mirror of
https://github.com/NVIDIA/TensorRT-LLM.git
synced 2026-01-14 06:27:45 +08:00
381 lines
39 KiB
HTML
381 lines
39 KiB
HTML
<!DOCTYPE html>
|
|
<html class="writer-html5" lang="en" data-content_root="../../../">
|
|
<head>
|
|
<meta charset="utf-8" />
|
|
<meta name="viewport" content="width=device-width, initial-scale=1.0" />
|
|
<title>tensorrt_llm.layers.mlp — tensorrt_llm documentation</title>
|
|
<link rel="stylesheet" type="text/css" href="../../../_static/pygments.css?v=80d5e7a1" />
|
|
<link rel="stylesheet" type="text/css" href="../../../_static/css/theme.css?v=19f00094" />
|
|
|
|
|
|
<!--[if lt IE 9]>
|
|
<script src="../../../_static/js/html5shiv.min.js"></script>
|
|
<![endif]-->
|
|
|
|
<script src="../../../_static/jquery.js?v=5d32c60e"></script>
|
|
<script src="../../../_static/_sphinx_javascript_frameworks_compat.js?v=2cd50e6c"></script>
|
|
<script src="../../../_static/documentation_options.js?v=5929fcd5"></script>
|
|
<script src="../../../_static/doctools.js?v=888ff710"></script>
|
|
<script src="../../../_static/sphinx_highlight.js?v=dc90522c"></script>
|
|
<script src="../../../_static/js/theme.js"></script>
|
|
<link rel="index" title="Index" href="../../../genindex.html" />
|
|
<link rel="search" title="Search" href="../../../search.html" />
|
|
</head>
|
|
|
|
<body class="wy-body-for-nav">
|
|
<div class="wy-grid-for-nav">
|
|
<nav data-toggle="wy-nav-shift" class="wy-nav-side">
|
|
<div class="wy-side-scroll">
|
|
<div class="wy-side-nav-search" >
|
|
|
|
|
|
|
|
<a href="../../../index.html" class="icon icon-home">
|
|
tensorrt_llm
|
|
</a>
|
|
<div role="search">
|
|
<form id="rtd-search-form" class="wy-form" action="../../../search.html" method="get">
|
|
<input type="text" name="q" placeholder="Search docs" aria-label="Search docs" />
|
|
<input type="hidden" name="check_keywords" value="yes" />
|
|
<input type="hidden" name="area" value="default" />
|
|
</form>
|
|
</div>
|
|
</div><div class="wy-menu wy-menu-vertical" data-spy="affix" role="navigation" aria-label="Navigation menu">
|
|
<p class="caption" role="heading"><span class="caption-text">Contents:</span></p>
|
|
<ul>
|
|
<li class="toctree-l1"><a class="reference internal" href="../../../architecture.html">TensorRT-LLM Architecture</a></li>
|
|
<li class="toctree-l1"><a class="reference internal" href="../../../gpt_runtime.html">C++ GPT Runtime</a></li>
|
|
<li class="toctree-l1"><a class="reference internal" href="../../../batch_manager.html">The Batch Manager in TensorRT-LLM</a></li>
|
|
<li class="toctree-l1"><a class="reference internal" href="../../../gpt_attention.html">Multi-head, Multi-query and Group-query Attention</a></li>
|
|
<li class="toctree-l1"><a class="reference internal" href="../../../precision.html">Numerical Precision</a></li>
|
|
<li class="toctree-l1"><a class="reference internal" href="../../../installation.html">Build TensorRT-LLM</a></li>
|
|
<li class="toctree-l1"><a class="reference internal" href="../../../performance.html">Performance of TensorRT-LLM</a></li>
|
|
<li class="toctree-l1"><a class="reference internal" href="../../../2023-05-19-how-to-debug.html">How to debug</a></li>
|
|
<li class="toctree-l1"><a class="reference internal" href="../../../2023-05-17-how-to-add-a-new-model.html">How to add a new model</a></li>
|
|
<li class="toctree-l1"><a class="reference internal" href="../../../graph-rewriting.html">Graph Rewriting Module</a></li>
|
|
<li class="toctree-l1"><a class="reference internal" href="../../../memory.html">Memory Usage of TensorRT-LLM</a></li>
|
|
</ul>
|
|
<p class="caption" role="heading"><span class="caption-text">Python API</span></p>
|
|
<ul>
|
|
<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 class="caption" role="heading"><span class="caption-text">C++ API</span></p>
|
|
<ul>
|
|
<li class="toctree-l1"><a class="reference internal" href="../../../_cpp_gen/runtime.html">Runtime</a></li>
|
|
</ul>
|
|
<p class="caption" role="heading"><span class="caption-text">Blogs</span></p>
|
|
<ul>
|
|
<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>
|
|
</ul>
|
|
|
|
</div>
|
|
</div>
|
|
</nav>
|
|
|
|
<section data-toggle="wy-nav-shift" class="wy-nav-content-wrap"><nav class="wy-nav-top" aria-label="Mobile navigation menu" >
|
|
<i data-toggle="wy-nav-top" class="fa fa-bars"></i>
|
|
<a href="../../../index.html">tensorrt_llm</a>
|
|
</nav>
|
|
|
|
<div class="wy-nav-content">
|
|
<div class="rst-content">
|
|
<div role="navigation" aria-label="Page navigation">
|
|
<ul class="wy-breadcrumbs">
|
|
<li><a href="../../../index.html" class="icon icon-home" aria-label="Home"></a></li>
|
|
<li class="breadcrumb-item"><a href="../../index.html">Module code</a></li>
|
|
<li class="breadcrumb-item active">tensorrt_llm.layers.mlp</li>
|
|
<li class="wy-breadcrumbs-aside">
|
|
</li>
|
|
</ul>
|
|
<hr/>
|
|
</div>
|
|
<div role="main" class="document" itemscope="itemscope" itemtype="http://schema.org/Article">
|
|
<div itemprop="articleBody">
|
|
|
|
<h1>Source code for tensorrt_llm.layers.mlp</h1><div class="highlight"><pre>
|
|
<span></span><span class="c1"># SPDX-FileCopyrightText: Copyright (c) 2022-2023 NVIDIA CORPORATION & 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 "License");</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 "AS IS" 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="nn">numpy</span> <span class="k">as</span> <span class="nn">np</span>
|
|
|
|
<span class="kn">from</span> <span class="nn">.._utils</span> <span class="kn">import</span> <span class="n">trt_dtype_to_np</span>
|
|
<span class="kn">from</span> <span class="nn">..functional</span> <span class="kn">import</span> <span class="n">ACT2FN</span>
|
|
<span class="kn">from</span> <span class="nn">..module</span> <span class="kn">import</span> <span class="n">Module</span>
|
|
<span class="kn">from</span> <span class="nn">..quantization</span> <span class="kn">import</span> <span class="n">QuantMode</span>
|
|
<span class="kn">from</span> <span class="nn">..quantization.layers</span> <span class="kn">import</span> <span class="n">FP8Linear</span><span class="p">,</span> <span class="n">FP8RowLinear</span>
|
|
<span class="kn">from</span> <span class="nn">.linear</span> <span class="kn">import</span> <span class="n">ColumnLinear</span><span class="p">,</span> <span class="n">RowLinear</span>
|
|
|
|
|
|
<div class="viewcode-block" id="MLP">
|
|
<a class="viewcode-back" href="../../../python-api/tensorrt_llm.layers.html#tensorrt_llm.layers.mlp.MLP">[docs]</a>
|
|
<span class="k">class</span> <span class="nc">MLP</span><span class="p">(</span><span class="n">Module</span><span class="p">):</span>
|
|
|
|
<span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</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">hidden_act</span><span class="p">,</span>
|
|
<span class="n">bias</span><span class="o">=</span><span class="kc">True</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">tp_group</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span>
|
|
<span class="n">tp_size</span><span class="o">=</span><span class="mi">1</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="n">instance_id</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="nb">super</span><span class="p">()</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span>
|
|
<span class="k">if</span> <span class="n">hidden_act</span> <span class="ow">not</span> <span class="ow">in</span> <span class="n">ACT2FN</span><span class="p">:</span>
|
|
<span class="k">raise</span> <span class="ne">ValueError</span><span class="p">(</span>
|
|
<span class="s1">'unsupported activation function: </span><span class="si">{}</span><span class="s1">'</span><span class="o">.</span><span class="n">format</span><span class="p">(</span><span class="n">hidden_act</span><span class="p">))</span>
|
|
<span class="n">fc_output_size</span> <span class="o">=</span> <span class="mi">2</span> <span class="o">*</span> <span class="n">ffn_hidden_size</span> <span class="k">if</span> <span class="n">hidden_act</span> <span class="o">==</span> <span class="s1">'swiglu'</span> <span class="k">else</span> <span class="n">ffn_hidden_size</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">use_fp8_qdq</span> <span class="o">=</span> <span class="n">quant_mode</span><span class="o">.</span><span class="n">has_fp8_qdq</span><span class="p">()</span>
|
|
|
|
<span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">use_fp8_qdq</span><span class="p">:</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">fc</span> <span class="o">=</span> <span class="n">FP8Linear</span><span class="p">(</span><span class="n">hidden_size</span><span class="p">,</span>
|
|
<span class="n">fc_output_size</span><span class="p">,</span>
|
|
<span class="n">bias</span><span class="o">=</span><span class="n">bias</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">tp_group</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">tp_size</span><span class="p">,</span>
|
|
<span class="n">gather_output</span><span class="o">=</span><span class="kc">False</span><span class="p">)</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">proj</span> <span class="o">=</span> <span class="n">FP8RowLinear</span><span class="p">(</span><span class="n">ffn_hidden_size</span><span class="p">,</span>
|
|
<span class="n">hidden_size</span><span class="p">,</span>
|
|
<span class="n">bias</span><span class="o">=</span><span class="n">bias</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">tp_group</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">tp_size</span><span class="p">,</span>
|
|
<span class="n">instance_id</span><span class="o">=</span><span class="n">instance_id</span><span class="p">)</span>
|
|
<span class="k">else</span><span class="p">:</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">fc</span> <span class="o">=</span> <span class="n">ColumnLinear</span><span class="p">(</span><span class="n">hidden_size</span><span class="p">,</span>
|
|
<span class="n">fc_output_size</span><span class="p">,</span>
|
|
<span class="n">bias</span><span class="o">=</span><span class="n">bias</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">tp_group</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">tp_size</span><span class="p">,</span>
|
|
<span class="n">gather_output</span><span class="o">=</span><span class="kc">False</span><span class="p">)</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">proj</span> <span class="o">=</span> <span class="n">RowLinear</span><span class="p">(</span><span class="n">ffn_hidden_size</span><span class="p">,</span>
|
|
<span class="n">hidden_size</span><span class="p">,</span>
|
|
<span class="n">bias</span><span class="o">=</span><span class="n">bias</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">tp_group</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">tp_size</span><span class="p">,</span>
|
|
<span class="n">instance_id</span><span class="o">=</span><span class="n">instance_id</span><span class="p">)</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">hidden_act</span> <span class="o">=</span> <span class="n">hidden_act</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">dtype</span> <span class="o">=</span> <span class="n">dtype</span>
|
|
|
|
<div class="viewcode-block" id="MLP.forward">
|
|
<a class="viewcode-back" href="../../../python-api/tensorrt_llm.layers.html#tensorrt_llm.layers.mlp.MLP.forward">[docs]</a>
|
|
<span class="k">def</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">workspace</span><span class="o">=</span><span class="kc">None</span><span class="p">):</span>
|
|
<span class="n">inter</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">fc</span><span class="p">(</span><span class="n">hidden_states</span><span class="p">)</span>
|
|
<span class="n">inter</span> <span class="o">=</span> <span class="n">ACT2FN</span><span class="p">[</span><span class="bp">self</span><span class="o">.</span><span class="n">hidden_act</span><span class="p">](</span><span class="n">inter</span><span class="p">)</span>
|
|
<span class="n">output</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">proj</span><span class="p">(</span><span class="n">inter</span><span class="p">,</span> <span class="n">workspace</span><span class="p">)</span>
|
|
<span class="k">return</span> <span class="n">output</span></div>
|
|
</div>
|
|
|
|
|
|
|
|
<div class="viewcode-block" id="GatedMLP">
|
|
<a class="viewcode-back" href="../../../python-api/tensorrt_llm.layers.html#tensorrt_llm.layers.mlp.GatedMLP">[docs]</a>
|
|
<span class="k">class</span> <span class="nc">GatedMLP</span><span class="p">(</span><span class="n">MLP</span><span class="p">):</span>
|
|
|
|
<span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</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">hidden_act</span><span class="p">,</span>
|
|
<span class="n">bias</span><span class="o">=</span><span class="kc">True</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">tp_group</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span>
|
|
<span class="n">tp_size</span><span class="o">=</span><span class="mi">1</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="n">instance_id</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="bp">self</span><span class="o">.</span><span class="n">use_fp8_qdq</span> <span class="o">=</span> <span class="n">quant_mode</span><span class="o">.</span><span class="n">has_fp8_qdq</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">hidden_size</span><span class="p">,</span>
|
|
<span class="n">ffn_hidden_size</span><span class="p">,</span>
|
|
<span class="n">hidden_act</span><span class="p">,</span>
|
|
<span class="n">bias</span><span class="o">=</span><span class="n">bias</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">tp_group</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">tp_size</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="n">instance_id</span><span class="o">=</span><span class="n">instance_id</span><span class="p">)</span>
|
|
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">hidden_size</span> <span class="o">=</span> <span class="n">hidden_size</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">ffn_hidden_size</span> <span class="o">=</span> <span class="n">ffn_hidden_size</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">hidden_act</span> <span class="o">=</span> <span class="n">hidden_act</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">bias</span> <span class="o">=</span> <span class="n">bias</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">dtype</span> <span class="o">=</span> <span class="n">dtype</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">tp_group</span> <span class="o">=</span> <span class="n">tp_group</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">tp_size</span> <span class="o">=</span> <span class="n">tp_size</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">quant_mode</span> <span class="o">=</span> <span class="n">quant_mode</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">instance_id</span> <span class="o">=</span> <span class="n">instance_id</span>
|
|
|
|
<span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">use_fp8_qdq</span><span class="p">:</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">gate</span> <span class="o">=</span> <span class="n">FP8Linear</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">bias</span><span class="o">=</span><span class="n">bias</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">tp_group</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">tp_size</span><span class="p">,</span>
|
|
<span class="n">gather_output</span><span class="o">=</span><span class="kc">False</span><span class="p">)</span>
|
|
<span class="k">else</span><span class="p">:</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">gate</span> <span class="o">=</span> <span class="n">ColumnLinear</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">bias</span><span class="o">=</span><span class="n">bias</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">tp_group</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">tp_size</span><span class="p">,</span>
|
|
<span class="n">gather_output</span><span class="o">=</span><span class="kc">False</span><span class="p">)</span>
|
|
|
|
<div class="viewcode-block" id="GatedMLP.forward">
|
|
<a class="viewcode-back" href="../../../python-api/tensorrt_llm.layers.html#tensorrt_llm.layers.mlp.GatedMLP.forward">[docs]</a>
|
|
<span class="k">def</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">workspace</span><span class="o">=</span><span class="kc">None</span><span class="p">):</span>
|
|
<span class="n">inter</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">fc</span><span class="p">(</span><span class="n">hidden_states</span><span class="p">)</span>
|
|
<span class="n">inter</span> <span class="o">=</span> <span class="n">ACT2FN</span><span class="p">[</span><span class="bp">self</span><span class="o">.</span><span class="n">hidden_act</span><span class="p">](</span><span class="n">inter</span><span class="p">)</span>
|
|
<span class="n">gate</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">gate</span><span class="p">(</span><span class="n">hidden_states</span><span class="p">)</span>
|
|
<span class="n">output</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">proj</span><span class="p">(</span><span class="n">inter</span> <span class="o">*</span> <span class="n">gate</span><span class="p">,</span> <span class="n">workspace</span><span class="p">)</span>
|
|
<span class="k">return</span> <span class="n">output</span></div>
|
|
</div>
|
|
|
|
|
|
|
|
<div class="viewcode-block" id="FusedGatedMLP">
|
|
<a class="viewcode-back" href="../../../python-api/tensorrt_llm.layers.html#tensorrt_llm.layers.mlp.FusedGatedMLP">[docs]</a>
|
|
<span class="k">class</span> <span class="nc">FusedGatedMLP</span><span class="p">(</span><span class="n">GatedMLP</span><span class="p">):</span>
|
|
|
|
<span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</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">hidden_act</span><span class="p">,</span>
|
|
<span class="n">bias</span><span class="o">=</span><span class="kc">True</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">tp_group</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span>
|
|
<span class="n">tp_size</span><span class="o">=</span><span class="mi">1</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="n">instance_id</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="nb">super</span><span class="p">()</span><span class="o">.</span><span class="fm">__init__</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">hidden_act</span><span class="p">,</span>
|
|
<span class="n">bias</span><span class="o">=</span><span class="n">bias</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">tp_group</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">tp_size</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="n">instance_id</span><span class="o">=</span><span class="n">instance_id</span><span class="p">)</span>
|
|
|
|
<div class="viewcode-block" id="FusedGatedMLP.forward">
|
|
<a class="viewcode-back" href="../../../python-api/tensorrt_llm.layers.html#tensorrt_llm.layers.mlp.FusedGatedMLP.forward">[docs]</a>
|
|
<span class="k">def</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">workspace</span><span class="o">=</span><span class="kc">None</span><span class="p">):</span>
|
|
<span class="c1"># Combine the following pattern</span>
|
|
<span class="c1">#</span>
|
|
<span class="c1"># SiLU(FC(x)) + Gate(x)</span>
|
|
<span class="c1">#</span>
|
|
<span class="c1"># into:</span>
|
|
<span class="c1">#</span>
|
|
<span class="c1"># SwiGLU(FusedFC(x))</span>
|
|
<span class="c1">#</span>
|
|
<span class="c1"># Upside is we don't need to modify 4 different weight loading paths just to concat weights</span>
|
|
|
|
<span class="n">_np_dtype</span> <span class="o">=</span> <span class="n">trt_dtype_to_np</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="n">concat_weight</span> <span class="o">=</span> <span class="n">np</span><span class="o">.</span><span class="n">concatenate</span><span class="p">(</span>
|
|
<span class="p">[</span><span class="bp">self</span><span class="o">.</span><span class="n">gate</span><span class="o">.</span><span class="n">weight</span><span class="o">.</span><span class="n">raw_value</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">fc</span><span class="o">.</span><span class="n">weight</span><span class="o">.</span><span class="n">raw_value</span><span class="p">],</span>
|
|
<span class="n">axis</span><span class="o">=</span><span class="mi">0</span><span class="p">)</span><span class="o">.</span><span class="n">astype</span><span class="p">(</span><span class="n">_np_dtype</span><span class="p">)</span>
|
|
<span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">bias</span><span class="p">:</span>
|
|
<span class="n">concat_bias</span> <span class="o">=</span> <span class="n">np</span><span class="o">.</span><span class="n">concatenate</span><span class="p">(</span>
|
|
<span class="p">[</span><span class="bp">self</span><span class="o">.</span><span class="n">gate</span><span class="o">.</span><span class="n">bias</span><span class="o">.</span><span class="n">raw_value</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">fc</span><span class="o">.</span><span class="n">bias</span><span class="o">.</span><span class="n">raw_value</span><span class="p">],</span>
|
|
<span class="n">axis</span><span class="o">=</span><span class="mi">0</span><span class="p">)</span><span class="o">.</span><span class="n">astype</span><span class="p">(</span><span class="n">_np_dtype</span><span class="p">)</span>
|
|
|
|
<span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">use_fp8_qdq</span><span class="p">:</span>
|
|
<span class="n">gate_weights_scaling_factor</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">gate</span><span class="o">.</span><span class="n">weights_scaling_factor</span><span class="o">.</span><span class="n">raw_value</span>
|
|
<span class="n">fc_weights_scaling_factor</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">fc</span><span class="o">.</span><span class="n">weights_scaling_factor</span><span class="o">.</span><span class="n">raw_value</span>
|
|
<span class="n">fc_activation_scaling_factor</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">fc</span><span class="o">.</span><span class="n">activation_scaling_factor</span><span class="o">.</span><span class="n">raw_value</span>
|
|
<span class="n">gate_activation_scaling_factor</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">gate</span><span class="o">.</span><span class="n">activation_scaling_factor</span><span class="o">.</span><span class="n">raw_value</span>
|
|
<span class="k">assert</span> <span class="n">fc_activation_scaling_factor</span> <span class="o">==</span> <span class="n">gate_activation_scaling_factor</span><span class="p">,</span> <span class="s2">"Activation scales should be identical"</span>
|
|
|
|
<span class="c1"># Remove dangling TRT-LLM parameter references after the graph rewrite.</span>
|
|
<span class="k">for</span> <span class="n">param</span><span class="p">,</span> <span class="n">_</span> <span class="ow">in</span> <span class="nb">list</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">gate</span><span class="o">.</span><span class="n">named_parameters</span><span class="p">()):</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">gate</span><span class="o">.</span><span class="n">_parameters</span><span class="o">.</span><span class="n">pop</span><span class="p">(</span><span class="n">param</span><span class="p">)</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">gate</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">use_fp8_qdq</span><span class="p">:</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">fc</span> <span class="o">=</span> <span class="n">FP8Linear</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="bp">self</span><span class="o">.</span><span class="n">ffn_hidden_size</span> <span class="o">*</span> <span class="mi">2</span><span class="p">,</span>
|
|
<span class="n">bias</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">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">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">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">tp_size</span><span class="p">,</span>
|
|
<span class="n">gather_output</span><span class="o">=</span><span class="kc">False</span><span class="p">)</span>
|
|
<span class="k">else</span><span class="p">:</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">fc</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">hidden_size</span><span class="p">,</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">ffn_hidden_size</span> <span class="o">*</span> <span class="mi">2</span><span class="p">,</span>
|
|
<span class="n">bias</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">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">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">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">tp_size</span><span class="p">,</span>
|
|
<span class="n">gather_output</span><span class="o">=</span><span class="kc">False</span><span class="p">)</span>
|
|
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">fc</span><span class="o">.</span><span class="n">weight</span><span class="o">.</span><span class="n">value</span> <span class="o">=</span> <span class="n">concat_weight</span>
|
|
|
|
<span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">use_fp8_qdq</span><span class="p">:</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">fc</span><span class="o">.</span><span class="n">activation_scaling_factor</span><span class="o">.</span><span class="n">value</span> <span class="o">=</span> <span class="n">fc_activation_scaling_factor</span>
|
|
<span class="c1"># TODO: need to align with quantization toolkit; preferably put a constraint to equalize</span>
|
|
<span class="c1"># fc/gate weight scaling factor to allow horizontal fusion without accuracy loss</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">fc</span><span class="o">.</span><span class="n">weights_scaling_factor</span><span class="o">.</span><span class="n">value</span> <span class="o">=</span> <span class="nb">max</span><span class="p">(</span>
|
|
<span class="n">gate_weights_scaling_factor</span><span class="p">,</span> <span class="n">fc_weights_scaling_factor</span><span class="p">)</span>
|
|
|
|
<span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">bias</span><span class="p">:</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">fc</span><span class="o">.</span><span class="n">bias</span><span class="o">.</span><span class="n">value</span> <span class="o">=</span> <span class="n">concat_bias</span>
|
|
<span class="n">inter</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">fc</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">hidden_act</span> <span class="o">==</span> <span class="s1">'silu'</span><span class="p">:</span>
|
|
<span class="n">inter</span> <span class="o">=</span> <span class="n">ACT2FN</span><span class="p">[</span><span class="s1">'swiglu'</span><span class="p">](</span><span class="n">inter</span><span class="p">)</span>
|
|
<span class="k">else</span><span class="p">:</span>
|
|
<span class="k">raise</span> <span class="ne">NotImplementedError</span><span class="p">(</span>
|
|
<span class="sa">f</span><span class="s2">"Activation </span><span class="si">{</span><span class="bp">self</span><span class="o">.</span><span class="n">hidden_act</span><span class="si">}</span><span class="s2"> not yet implemented for FusedGatedMLP"</span>
|
|
<span class="p">)</span>
|
|
<span class="n">output</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">proj</span><span class="p">(</span><span class="n">inter</span><span class="p">,</span> <span class="n">workspace</span><span class="p">)</span>
|
|
<span class="k">return</span> <span class="n">output</span></div>
|
|
</div>
|
|
|
|
</pre></div>
|
|
|
|
</div>
|
|
</div>
|
|
<footer>
|
|
|
|
<hr/>
|
|
|
|
<div role="contentinfo">
|
|
<p>© Copyright 2023, NVidia.</p>
|
|
</div>
|
|
|
|
Built with <a href="https://www.sphinx-doc.org/">Sphinx</a> using a
|
|
<a href="https://github.com/readthedocs/sphinx_rtd_theme">theme</a>
|
|
provided by <a href="https://readthedocs.org">Read the Docs</a>.
|
|
|
|
|
|
</footer>
|
|
</div>
|
|
</div>
|
|
</section>
|
|
</div>
|
|
<script>
|
|
jQuery(function () {
|
|
SphinxRtdTheme.Navigation.enable(true);
|
|
});
|
|
</script>
|
|
|
|
</body>
|
|
</html> |