From b2b9c427246d895a943fa6a5f8e9b702eff03559 Mon Sep 17 00:00:00 2001 From: Vaibhav Srivastav Date: Fri, 9 Aug 2024 15:01:34 +0200 Subject: [PATCH] Update documentation for Supported models (#2386) * Minor doc fixes * up. * Other minor updates. --- README.md | 34 ++++++++++++++++--- docs/source/conceptual/quantization.md | 4 +-- docs/source/quicktour.md | 2 +- docs/source/supported_models.md | 8 ++--- .../text_generation_server/models/__init__.py | 6 ++-- update_doc.py | 2 +- 6 files changed, 41 insertions(+), 15 deletions(-) diff --git a/README.md b/README.md index cf7f1d22..803e9172 100644 --- a/README.md +++ b/README.md @@ -13,7 +13,7 @@ Swagger API documentation -A Rust, Python and gRPC server for text generation inference. Used in production at [HuggingFace](https://huggingface.co) +A Rust, Python and gRPC server for text generation inference. Used in production at [Hugging Face](https://huggingface.co) to power Hugging Chat, the Inference API and Inference Endpoint. @@ -42,6 +42,7 @@ Text Generation Inference (TGI) is a toolkit for deploying and serving Large Lan - Tensor Parallelism for faster inference on multiple GPUs - Token streaming using Server-Sent Events (SSE) - Continuous batching of incoming requests for increased total throughput +- [Messages API](https://huggingface.co/docs/text-generation-inference/en/messages_api) compatible with Open AI Chat Completion API - Optimized transformers code for inference using [Flash Attention](https://github.com/HazyResearch/flash-attention) and [Paged Attention](https://github.com/vllm-project/vllm) on the most popular architectures - Quantization with : - [bitsandbytes](https://github.com/TimDettmers/bitsandbytes) @@ -49,7 +50,7 @@ Text Generation Inference (TGI) is a toolkit for deploying and serving Large Lan - [EETQ](https://github.com/NetEase-FuXi/EETQ) - [AWQ](https://github.com/casper-hansen/AutoAWQ) - [Marlin](https://github.com/IST-DASLab/marlin) - - [fp8]() + - [fp8](https://developer.nvidia.com/blog/nvidia-arm-and-intel-publish-fp8-specification-for-standardization-as-an-interchange-format-for-ai/) - [Safetensors](https://github.com/huggingface/safetensors) weight loading - Watermarking with [A Watermark for Large Language Models](https://arxiv.org/abs/2301.10226) - Logits warper (temperature scaling, top-p, top-k, repetition penalty, more details see [transformers.LogitsProcessor](https://huggingface.co/docs/transformers/internal/generation_utils#transformers.LogitsProcessor)) @@ -94,6 +95,29 @@ curl 127.0.0.1:8080/generate_stream \ -H 'Content-Type: application/json' ``` +You can also use [TGI's Messages API](https://huggingface.co/docs/text-generation-inference/en/messages_api) to obtain Open AI Chat Completion API compatible responses. + +```bash +curl localhost:3000/v1/chat/completions \ + -X POST \ + -d '{ + "model": "tgi", + "messages": [ + { + "role": "system", + "content": "You are a helpful assistant." + }, + { + "role": "user", + "content": "What is deep learning?" + } + ], + "stream": true, + "max_tokens": 20 +}' \ + -H 'Content-Type: application/json' +``` + **Note:** To use NVIDIA GPUs, you need to install the [NVIDIA Container Toolkit](https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/install-guide.html). We also recommend using NVIDIA drivers with CUDA version 12.2 or higher. For running the Docker container on a machine with no GPUs or CUDA support, it is enough to remove the `--gpus all` flag and add `--disable-custom-kernels`, please note CPU is not the intended platform for this project, so performance might be subpar. **Note:** TGI supports AMD Instinct MI210 and MI250 GPUs. Details can be found in the [Supported Hardware documentation](https://huggingface.co/docs/text-generation-inference/supported_models#supported-hardware). To use AMD GPUs, please use `docker run --device /dev/kfd --device /dev/dri --shm-size 1g -p 8080:80 -v $volume:/data ghcr.io/huggingface/text-generation-inference:2.2.0-rocm --model-id $model` instead of the command above. @@ -122,7 +146,7 @@ For example, if you want to serve the gated Llama V2 model variants: or with Docker: ```shell -model=meta-llama/Llama-2-7b-chat-hf +model=meta-llama/Meta-Llama-3.1-8B-Instruct volume=$PWD/data # share a volume with the Docker container to avoid downloading weights every run token= @@ -234,7 +258,7 @@ text-generation-launcher --model-id mistralai/Mistral-7B-Instruct-v0.2 ### Quantization -You can also quantize the weights with bitsandbytes to reduce the VRAM requirement: +You can also run pre-quantized weights (AWQ, GPTQ, Marlin) or on-the-fly quantize weights with bitsandbytes, EETQ, fp8, to reduce the VRAM requirement: ```shell text-generation-launcher --model-id mistralai/Mistral-7B-Instruct-v0.2 --quantize @@ -242,6 +266,8 @@ text-generation-launcher --model-id mistralai/Mistral-7B-Instruct-v0.2 --quantiz 4bit quantization is available using the [NF4 and FP4 data types from bitsandbytes](https://arxiv.org/pdf/2305.14314.pdf). It can be enabled by providing `--quantize bitsandbytes-nf4` or `--quantize bitsandbytes-fp4` as a command line argument to `text-generation-launcher`. +Read more about quantization in the [Quantization documentation](https://huggingface.co/docs/text-generation-inference/en/conceptual/quantization). + ## Develop ```shell diff --git a/docs/source/conceptual/quantization.md b/docs/source/conceptual/quantization.md index 7507687f..a1ebe7e7 100644 --- a/docs/source/conceptual/quantization.md +++ b/docs/source/conceptual/quantization.md @@ -11,7 +11,7 @@ We recommend using the official quantization scripts for creating your quants: For on-the-fly quantization you simply need to pass one of the supported quantization types and TGI takes care of the rest. -## Quantization with bitsandbytes +## Quantization with bitsandbytes, EETQ & fp8 bitsandbytes is a library used to apply 8-bit and 4-bit quantization to models. Unlike GPTQ quantization, bitsandbytes doesn't require a calibration dataset or any post-processing – weights are automatically quantized on load. However, inference with bitsandbytes is slower than GPTQ or FP16 precision. @@ -32,7 +32,7 @@ docker run --gpus all --shm-size 1g -p 8080:80 -v $volume:/data ghcr.io/huggingf You can get more information about 8-bit quantization by reading this [blog post](https://huggingface.co/blog/hf-bitsandbytes-integration), and 4-bit quantization by reading [this blog post](https://huggingface.co/blog/4bit-transformers-bitsandbytes). -Use `eetq` or `fp8` for other quantization schemes. +Similarly you can use pass you can pass `--quantize eetq` or `--quantize fp8` for respective quantization schemes. In addition to this, TGI allows creating GPTQ quants directly by passing the model weights and a calibration dataset. diff --git a/docs/source/quicktour.md b/docs/source/quicktour.md index 2313c69b..18e1a107 100644 --- a/docs/source/quicktour.md +++ b/docs/source/quicktour.md @@ -21,7 +21,7 @@ TGI supports various hardware. Make sure to check the [Using TGI with Nvidia GPU ## Consuming TGI -Once TGI is running, you can use the `generate` endpoint by doing requests. To learn more about how to query the endpoints, check the [Consuming TGI](./basic_tutorials/consuming_tgi) section, where we show examples with utility libraries and UIs. Below you can see a simple snippet to query the endpoint. +Once TGI is running, you can use the `generate` endpoint or the Open AI Chat Completion API compatible [Messages API](https://huggingface.co/docs/text-generation-inference/en/messages_api) by doing requests. To learn more about how to query the endpoints, check the [Consuming TGI](./basic_tutorials/consuming_tgi) section, where we show examples with utility libraries and UIs. Below you can see a simple snippet to query the endpoint. diff --git a/docs/source/supported_models.md b/docs/source/supported_models.md index b78104df..832f88ef 100644 --- a/docs/source/supported_models.md +++ b/docs/source/supported_models.md @@ -1,22 +1,22 @@ # Supported Models and Hardware -Text Generation Inference enables serving optimized models on specific hardware for the highest performance. The following sections list which models are hardware are supported. +Text Generation Inference enables serving optimized models on specific hardware for the highest performance. The following sections list which models (VLMs & LLMs) are supported. ## Supported Models - [Deepseek V2](https://huggingface.co/deepseek-ai/DeepSeek-V2) - [Idefics 2](https://huggingface.co/HuggingFaceM4/idefics2-8b) (Multimodal) - [Llava Next (1.6)](https://huggingface.co/llava-hf/llava-v1.6-vicuna-13b-hf) (Multimodal) -- [Llama](https://huggingface.co/meta-llama/Meta-Llama-3-8B-Instruct) +- [Llama](https://huggingface.co/collections/meta-llama/llama-31-669fc079a0c406a149a5738f) - [Phi 3](https://huggingface.co/microsoft/Phi-3-mini-4k-instruct) - [Gemma](https://huggingface.co/google/gemma-7b) - [PaliGemma](https://huggingface.co/google/paligemma-3b-pt-224) -- [Gemma2](https://huggingface.co/google/gemma2-9b) +- [Gemma2](https://huggingface.co/collections/google/gemma-2-release-667d6600fd5220e7b967f315) - [Cohere](https://huggingface.co/CohereForAI/c4ai-command-r-plus) - [Dbrx](https://huggingface.co/databricks/dbrx-instruct) - [Mamba](https://huggingface.co/state-spaces/mamba-2.8b-slimpj) -- [Mistral](https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.2) +- [Mistral](https://huggingface.co/mistralai/Mistral-Nemo-Instruct-2407) - [Mixtral](https://huggingface.co/mistralai/Mixtral-8x22B-Instruct-v0.1) - [Gpt Bigcode](https://huggingface.co/bigcode/gpt_bigcode-santacoder) - [Phi](https://huggingface.co/microsoft/phi-1_5) diff --git a/server/text_generation_server/models/__init__.py b/server/text_generation_server/models/__init__.py index da14d083..960b426b 100644 --- a/server/text_generation_server/models/__init__.py +++ b/server/text_generation_server/models/__init__.py @@ -180,7 +180,7 @@ class ModelType(enum.Enum): LLAMA = { "type": "llama", "name": "Llama", - "url": "https://huggingface.co/meta-llama/Meta-Llama-3-8B-Instruct", + "url": "https://huggingface.co/collections/meta-llama/llama-31-669fc079a0c406a149a5738f", } PHI3 = { "type": "phi3", @@ -200,7 +200,7 @@ class ModelType(enum.Enum): GEMMA2 = { "type": "gemma2", "name": "Gemma2", - "url": "https://huggingface.co/google/gemma2-9b", + "url": "https://huggingface.co/collections/google/gemma-2-release-667d6600fd5220e7b967f315", } COHERE = { "type": "cohere", @@ -220,7 +220,7 @@ class ModelType(enum.Enum): MISTRAL = { "type": "mistral", "name": "Mistral", - "url": "https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.2", + "url": "https://huggingface.co/mistralai/Mistral-Nemo-Instruct-2407", } MIXTRAL = { "type": "mixtral", diff --git a/update_doc.py b/update_doc.py index 428d4452..e887e1c6 100644 --- a/update_doc.py +++ b/update_doc.py @@ -7,7 +7,7 @@ import os TEMPLATE = """ # Supported Models and Hardware -Text Generation Inference enables serving optimized models on specific hardware for the highest performance. The following sections list which models are hardware are supported. +Text Generation Inference enables serving optimized models on specific hardware for the highest performance. The following sections list which models (VLMs & LLMs) are supported. ## Supported Models