="n">Tensor</span><span class="p">(</span>
<span class="n">name</span><span class="o">=</span><span class="s1">&#39;cross_attention_mask&#39;</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="n">trt</span><span class="o">.</span><span class="n">bool</span><span class="p">,</span>
<span class="n">shape</span><span class="o">=</span><span class="p">[</span><span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="o">-</span><span class="mi">1</span><span class="p">],</span>
<span class="n">dim_range</span><span class="o">=</span><span class="n">OrderedDict</span><span class="p">([</span>
<span class="p">(</span><span class="s1">&#39;decoder_num_tokens_2&#39;</span><span class="p">,</span>
<span class="p">[</span><span class="n">decoder_num_tokens_range</span>
<span class="p">]),</span> <span class="c1"># TODO should use same name as input_ids</span>
<span class="p">(</span><span class="s1">&#39;encoder_input_len_2&#39;</span><span class="p">,</span> <span class="p">[</span><span class="n">encoder_input_len_range</span><span class="p">]),</span>
<span class="p">]),</span>
<span class="p">)</span>
<span class="n">cross_attention_packed_mask</span> <span class="o">=</span> <span class="n">Tensor</span><span class="p">(</span>
<span class="n">name</span><span class="o">=</span><span class="s1">&#39;cross_attention_packed_mask&#39;</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="n">trt</span><span class="o">.</span><span class="n">int32</span><span class="p">,</span>
<span class="n">shape</span><span class="o">=</span><span class="p">[</span><span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="o">-</span><span class="mi">1</span><span class="p">],</span>
<span class="n">dim_range</span><span class="o">=</span><span class="n">OrderedDict</span><span class="p">([</span>
<span class="p">(</span><span class="s1">&#39;cross_packed_mask_dim0&#39;</span><span class="p">,</span> <span class="p">[</span><span class="n">cross_packed_mask_dim0_range</span><span class="p">]),</span>
<span class="p">(</span><span class="s1">&#39;cross_packed_mask_dim1&#39;</span><span class="p">,</span> <span class="p">[</span><span class="n">cross_packed_mask_dim1_range</span><span class="p">]),</span>
<span class="p">]),</span>
<span class="p">)</span>
<span class="c1"># create the attention_mask_params.</span>
<span class="n">attention_mask_params</span> <span class="o">=</span> <span class="n">AttentionMaskParams</span><span class="p">(</span>
<span class="n">attention_mask</span><span class="p">,</span> <span class="kc">None</span><span class="p">,</span> <span class="n">cross_attention_mask</span><span class="p">,</span>
<span class="n">cross_attention_packed_mask</span><span class="p">)</span>
<span class="n">cache_indirection</span> <span class="o">=</span> <span class="n">Tensor</span><span class="p">(</span>
<span class="n">name</span><span class="o">=</span><span class="s1">&#39;cache_indirection&#39;</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="n">trt</span><span class="o">.</span><span class="n">int32</span><span class="p">,</span>
<span class="n">shape</span><span class="o">=</span><span class="p">[</span><span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="o">-</span><span class="mi">1</span><span class="p">],</span>
<span class="n">dim_range</span><span class="o">=</span><span class="n">OrderedDict</span><span class="p">([</span>
<span class="p">(</span><span class="s1">&#39;batch_size_cache&#39;</span><span class="p">,</span> <span class="p">[</span><span class="n">bs_range</span><span class="p">]),</span>
<span class="p">(</span><span class="s1">&#39;beam_width&#39;</span><span class="p">,</span> <span class="p">[</span><span class="n">beam_width_range</span><span class="p">]),</span>
<span class="p">(</span><span class="s1">&#39;max_seq_len&#39;</span><span class="p">,</span> <span class="p">[</span><span class="n">max_output_len_range</span><span class="p">]),</span>
<span class="p">]),</span>
<span class="p">)</span>
<span class="n">layers_range</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">mapping</span><span class="o">.</span><span class="n">pp_layers</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">transformer</span><span class="o">.</span><span class="n">total_num_layers</span><span class="p">)</span>
<span class="n">num_pp_layers</span> <span class="o">=</span> <span class="nb">len</span><span class="p">(</span><span class="n">layers_range</span><span class="p">)</span>
<span class="n">host_max_attention_window_sizes</span> <span class="o">=</span> <span class="kc">None</span>
<span class="n">host_sink_token_length</span> <span class="o">=</span> <span class="kc">None</span>
<span class="k">if</span> <span class="n">use_gpt_attention_plugin</span><span class="p">:</span>
<span class="n">host_max_attention_window_sizes</span> <span class="o">=</span> <span class="n">Tensor</span><span class="p">(</span>
<span class="n">name</span><span class="o">=</span><span class="sa">f</span><span class="s1">&#39;host_max_attention_window_sizes&#39;</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="n">trt</span><span class="o">.</span><span class="n">int32</span><span class="p">,</span>
<span class="n">shape</span><span class="o">=</span><span class="p">[</span><span class="n">num_pp_layers</span><span class="p">],</span>
<span class="n">dim_range</span><span class="o">=</span><span class="n">OrderedDict</span><span class="p">([(</span><span class="s1">&#39;num_layers&#39;</span><span class="p">,</span> <span class="p">[</span><span class="n">num_pp_layers</span><span class="p">])]))</span>
<span class="n">host_sink_token_length</span> <span class="o">=</span> <span class="n">Tensor</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s1">&#39;host_sink_token_length&#39;</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="n">trt</span><span class="o">.</span><span class="n">int32</span><span class="p">,</span>
<span class="n">shape</span><span class="o">=</span><span class="p">[</span><span class="mi">1</span><span class="p">],</span>
<span class="n">dim_range</span><span class="o">=</span><span class="n">OrderedDict</span><span class="p">([(</span><span class="s1">&#39;scalar&#39;</span><span class="p">,</span>
<span class="p">[</span><span class="mi">1</span><span class="p">])]))</span>
<span class="c1"># TODO LoRA for mllama is not verified.</span>
<span class="n">lora_weights_pointers</span> <span class="o">=</span> <span class="kc">None</span>
<span class="n">lora_ranks</span> <span class="o">=</span> <span class="kc">None</span>
<span class="n">lora_params</span> <span class="o">=</span> <span class="kc">None</span>
<span class="k">if</span> <span class="n">use_lora_plugin</span><span class="p">:</span>
<span class="n">lora_weights_pointers</span> <span class="o">=</span> <span class="p">[]</span>
<span class="n">lora_ranks</span> <span class="o">=</span> <span class="p">[]</span>
<span class="n">missing_qkv_modules</span> <span class="o">=</span> <span class="p">[]</span>
<span class="k">if</span> <span class="nb">any</span><span class="p">(</span><span class="n">x</span> <span class="ow">in</span> <span class="n">lora_target_modules</span>
<span class="k">for</span> <span class="n">x</span> <span class="ow">in</span> <span class="p">[</span><span class="s2">&quot;attn_q&quot;</span><span class="p">,</span> <span class="s2">&quot;attn_k&quot;</span><span class="p">,</span> <span class="s2">&quot;attn_v&quot;</span><span class="p">]):</span>
<span class="k">for</span> <span class="n">lora_module</span> <span class="ow">in</span> <span class="p">[</span>
<span class="s2">&quot;attn_q&quot;</span><span class="p">,</span>
<span class="s2">&quot;attn_k&quot;</span><span class="p">,</span>
<span class="s2">&quot;attn_v&quot;</span><span class="p">,</span>
<span class="p">]:</span>
<span class="k">if</span> <span class="n">lora_module</span> <span class="ow">not</span> <span class="ow">in</span> <span class="n">lora_target_modules</span><span class="p">:</span>
<span class="n">missing_qkv_modules</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">lora_module</span><span class="p">)</span>
<span class="k">if</span> <span class="nb">any</span><span class="p">(</span><span class="n">x</span> <span class="ow">in</span> <span class="n">lora_target_modules</span>
<span class="k">for</span> <span class="n">x</span> <span class="ow">in</span> <span class="p">[</span><span class="s2">&quot;cross_attn_q&quot;</span><span class="p">,</span> <span class="s2">&quot;cross_attn_k&quot;</span><span class="p">,</span> <span class="s2">&quot;cross_attn_v&quot;</span><span class="p">]):</span>
<span class="k">for</span> <span class="n">lora_module</span> <span class="ow">in</span> <span class="p">[</span>
<span class="s2">&quot;cross_attn_q&quot;</span><span class="p">,</span> <span class="s2">&quot;cross_attn_k&quot;</span><span class="p">,</span> <span class="s2">&quot;cross_attn_v&quot;</span>
<span class="p">]:</span>
<span class="k">if</span> <span class="n">lora_module</span> <span class="ow">not</span> <span class="ow">in</span> <span class="n">lora_target_modules</span><span class="p">:</span>
<span class="n">missing_qkv_modules</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">lora_module</span><span class="p">)</span>
<span class="c1"># For LoRA</span>
<span class="k">for</span> <span class="n">i</span> <span class="ow">in</span> <span class="n">layers_range</span><span class="p">:</span>
<span class="n">lora_weight_pointer_dict</span> <span class="o">=</span> <span class="p">{}</span>
<span class="n">lora_rank_dict</span> <span class="o">=</span> <span class="p">{}</span>
<span class="k">for</span> <span class="n">lora_module</span> <span class="ow">in</span> <span class="p">(</span><span class="n">lora_target_modules</span> <span class="o">+</span> <span class="n">missing_qkv_modules</span><span class="p">):</span>
<span class="n">lora_weight_pointer</span> <span class="o">=</span> <span class="n">Tensor</span><span class="p">(</span>
<span class="n">name</span><span class="o">=</span><span class="sa">f</span><span class="s1">&#39;</span><span class="si">{</span><span class="n">lora_module</span><span class="si">}</span><span class="s1">_lora_weights_pointers_</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">&#39;</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="n">trt</span><span class="o">.</span><span class="n">int64</span><span class="p">,</span>
<span class="n">shape</span><span class="o">=</span><span class="p">[</span><span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="mi">3</span><span class="p">],</span>
<span class="n">dim_range</span><span class="o">=</span><span class="n">OrderedDict</span><span class="p">([(</span><span class="s1">&#39;batch_size_beam_width&#39;</span><span class="p">,</span>
<span class="p">[</span><span class="n">bb_range</span><span class="p">]),</span>
<span class="p">(</span><span class="s1">&#39;in_out_scales&#39;</span><span class="p">,</span> <span class="p">[</span><span class="mi">3</span><span class="p">])]))</span>
<span class="n">lora_weight_pointer_dict</span><span class="o">.</span><span class="n">update</span><span class="p">({</span>
<span class="sa">f</span><span class="s1">&#39;</span><span class="si">{</span><span class="n">lora_module</span><span class="si">}</span><span class="s1">_lora_weights_pointers&#39;</span><span class="p">:</span>
<span class="n">lora_weight_pointer</span>
<span class="p">})</span>
<span class="n">lora_rank</span> <span class="o">=</span> <span class="n">Tensor</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="sa">f</span><span class="s1">&#39;</span><span class="si">{</span><span class="n">lora_module</span><span class="si">}</span><span class="s1">_lora_ranks_</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">&#39;</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="n">trt</span><span class="o">.</span><span class="n">int32</span><span class="p">,</span>
<span class="n">shape</span><span class="o">=</span><span class="p">[</span><span class="o">-</span><span class="mi">1</span><span class="p">],</span>
<span class="n">dim_range</span><span class="o">=</span><span class="n">OrderedDict</span><span class="p">([</span>
<span class="p">(</span><span class="s1">&#39;batch_size_beam_width&#39;</span><span class="p">,</span> <span class="p">[</span><span class="n">bb_range</span><span class="p">])</span>
<span class="p">]))</span>
<span class="n">lora_rank_dict</span><span class="o">.</span><span class="n">update</span><span class="p">(</span>
<span class="p">{</span><span class="sa">f</span><span class="s1">&#39;</span><span class="si">{</span><span class="n">lora_module</span><span class="si">}</span><span class="s1">_lora_ranks&#39;</span><span class="p">:</span> <span class="n">lora_rank</span><span class="p">})</span>
<span class="n">lora_weights_pointers</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">lora_weight_pointer_dict</span><span class="p">)</span>
<span class="n">lora_ranks</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">lora_rank_dict</span><span class="p">)</span>
<span class="c1"># For cross attention, we need to use encoder_input_lengths (in CPU) to pass</span>
<span class="c1"># as the host_context_lengths to the lora_plugin. But for self attention, we</span>
<span class="c1"># should keep using the original host_context_lengths. Therefore, we keep both</span>
<span class="c1"># of them in the lora_params.</span>
<span class="n">host_encoder_input_lengths</span> <span class="o">=</span> <span class="kc">None</span>
<span class="k">if</span> <span class="n">remove_input_padding</span><span class="p">:</span>
<span class="n">host_encoder_input_lengths</span> <span class="o">=</span> <span class="n">Tensor</span><span class="p">(</span>
<span class="n">name</span><span class="o">=</span><span class="s2">&quot;host_encoder_input_lengths&quot;</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="n">trt</span><span class="o">.</span><span class="n">int32</span><span class="p">,</span>
<span class="n">shape</span><span class="o">=</span><span class="p">[</span><span class="o">-</span><span class="mi">1</span><span class="p">],</span>
<span class="n">dim_range</span><span class="o">=</span><span class="n">OrderedDict</span><span class="p">([(</span><span class="s2">&quot;batch_size_beam_width&quot;</span><span class="p">,</span> <span class="p">[</span><span class="n">bb_range</span><span class="p">])</span>
<span class="p">]),</span>
<span class="p">)</span>
<span class="n">lora_params</span> <span class="o">=</span> <span class="n">LoraParams</span><span class="p">(</span>
<span class="n">lora_ranks</span><span class="o">=</span><span class="n">lora_ranks</span><span class="p">,</span>
<span class="n">lora_weights_pointers</span><span class="o">=</span><span class="n">lora_weights_pointers</span><span class="p">,</span>
<span class="n">host_context_lengths</span><span class="o">=</span><span class="n">host_context_lengths</span><span class="p">,</span>
<span class="n">max_context_length</span><span class="o">=</span><span class="n">max_decoder_input_len</span><span class="p">,</span>
<span class="n">max_encoder_context_length</span><span class="o">=</span><span class="n">max_encoder_input_len</span><span class="p">,</span>
<span class="n">host_request_types</span><span class="o">=</span><span class="n">host_request_types</span><span class="p">,</span>
<span class="n">host_encoder_input_lengths</span><span class="o">=</span><span class="n">host_encoder_input_lengths</span><span class="p">,</span>
<span class="p">)</span>
<span class="n">kv_cache_block_offsets</span> <span class="o">=</span> <span class="kc">None</span>
<span class="n">host_kv_cache_block_offsets</span> <span class="o">=</span> <span class="kc">None</span>
<span class="n">host_kv_cache_pool_pointers</span> <span class="o">=</span> <span class="kc">None</span>
<span class="n">host_kv_cache_pool_mapping</span> <span class="o">=</span> <span class="kc">None</span>
<span class="n">cross_kv_cache_block_offsets</span> <span class="o">=</span> <span class="kc">None</span>
<span class="n">host_cross_kv_cache_block_offsets</span> <span class="o">=</span> <span class="kc">None</span>
<span class="n">host_cross_kv_cache_pool_pointers</span> <span class="o">=</span> <span class="kc">None</span>
<span class="n">host_cross_kv_cache_pool_mapping</span> <span class="o">=</span> <span class="kc">None</span>
<span class="k">if</span> <span class="n">use_cache</span><span class="p">:</span>
<span class="k">if</span> <span class="ow">not</span> <span class="n">paged_kv_cache</span><span class="p">:</span>
<span class="k">for</span> <span class="n">i</span> <span class="ow">in</span> <span class="n">layers_range</span><span class="p">:</span>
<span class="n">kv_dim_range</span> <span class="o">=</span> <span class="n">OrderedDict</span><span class="p">([</span>
<span class="p">(</span><span class="s1">&#39;batch_size_beam_width&#39;</span><span class="p">,</span> <span class="p">[</span><span class="n">bb_range</span><span class="p">]),</span>
<span class="p">(</span><span class="s1">&#39;kv&#39;</span><span class="p">,</span> <span class="p">[</span><span class="mi">2</span><span class="p">]),</span>
<span class="p">(</span><span class="s1">&#39;num_heads&#39;</span><span class="p">,</span> <span class="p">[</span><span class="n">num_kv_heads</span><span class="p">]),</span>
<span class="p">(</span><span class="s1">&#39;past_key_len&#39;</span><span class="p">,</span> <span class="p">[</span><span class="n">max_output_len_range</span><span class="p">]),</span>
<span class="p">(</span><span class="s1">&#39;head_size&#39;</span><span class="p">,</span> <span class="p">[</span><span class="n">head_size</span><span class="p">]),</span>
<span class="p">])</span>
<span class="n">kv</span> <span class="o">=</span> <span class="n">Tensor</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="sa">f</span><span class="s1">&#39;past_key_value_</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">&#39;</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">_kv_dtype</span><span class="p">,</span>
<span class="n">shape</span><span class="o">=</span><span class="p">[</span><span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="mi">2</span><span class="p">,</span> <span class="n">num_kv_heads</span><span class="p">,</span> <span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="n">head_size</span><span class="p">],</span>
<span class="n">dim_range</span><span class="o">=</span><span class="n">kv_dim_range</span><span class="p">)</span>
<span class="n">past_key_value</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">kv</span><span class="p">)</span>
<span class="k">if</span> <span class="n">i</span> <span class="ow">in</span> <span class="bp">self</span><span class="o">.</span><span class="n">transformer</span><span class="o">.</span><span class="n">cross_attention_layers</span><span class="p">:</span>
<span class="n">xa_layer_id</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">transformer</span><span class="o">.</span><span class="n">cross_attention_layers</span><span class="o">.</span><span class="n">index</span><span class="p">(</span>
<span class="n">i</span><span class="p">)</span> <span class="o">+</span> <span class="n">layers_range</span><span class="p">[</span><span class="o">-</span><span class="mi">1</span><span class="p">]</span>
<span class="n">cross_kv_dim_range</span> <span class="o">=</span> <span class="n">OrderedDict</span><span class="p">([</span>
<span class="p">(</span><span class="s1">&#39;batch_size_beam_width&#39;</span><span class="p">,</span> <span class="p">[</span><span class="n">bb_range</span><span class="p">]),</span>
<span class="p">(</span><span class="s1">&#39;kv&#39;</span><span class="p">,</span> <span class="p">[</span><span class="mi">2</span><span class="p">]),</span>
<span class="p">(</span><span class="s1">&#39;cross_num_heads&#39;</span><span class="p">,</span> <span class="p">[</span><span class="n">encoder_num_kv_heads</span><span class="p">]),</span>
<span class="p">(</span><span class="s1">&#39;cross_past_key_len&#39;</span><span class="p">,</span> <span class="p">[</span><span class="n">encoder_input_len_range</span><span class="p">]),</span>
<span class="p">(</span><span class="s1">&#39;cross_head_size&#39;</span><span class="p">,</span> <span class="p">[</span><span class="n">encoder_head_size</span><span class="p">]),</span>
<span class="p">])</span>
<span class="n">cross_kv</span> <span class="o">=</span> <span class="n">Tensor</span><span class="p">(</span>
<span class="n">name</span><span class="o">=</span><span class="sa">f</span><span class="s1">&#39;cross_past_key_value_</span><span class="si">{</span><span class="n">xa_layer_id</span><span class="si">}</span><span class="s1">&#39;</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">_kv_dtype</span><span class="p">,</span>
<span class="n">shape</span><span class="o">=</span><span class="p">[</span>
<span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="mi">2</span><span class="p">,</span> <span class="n">encoder_num_kv_heads</span><span class="p">,</span> <span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="n">encoder_head_size</span>
<span class="p">],</span>
<span class="n">dim_range</span><span class="o">=</span><span class="n">cross_kv_dim_range</span><span class="p">)</span>
<span class="n">past_key_value</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">kv</span><span class="p">)</span>
<span class="c1"># TODO: Remove this when TRT fix the named dimension</span>
<span class="k">if</span> <span class="ow">not</span> <span class="n">remove_input_padding</span><span class="p">:</span>
<span class="n">assertion</span><span class="p">(</span>
<span class="n">shape</span><span class="p">(</span>
<span class="n">input_ids</span> <span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">mapping</span><span class="o">.</span><span class="n">is_first_pp_rank</span><span class="p">()</span> <span class="k">else</span>
<span class="n">hidden_states</span><span class="p">,</span> <span class="mi">0</span><span class="p">)</span> <span class="o">==</span> <span class="n">shape</span><span class="p">(</span><span class="n">kv</span><span class="p">,</span> <span class="mi">0</span><span class="p">),</span> <span class="s1">&#39;batch size&#39;</span><span class="p">)</span>
<span class="k">else</span><span class="p">:</span> <span class="c1"># paged_kv_cache == True</span>
<span class="c1"># PagedKV setup for KV cache of self-attention</span>
<span class="n">max_blocks_per_seq_range</span> <span class="o">=</span> <span class="p">[[</span>
<span class="n">math</span><span class="o">.</span><span class="n">ceil</span><span class="p">(</span><span class="n">max_output_len_range</span><span class="p">[</span><span class="mi">0</span><span class="p">]</span> <span class="o">/</span> <span class="n">tokens_per_block</span><span class="p">),</span>
<span class="n">math</span><span class="o">.</span><span class="n">ceil</span><span class="p">(</span><span class="n">max_output_len_range</span><span class="p">[</span><span class="mi">1</span><span class="p">]</span> <span class="o">/</span> <span class="n">tokens_per_block</span><span class="p">),</span>
<span class="n">math</span><span class="o">.</span><span class="n">ceil</span><span class="p">(</span><span class="n">max_output_len_range</span><span class="p">[</span><span class="mi">2</span><span class="p">]</span> <span class="o">/</span> <span class="n">tokens_per_block</span><span class="p">)</span>
<span class="p">]]</span>
<span class="n">max_blocks_per_seq_range</span> <span class="o">=</span> <span class="p">[[</span>
<span class="n">x</span> <span class="k">for</span> <span class="n">x</span> <span class="ow">in</span> <span class="n">max_blocks_per_seq_range</span><span class="p">[</span><span class="mi">0</span><span class="p">]</span>
<span class="p">]]</span>
<span class="c1"># PagedKV setup for KV cache of cross-attention</span>
<span class="n">max_cross_blocks_per_seq_range</span> <span class="o">=</span> <span class="p">[[</span>
<span class="n">math</span><span class="o">.</span><span class="n">ceil</span><span class="p">(</span><span class="n">encoder_input_len_range</span><span class="p">[</span><span class="mi">0</span><span class="p">]</span> <span class="o">/</span> <span class="n">tokens_per_block</span><span class="p">),</span>
<span class="n">math</span><span class="o">.</span><span class="n">ceil</span><span class="p">(</span><span class="n">encoder_input_len_range</span><span class="p">[</span><span class="mi">1</span><span class="p">]</span> <span class="o">/</span> <span class="n">tokens_per_block</span><span class="p">),</span>
<span class="n">math</span><span class="o">.</span><span class="n">ceil</span><span class="p">(</span><span class="n">encoder_input_len_range</span><span class="p">[</span><span class="mi">2</span><span class="p">]</span> <span class="o">/</span> <span class="n">tokens_per_block</span><span class="p">)</span>
<span class="p">]]</span>
<span class="n">max_cross_blocks_per_seq_range</span> <span class="o">=</span> <span class="p">[[</span>
<span class="n">x</span> <span class="k">for</span> <span class="n">x</span> <span class="ow">in</span> <span class="n">max_cross_blocks_per_seq_range</span><span class="p">[</span><span class="mi">0</span><span class="p">]</span>
<span class="p">]]</span>
<span class="n">num_kv_cache_pools</span> <span class="o">=</span> <span class="mi">2</span>
<span class="n">kv_cache_block_offsets</span> <span class="o">=</span> <span class="n">Tensor</span><span class="p">(</span>
<span class="n">name</span><span class="o">=</span><span class="sa">f</span><span class="s1">&#39;kv_cache_block_offsets&#39;</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="n">trt</span><span class="o">.</span><span class="n">int32</span><span class="p">,</span>
<span class="n">shape</span><span class="o">=</span><span class="p">[</span><span class="n">num_kv_cache_pools</span><span class="p">,</span> <span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="mi">2</span><span class="p">,</span> <span class="o">-</span><span class="mi">1</span><span class="p">],</span>
<span class="n">dim_range</span><span class="o">=</span><span class="n">OrderedDict</span><span class="p">([</span>
<span class="p">(</span><span class="s1">&#39;num_kv_cache_pools&#39;</span><span class="p">,</span> <span class="p">[</span><span class="n">num_kv_cache_pools</span><span class="p">]),</span>
<span class="p">(</span><span class="s1">&#39;batch_size_beam_width&#39;</span><span class="p">,</span> <span class="p">[</span><span class="n">bb_range</span><span class="p">]),</span>
<span class="p">(</span><span class="s1">&#39;kv&#39;</span><span class="p">,</span> <span class="p">[</span><span class="mi">2</span><span class="p">]),</span>
<span class="p">(</span><span class="s1">&#39;max_blocks_per_seq&#39;</span><span class="p">,</span> <span class="n">max_blocks_per_seq_range</span><span class="p">),</span>
<span class="p">]))</span>
<span class="n">host_kv_cache_block_offsets</span> <span class="o">=</span> <span class="n">Tensor</span><span class="p">(</span>
<span class="n">name</span><span class="o">=</span><span class="sa">f</span><span class="s1">&#39;host_kv_cache_block_offsets&#39;</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="n">trt</span><span class="o">.</span><span class="n">int32</span><span class="p">,</span>
<span class="n">shape</span><span class="o">=</span><span class="p">[</span><span class="n">num_kv_cache_pools</span><span class="p">,</span> <span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="mi">2</span><span class="p">,</span> <span class="o">-</span><span class="mi">1</span><span class="p">],</span>
<span class="n">dim_range</span><span class="o">=</span><span class="n">OrderedDict</span><span class="p">([</span>
<span class="p">(</span><span class="s1">&#39;num_kv_cache_pools&#39;</span><span class="p">,</span> <span class="p">[</span><span class="n">num_kv_cache_pools</span><span class="p">]),</span>
<span class="p">(</span><span class="s1">&#39;batch_size_beam_width&#39;</span><span class="p">,</span> <span class="p">[</span><span class="n">bb_range</span><span class="p">]),</span>
<span class="p">(</span><span class="s1">&#39;kv&#39;</span><span class="p">,</span> <span class="p">[</span><span class="mi">2</span><span class="p">]),</span>
<span class="p">(</span><span class="s1">&#39;max_blocks_per_seq&#39;</span><span class="p">,</span> <span class="n">max_blocks_per_seq_range</span><span class="p">),</span>
<span class="p">]))</span>
<span class="n">host_kv_cache_pool_pointers</span> <span class="o">=</span> <span class="n">Tensor</span><span class="p">(</span>
<span class="n">name</span><span class="o">=</span><span class="sa">f</span><span class="s1">&#39;host_kv_cache_pool_pointers&#39;</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="n">trt</span><span class="o">.</span><span class="n">int64</span><span class="p">,</span>
<span class="n">shape</span><span class="o">=</span><span class="p">[</span><span class="n">num_kv_cache_pools</span><span class="p">,</span> <span class="mi">2</span><span class="p">],</span>
<span class="n">dim_range</span><span class="o">=</span><span class="n">OrderedDict</span><span class="p">([</span>
<span class="p">(</span><span class="s1">&#39;num_kv_cache_pools&#39;</span><span class="p">,</span> <span class="p">[</span><span class="n">num_kv_cache_pools</span><span class="p">]),</span>
<span class="p">(</span><span class="s1">&#39;num_pools&#39;</span><span class="p">,</span> <span class="p">[</span><span class="mi">2</span><span class="p">]),</span>
<span class="p">]))</span>
<span class="n">host_kv_cache_pool_mapping</span> <span class="o">=</span> <span class="n">Tensor</span><span class="p">(</span>
<span class="n">name</span><span class="o">=</span><span class="sa">f</span><span class="s2">&quot;host_kv_cache_pool_mapping&quot;</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="n">trt</span><span class="o">.</span><span class="n">int32</span><span class="p">,</span>
<span class="c1"># 2: (Index of pool, Index of layer within pool)</span>
<span class="n">shape</span><span class="o">=</span><span class="p">[</span><span class="n">num_pp_layers</span><span class="p">,</span> <span class="mi">2</span><span class="p">],</span>
<span class="n">dim_range</span><span class="o">=</span><span class="n">OrderedDict</span><span class="p">([</span>
<span class="p">(</span><span class="s1">&#39;pools_mapping&#39;</span><span class="p">,</span> <span class="p">[</span><span class="n">num_pp_layers</span><span class="p">]),</span>
<span class="p">(</span><span class="s1">&#39;layer_cache_pool_locator&#39;</span><span class="p">,</span> <span class="p">[</span><span class="mi">2</span><span class="p">]),</span>
<span class="p">]))</span>
<span class="c1"># paged blocks for cross kv</span>
<span class="n">cross_kv_cache_block_offsets</span> <span class="o">=</span> <span class="n">Tensor</span><span class="p">(</span>
<span class="n">name</span><span class="o">=</span><span class="sa">f</span><span class="s1">&#39;cross_kv_cache_block_offsets&#39;</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="n">trt</span><span class="o">.</span><span class="n">int32</span><span class="p">,</span>
<span class="n">shape</span><span class="o">=</span><span class="p">[</span><span class="n">num_kv_cache_pools</span><span class="p">,</span> <span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="mi">2</span><span class="p">,</span> <span class="o">-</span><span class="mi">1</span><span class="p">],</span>
<span class="n">dim_range</span><span class="o">=</span><span class="n">OrderedDict</span><span class="p">([</span>
<span class="p">(</span><span class="s1">&#39;num_kv_cache_pools&#39;</span><span class="p">,</span> <span class="p">[</span><span class="n">num_kv_cache_pools</span><span class="p">]),</span>
<span class="p">(</span><span class="s1">&#39;batch_size_beam_width&#39;</span><span class="p">,</span> <span class="p">[</span><span class="n">bb_range</span><span class="p">]),</span>
<span class="p">(</span><span class="s1">&#39;kv&#39;</span><span class="p">,</span> <span class="p">[</span><span class="mi">2</span><span class="p">]),</span>
<span class="p">(</span><span class="s1">&#39;max_cross_blocks_per_seq&#39;</span><span class="p">,</span>
<span class="n">max_cross_blocks_per_seq_range</span><span class="p">),</span>
<span class="p">]))</span>
<span class="n">host_cross_kv_cache_block_offsets</span> <span class="o">=</span> <span class="n">Tensor</span><span class="p">(</span>
<span class="n">name</span><span class="o">=</span><span class="sa">f</span><span class="s1">&#39;host_cross_kv_cache_block_offsets&#39;</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="n">trt</span><span class="o">.</span><span class="n">int32</span><span class="p">,</span>
<span class="n">shape</span><span class="o">=</span><span class="p">[</span><span class="n">num_kv_cache_pools</span><span class="p">,</span> <span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="mi">2</span><span class="p">,</span> <span class="o">-</span><span class="mi">1</span><span class="p">],</span>
<span class="n">dim_range</span><span class="o">=</span><span class="n">OrderedDict</span><span class="p">([</span>
<span class="p">(</span><span class="s1">&#39;num_kv_cache_pools&#39;</span><span class="p">,</span> <span class="p">[</span><span class="n">num_kv_cache_pools</span><span class="p">]),</span>
<span class="p">(</span><span class="s1">&#39;batch_size_beam_width&#39;</span><span class="p">,</span> <span class="p">[</span><span class="n">bb_range</span><span class="p">]),</span>
<span class="p">(</span><span class="s1">&#39;kv&#39;</span><span class="p">,</span> <span class="p">[</span><span class="mi">2</span><span class="p">]),</span>
<span class="p">(</span><span class="s1">&#39;max_cross_blocks_per_seq&#39;</span><span class="p">,</span>
<span class="n">max_cross_blocks_per_seq_range</span><span class="p">),</span>
<span class="p">]))</span>
<span class="n">host_cross_kv_cache_pool_pointers</span> <span class="o">=</span> <span class="n">Tensor</span><span class="p">(</span>
<span class="n">name</span><span class="o">=</span><span class="sa">f</span><span class="s1">&#39;host_cross_kv_cache_pool_pointers&#39;</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="n">trt</span><span class="o">.</span><span class="n">int64</span><span class="p">,</span>
<span class="n">shape</span><span class="o">=</span><span class="p">[</span><span class="n">num_kv_cache_pools</span><span class="p">,</span> <span class="mi">2</span><span class="p">],</span>
<span class="n">dim_range</span><span class="o">=</span><span class="n">OrderedDict</span><span class="p">([</span>
<span class="p">(</span><span class="s1">&#39;num_kv_cache_pools&#39;</span><span class="p">,</span> <span class="p">[</span><span class="n">num_kv_cache_pools</span><span class="p">]),</span>
<span class="p">(</span><span class="s1">&#39;num_pools&#39;</span><span class="p">,</span> <span class="p">[</span><span class="mi">2</span><span class="p">]),</span>
<span class="p">]))</span>
<span class="n">host_cross_kv_cache_pool_mapping</span> <span class="o">=</span> <span class="n">Tensor</span><span class="p">(</span>
<span class="n">name</span><span class="o">=</span><span class="sa">f</span><span class="s2">&quot;host_cross_kv_cache_pool_mapping&quot;</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="n">trt</span><span class="o">.</span><span class="n">int32</span><span class="p">,</span>
<span class="c1"># 2: (Index of pool, Index of layer within pool)</span>
<span class="n">shape</span><span class="o">=</span><span class="p">[</span><span class="n">num_pp_layers</span><span class="p">,</span> <span class="mi">2</span><span class="p">],</span>
<span class="n">dim_range</span><span class="o">=</span><span class="n">OrderedDict</span><span class="p">([</span>
<span class="p">(</span><span class="s1">&#39;pools_mapping&#39;</span><span class="p">,</span> <span class="p">[</span><span class="n">num_pp_layers</span><span class="p">]),</span>
<span class="p">(</span><span class="s1">&#39;layer_cache_pool_locator&#39;</span><span class="p">,</span> <span class="p">[</span><span class="mi">2</span><span class="p">]),</span>
<span class="p">]))</span>
<span class="k">for</span> <span class="n">i</span> <span class="ow">in</span> <span class="n">layers_range</span><span class="p">:</span>
<span class="n">past_key_value</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="kc">None</span><span class="p">)</span>
<span class="n">kv_cache_params</span> <span class="o">=</span> <span class="n">KeyValueCacheParams</span><span class="p">(</span>
<span class="n">past_key_value</span><span class="o">=</span><span class="n">past_key_value</span><span class="p">,</span>
<span class="n">host_past_key_value_lengths</span><span class="o">=</span><span class="n">host_past_key_value_lengths</span><span class="p">,</span>
<span class="n">host_max_attention_window_sizes</span><span class="o">=</span><span class="n">host_max_attention_window_sizes</span><span class="p">,</span>
<span class="n">host_sink_token_length</span><span class="o">=</span><span class="n">host_sink_token_length</span><span class="p">,</span>
<span class="n">cache_indirection</span><span class="o">=</span><span class="n">cache_indirection</span><span class="p">,</span>
<span class="n">kv_cache_block_offsets</span><span class="o">=</span><span class="n">kv_cache_block_offsets</span><span class="p">,</span>
<span class="n">host_kv_cache_block_offsets</span><span class="o">=</span><span class="n">host_kv_cache_block_offsets</span><span class="p">,</span>
<span class="n">host_kv_cache_pool_pointers</span><span class="o">=</span><span class="n">host_kv_cache_pool_pointers</span><span class="p">,</span>
<span class="n">host_kv_cache_pool_mapping</span><span class="o">=</span><span class="n">host_kv_cache_pool_mapping</span><span class="p">,</span>
<span class="n">cross_kv_cache_block_offsets</span><span class="o">=</span><span class="n">cross_kv_cache_block_offsets</span><span class="p">,</span>
<span class="n">host_cross_kv_cache_block_offsets</span><span class="o">=</span>
<span class="n">host_cross_kv_cache_block_offsets</span><span class="p">,</span>
<span class="n">host_cross_kv_cache_pool_pointers</span><span class="o">=</span>
<span class="n">host_cross_kv_cache_pool_pointers</span><span class="p">,</span>
<span class="n">host_cross_kv_cache_pool_mapping</span><span class="o">=</span>
<span class="n">host_cross_kv_cache_pool_mapping</span><span class="p">,</span>
<span class="p">)</span>
<span class="n">attention_params</span> <span class="o">=</span> <span class="n">AttentionParams</span><span class="p">(</span>
<span class="n">sequence_length</span><span class="o">=</span><span class="n">sequence_length</span><span class="p">,</span>
<span class="n">context_lengths</span><span class="o">=</span><span class="n">context_lengths</span><span class="p">,</span>
<span class="n">host_context_lengths</span><span class="o">=</span><span class="n">host_context_lengths</span><span class="p">,</span>
<span class="n">max_context_length</span><span class="o">=</span><span class="n">max_decoder_input_len</span><span class="p">,</span>
<span class="n">host_request_types</span><span class="o">=</span><span class="n">host_request_types</span><span class="p">,</span>
<span class="n">host_runtime_perf_knobs</span><span class="o">=</span><span class="n">host_runtime_perf_knobs</span><span class="p">,</span>
<span class="n">host_context_progress</span><span class="o">=</span><span class="n">host_context_progress</span><span class="p">,</span>
<span class="n">encoder_input_lengths</span><span class="o">=</span><span class="n">encoder_input_lengths</span><span class="p">,</span>
<span class="n">encoder_max_input_length</span><span class="o">=</span><span class="n">encoder_max_input_length</span><span class="p">,</span>
<span class="p">)</span>
<span class="n">cross_kv_cache_gen</span> <span class="o">=</span> <span class="n">Tensor</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s1">&#39;cross_kv_cache_gen&#39;</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="n">trt</span><span class="o">.</span><span class="n">bool</span><span class="p">,</span>
<span class="n">shape</span><span class="o">=</span><span class="p">[</span><span class="mi">1</span><span class="p">],</span>
<span class="n">dim_range</span><span class="o">=</span><span class="n">OrderedDict</span><span class="p">([</span>
<span class="p">(</span><span class="s1">&#39;boolean&#39;</span><span class="p">,</span> <span class="p">[</span><span class="mi">1</span><span class="p">]),</span>
<span class="p">]))</span>
<span class="n">cross_kv_reuse</span> <span class="o">=</span> <span class="kc">None</span>
<span class="n">num_heads</span> <span class="o">=</span> <span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">transformer</span><span class="o">.</span><span class="n">num_heads</span> <span class="o">+</span> <span class="bp">self</span><span class="o">.</span><span class="n">mapping</span><span class="o">.</span><span class="n">tp_size</span> <span class="o">-</span>
<span class="mi">1</span><span class="p">)</span> <span class="o">//</span> <span class="bp">self</span><span class="o">.</span><span class="n">mapping</span><span class="o">.</span><span class="n">tp_size</span>
<span class="n">cross_kv_out_dim</span> <span class="o">=</span> <span class="mi">2</span> <span class="o">*</span> <span class="n">num_kv_heads</span> <span class="o">*</span> <span class="bp">self</span><span class="o">.</span><span class="n">transformer</span><span class="o">.</span><span class="n">head_size</span>
<span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">transformer</span><span class="o">.</span><span class="n">skip_cross_kv</span><span class="p">:</span>
<span class="k">if</span> <span class="n">remove_input_padding</span><span class="p">:</span>
<span class="n">cross_kv_reuse</span> <span class="o">=</span> <span class="n">Tensor</span><span class="p">(</span>
<span class="n">name</span><span class="o">=</span><span class="s2">&quot;cross_kv_reuse&quot;</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">_dtype</span><span class="p">,</span>
<span class="n">shape</span><span class="o">=</span><span class="p">[</span><span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="n">cross_kv_out_dim</span><span class="p">],</span>
<span class="n">dim_range</span><span class="o">=</span><span class="n">OrderedDict</span><span class="p">([</span>
<span class="p">(</span><span class="s2">&quot;encoder_num_tokens&quot;</span><span class="p">,</span> <span class="p">[</span><span class="n">encoder_num_tokens_range</span><span class="p">]),</span>
<span class="p">(</span><span class="s2">&quot;encoder_kv_size&quot;</span><span class="p">,</span> <span class="p">[</span><span class="n">cross_kv_out_dim</span><span class="p">]),</span>
<span class="p">]),</span>
<span class="p">)</span>
<span class="k">else</span><span class="p">:</span>
<span class="n">cross_kv_reuse</span> <span class="o">=</span> <span class="n">Tensor</span><span class="p">(</span>
<span class="n">name</span><span class="o">=</span><span class="s2">&quot;cross_kv_reuse&quot;</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">_dtype</span><span class="p">,</span>
<span class="n">shape</span><span class="o">=</span><span class="p">[</span><span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="n">cross_kv_out_dim</span><span class="p">],</span>
<span class="n">dim_range</span><span class="o">=</span><span class="n">OrderedDict</span><span class="p">([</span>
<span class="p">(</span><span class="s2">&quot;batch_size_beam_width_encoder&quot;</span><span class="p">,</span> <span class="p">[</span><span class="n">bb_range</span><span class="p">]),</span>
<span class="p">(</span><span class="s2">&quot;encoder_input_len&quot;</span><span class="p">,</span> <span class="p">[</span><span class="n">encoder_input_len_range</span><span class="p">]),</span>
<span class="p">(</span><span class="s2">&quot;encoder_kv_size&quot;</span><span class="p">,</span> <span class="p">[</span><span class="n">cross_kv_out_dim</span><span class="p">]),</span>
<span class="p">]),</span>
<span class="p">)</span>
<span class="n">skip_cross_attn_blocks</span> <span class="o">=</span> <span class="kc">None</span>
<span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">config</span><span class="o">.</span><span class="n">skip_cross_attn_blocks</span><span class="p">:</span>
<span class="n">skip_cross_attn_blocks</span> <span class="o">=</span> <span class="n">Tensor</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s1">&#39;skip_cross_attn_blocks&#39;</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="n">trt</span><span class="o">.</span><span class="n">bool</span><span class="p">,</span>
<span class="n">shape</span><span class="o">=</span><span class="p">[</span><span class="mi">1</span><span class="p">],</span>
<span class="n">dim_range</span><span class="o">=</span><span class="n">OrderedDict</span><span class="p">([</span>
<span class="p">(</span><span class="s1">&#39;boolean&#39;</span><span class="p">,</span> <span class="p">[</span><span class="mi">1</span><span class="p">]),</span>
<span class="p">]))</span>
<span class="n">prompt_embedding_table</span> <span class="o">=</span> <span class="kc">None</span>
<span class="n">tasks</span> <span class="o">=</span> <span class="kc">None</span>
<span class="n">prompt_vocab_size</span> <span class="o">=</span> <span class="kc">None</span>
<span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">mapping</span><span class="o">.</span><span class="n">is_first_pp_rank</span><span class="p">()</span> <span class="ow">and</span> <span class="n">prompt_embedding_table_size</span> <span class="o">&gt;</span> <span class="mi">0</span><span class="p">:</span>
<span class="n">p_embedding_range</span> <span class="o">=</span> <span class="p">[[</span>
<span class="mi">1</span><span class="p">,</span> <span class="n">prompt_embedding_table_size</span> <span class="o">//</span> <span class="mi">2</span><span class="p">,</span> <span class="n">prompt_embedding_table_size</span>
<span class="p">]]</span>
<span class="n">prompt_embedding_table</span> <span class="o">=</span> <span class="n">Tensor</span><span class="p">(</span>
<span class="n">name</span><span class="o">=</span><span class="s1">&#39;prompt_embedding_table&#39;</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">_dtype</span><span class="p">,</span>
<span class="n">shape</span><span class="o">=</span><span class="p">[</span><span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">transformer</span><span class="o">.</span><span class="n">hidden_size</span><span class="p">],</span>
<span class="n">dim_range</span><span class="o">=</span><span class="n">OrderedDict</span><span class="p">([</span>
<span class="p">(</span><span class="s1">&#39;prompt_embedding_table_size&#39;</span><span class="p">,</span> <span class="n">p_embedding_range</span><span class="p">),</span>
<span class="p">(</span><span class="s1">&#39;hidden_size&#39;</span><span class="p">,</span> <span class="p">[</span><span class="bp">self</span><span class="o">.</span><span class="n">transformer</span><span class="o">.</span><span class="n">hidden_size</span><span class="p">]),</span>
<span class="p">]))</span>
<span class="k">if</span> <span class="n">remove_input_padding</span><span class="p">:</span>
<span class="n">num_tokens_range</span> <span class="o">=</span> <span class="p">[</span>
<span class="mi">1</span><span class="p">,</span>
<span class="p">(</span><span class="n">max_decoder_input_len</span> <span class="o">*</span> <span class="n">max_batch_size</span> <span class="o">+</span> <span class="mi">1</span><span class="p">)</span> <span class="o">//</span> <span class="mi">2</span><span class="p">,</span>
<span class="n">max_decoder_input_len</span> <span class="o">*</span> <span class="n">max_batch_size</span><span class="p">,</span>
<span class="p">]</span>
<span class="n">tasks</span> <span class="o">=</span> <span class="n">Tensor</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s1">&#39;tasks&#39;</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="n">trt</span><span class="o">.</span><span class="n">int32</span><span class="p">,</span>
<span class="n">shape</span><span class="o">=</span><span class="p">[</span><span class="o">-</span><span class="mi">1</span><span class="p">],</span>
<span class="n">dim_range</span><span class="o">=</span><span class="n">OrderedDict</span><span class="p">([</span>
<span class="p">(</span><span class="s1">&#39;decoder_num_tokens&#39;</span><span class="p">,</span>
<span class="p">[</span><span class="n">decoder_num_tokens_range</span><span class="p">]),</span>
<span class="p">]))</span>
<span class="k">else</span><span class="p">:</span>
<span class="n">tasks</span> <span class="o">=</span> <span class="n">Tensor</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s1">&#39;tasks&#39;</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="n">trt</span><span class="o">.</span><span class="n">int32</span><span class="p">,</span>
<span class="n">shape</span><span class="o">=</span><span class="p">[</span><span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">],</span>
<span class="n">dim_range</span><span class="o">=</span><span class="n">OrderedDict</span><span class="p">([</span>
<span class="p">(</span><span class="s1">&#39;batch_size&#39;</span><span class="p">,</span> <span class="n">bs_range</span><span class="p">),</span>
<span class="p">(</span><span class="s1">&#39;broadcast_dim&#39;</span><span class="p">,</span> <span class="p">[</span><span class="mi">1</span><span class="p">]),</span>
<span class="p">]))</span>
<span class="n">prompt_vocab_size</span> <span class="o">=</span> <span class="n">Tensor</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s1">&#39;prompt_vocab_size&#39;</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="n">trt</span><span class="o">.</span><span class="n">int32</span><span class="p">,</span>
<span class="n">shape</span><span class="o">=</span><span class="p">[</span><span class="mi">1</span><span class="p">],</span>
<span class="n">dim_range</span><span class="o">=</span><span class="n">OrderedDict</span><span class="p">([(</span><span class="s1">&#39;size&#39;</span><span class="p">,</span> <span class="p">[</span><span class="mi">1</span><span class="p">])]))</span>
<span class="n">result</span> <span class="o">=</span> <span class="p">{</span>
<span class="s1">&#39;decoder_input_ids&#39;</span><span class="p">:</span> <span class="n">input_ids</span><span class="p">,</span>
<span class="s1">&#39;encoder_output&#39;</span><span class="p">:</span> <span class="n">encoder_output</span><span class="p">,</span>
<span class="s1">&#39;use_cache&#39;</span><span class="p">:</span> <span class="kc">True</span><span class="p">,</span>
<span class="s1">&#39;attention_mask_params&#39;</span><span class="p">:</span> <span class="n">attention_mask_params</span><span class="p">,</span>
<span class="s1">&#39;last_token_ids&#39;</span><span class="p">:</span> <span class="n">last_token_ids</span><span class="p">,</span>
<span class="s1">&#39;kv_cache_params&#39;</span><span class="p">:</span> <span class="n">kv_cache_params</span><span class="p">,</span>
<span class="s1">&#39;attention_params&#39;</span><span class="p">:</span> <span class="n">attention_params</span><span class="p">,</span>
<span class="s1">&#39;hidden_states&#39;</span><span class="p">:</span> <span class="n">hidden_states</span><span class="p">,</span>
<span class="s1">&#39;lora_params&#39;</span><span class="p">:</span> <span class="n">lora_params</span><span class="p">,</span>
<span class="s1">&#39;cross_kv_cache_gen&#39;</span><span class="p">:</span> <span class="n">cross_kv_cache_gen</span><span class="p">,</span>
<span class="s1">&#39;cross_kv_reuse&#39;</span><span class="p">:</span> <span class="n">cross_kv_reuse</span><span class="p">,</span>
<span class="s1">&#39;prompt_embedding_table&#39;</span><span class="p">:</span> <span class="n">prompt_embedding_table</span><span class="p">,</span>
<span class="s1">&#39;prompt_tasks&#39;</span><span class="p">:</span> <span class="n">tasks</span><span class="p">,</span>
<span class="s1">&#39;prompt_vocab_size&#39;</span><span class="p">:</span> <span class="n">prompt_vocab_size</span><span class="p">,</span>
<span class="s1">&#39;skip_cross_attn_blocks&#39;</span><span class="p">:</span> <span class="n">skip_cross_attn_blocks</span><span class="p">,</span>
<span class="p">}</span>
<span class="k">return</span> <span class="n">result</span></div>
<div class="viewcode-block" id="MLLaMAForCausalLM.use_lora">
<a class="viewcode-back" href="../../../../python-api/tensorrt_llm.models.html#tensorrt_llm.models.MLLaMAForCausalLM.use_lora">[docs]</a>
<span class="k">def</span><span class="w"> </span><span class="nf">use_lora</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">lora_config</span><span class="p">:</span> <span class="n">LoraConfig</span><span class="p">):</span>
<span class="n">use_lora</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">lora_config</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">trtllm_modules_to_hf_modules</span><span class="p">)</span></div>
<div class="viewcode-block" id="MLLaMAForCausalLM.from_hugging_face">
<a class="viewcode-back" href="../../../../python-api/tensorrt_llm.models.html#tensorrt_llm.models.MLLaMAForCausalLM.from_hugging_face">[docs]</a>
<span class="nd">@classmethod</span>
<span class="k">def</span><span class="w"> </span><span class="nf">from_hugging_face</span><span class="p">(</span>
<span class="bp">cls</span><span class="p">,</span>
<span class="n">hf_model_or_dir</span><span class="p">:</span> <span class="n">Union</span><span class="p">[</span><span class="nb">str</span><span class="p">,</span> <span class="s1">&#39;transformers.PreTrainedModel&#39;</span><span class="p">],</span>
<span class="n">dtype</span><span class="p">:</span> <span class="nb">str</span> <span class="o">=</span> <span class="s1">&#39;auto&#39;</span><span class="p">,</span>
<span class="n">mapping</span><span class="p">:</span> <span class="n">Optional</span><span class="p">[</span><span class="n">Mapping</span><span class="p">]</span> <span class="o">=</span> <span class="kc">None</span><span class="p">,</span>
<span class="n">quant_config</span><span class="p">:</span> <span class="n">Optional</span><span class="p">[</span><span class="n">QuantConfig</span><span class="p">]</span> <span class="o">=</span> <span class="kc">None</span><span class="p">,</span>
<span class="o">**</span><span class="n">kwargs</span><span class="p">):</span>
<span class="w"> </span><span class="sd">&#39;&#39;&#39; Create a MLLaMAForCausalLM object from give parameters</span>
<span class="sd"> &#39;&#39;&#39;</span>
<span class="kn">import</span><span class="w"> </span><span class="nn">transformers</span>
<span class="n">kwargs</span><span class="o">.</span><span class="n">pop</span><span class="p">(</span><span class="s1">&#39;load_by_shard&#39;</span><span class="p">,</span> <span class="kc">False</span><span class="p">)</span>
<span class="n">kwargs</span><span class="o">.</span><span class="n">pop</span><span class="p">(</span><span class="s1">&#39;load_model_on_cpu&#39;</span><span class="p">,</span> <span class="kc">False</span><span class="p">)</span>
<span class="n">quant_ckpt_path</span> <span class="o">=</span> <span class="n">kwargs</span><span class="o">.</span><span class="n">pop</span><span class="p">(</span><span class="s1">&#39;quant_ckpt_path&#39;</span><span class="p">,</span> <span class="kc">None</span><span class="p">)</span>
<span class="k">assert</span> <span class="n">hf_model_or_dir</span> <span class="ow">is</span> <span class="ow">not</span> <span class="kc">None</span>
<span class="n">use_preloading</span> <span class="o">=</span> <span class="nb">isinstance</span><span class="p">(</span><span class="n">hf_model_or_dir</span><span class="p">,</span>
<span class="n">transformers</span><span class="o">.</span><span class="n">PreTrainedModel</span><span class="p">)</span>
<span class="k">if</span> <span class="n">use_preloading</span><span class="p">:</span>
<span class="n">hf_model</span> <span class="o">=</span> <span class="n">hf_model_or_dir</span>
<span class="n">hf_config_or_dir</span> <span class="o">=</span> <span class="n">hf_model</span><span class="o">.</span><span class="n">config</span>
<span class="k">else</span><span class="p">:</span>
<span class="n">hf_model_dir</span> <span class="o">=</span> <span class="n">hf_model_or_dir</span>
<span class="n">hf_config_or_dir</span> <span class="o">=</span> <span class="n">hf_model_or_dir</span>
<span class="n">config</span> <span class="o">=</span> <span class="n">MLLaMAConfig</span><span class="o">.</span><span class="n">from_hugging_face</span><span class="p">(</span><span class="n">hf_config_or_dir</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="n">dtype</span><span class="p">,</span>
<span class="n">mapping</span><span class="o">=</span><span class="n">mapping</span><span class="p">,</span>
<span class="n">quant_config</span><span class="o">=</span><span class="n">quant_config</span><span class="p">,</span>
<span class="o">**</span><span class="n">kwargs</span><span class="p">)</span>
<span class="n">custom_dict</span> <span class="o">=</span> <span class="p">{</span>
<span class="s2">&quot;lm_head&quot;</span><span class="p">:</span> <span class="s2">&quot;language_model.lm_head&quot;</span><span class="p">,</span>
<span class="s2">&quot;transformer.ln_f&quot;</span><span class="p">:</span> <span class="s2">&quot;language_model.model.norm&quot;</span><span class="p">,</span>
<span class="s2">&quot;transformer&quot;</span><span class="p">:</span> <span class="s2">&quot;language_model.model&quot;</span><span class="p">,</span>
<span class="s2">&quot;self_attention&quot;</span><span class="p">:</span> <span class="s2">&quot;self_attn&quot;</span><span class="p">,</span>
<span class="s2">&quot;cross_attention&quot;</span><span class="p">:</span> <span class="s2">&quot;cross_attn&quot;</span><span class="p">,</span>
<span class="s2">&quot;vocab_embedding&quot;</span><span class="p">:</span> <span class="s2">&quot;embed_tokens&quot;</span><span class="p">,</span>
<span class="s2">&quot;gate_attn&quot;</span><span class="p">:</span> <span class="s2">&quot;cross_attn_attn_gate&quot;</span><span class="p">,</span>
<span class="s2">&quot;gate_ffwd&quot;</span><span class="p">:</span> <span class="s2">&quot;cross_attn_mlp_gate&quot;</span><span class="p">,</span>
<span class="s2">&quot;q_layernorm&quot;</span><span class="p">:</span> <span class="s2">&quot;q_norm&quot;</span><span class="p">,</span>
<span class="s2">&quot;k_layernorm&quot;</span><span class="p">:</span> <span class="s2">&quot;k_norm&quot;</span><span class="p">,</span>
<span class="p">}</span>
<span class="k">if</span> <span class="n">quant_ckpt_path</span> <span class="ow">is</span> <span class="ow">not</span> <span class="kc">None</span><span class="p">:</span>
<span class="n">hf_model_dir</span> <span class="o">=</span> <span class="n">quant_ckpt_path</span>
<span class="n">loader</span> <span class="o">=</span> <span class="n">ModelWeightsLoader</span><span class="p">(</span><span class="n">hf_model_dir</span><span class="p">,</span> <span class="n">custom_dict</span><span class="p">)</span>
<span class="n">model</span> <span class="o">=</span> <span class="bp">cls</span><span class="p">(</span><span class="n">config</span><span class="p">)</span>
<span class="n">loader</span><span class="o">.</span><span class="n">generate_tllm_weights</span><span class="p">(</span><span class="n">model</span><span class="p">)</span>
<span class="k">return</span> <span class="n">model</span></div>
</div>
</pre></div>
</article>
<footer class="prev-next-footer d-print-none">
<div class="prev-next-area">
</div>
</footer>
</div>
<div class="bd-sidebar-secondary"></div>
</div>
<footer class="bd-footer-content">
</footer>
</main>
</div>
</div>
<!-- Scripts loaded after <body> so the DOM is not blocked -->
<script defer src="../../../../_static/scripts/bootstrap.js?digest=8878045cc6db502f8baf"></script>
<script defer src="../../../../_static/scripts/pydata-sphinx-theme.js?digest=8878045cc6db502f8baf"></script>
<footer class="bd-footer">
<div class="bd-footer__inner bd-page-width">
<div class="footer-items__start">
<div class="footer-item">
<a class="footer-brand logo" href="https://www.nvidia.com">
<img src="../../../../_static/nvidia-logo-horiz-rgb-1c-blk-for-screen.svg" class="logo__image only-light" alt="NVIDIA"/>
<img src="../../../../_static/nvidia-logo-horiz-rgb-1c-wht-for-screen.svg" class="logo__image only-dark" alt="NVIDIA"/>
</a></div>
<div class="footer-item">
<div class="footer-links">
<a class="external" href="https://www.nvidia.com/en-us/about-nvidia/privacy-policy/">Privacy Policy</a>
|
<a class="external" href="https://www.nvidia.com/en-us/about-nvidia/privacy-center/">Manage My Privacy</a>
|
<a class="external" href="https://www.nvidia.com/en-us/preferences/start/">Do Not Sell or Share My Data</a>
|
<a class="external" href="https://www.nvidia.com/en-us/about-nvidia/terms-of-service/">Terms of Service</a>
|
<a class="external" href="https://www.nvidia.com/en-us/about-nvidia/accessibility/">Accessibility</a>
|
<a class="external" href="https://www.nvidia.com/en-us/about-nvidia/company-policies/">Corporate Policies</a>
|
<a class="external" href="https://www.nvidia.com/en-us/product-security/">Product Security</a>
|
<a class="external" href="https://www.nvidia.com/en-us/contact/">Contact</a>
</div>
</div>
<div class="footer-item">
<p class="copyright">
Copyright © 2025, NVidia.
<br/>
</p>
</div>
<div class="footer-item">
<div class="extra_footer">
<p>Last updated on August 06, 2025.</p>
<p>This page is generated by TensorRT-LLM commit <a href="https://github.com/NVIDIA/TensorRT-LLM/tree/a16ba64">a16ba64</a>.</p>
</div></div>
</div>
</div>
</footer>
</body>
</html>
Reference in New Issue View Git Blame Copy Permalink