TensorRT-LLMs/docs/source/architecture/add-model.md
Kaiyu Xie 66ef1df492
Update TensorRT-LLM (#1492)
* Update TensorRT-LLM

---------

Co-authored-by: Loki <lokravi@amazon.com>
2024-04-24 14:44:22 +08:00

17 lines
503 B
Markdown

(add-model)=
# Adding a Model
This document describes how to add a model in TensorRT-LLM.
TensorRT-LLM provides:
- Low-level functions, for example, `concat`, `add`, and `sum`.
- Basic layers, such as, `Linear` and `LayerNorm`.
- High-level layers, such as, `MLP` and `Attention`.
**Steps**
1. Create a model directory in `tensorrt_llm/tensorrt_llm/models`, for example `bloom`.
2. Write a `model.py` with TensorRT-LLM low level functions and basic layers. It's optional to use high level layers.