Merge branch 'huggingface:main' into abhinavkulkarni/add-awq-support

docs/source/_toctree.yml

@@ -17,10 +17,16 @@
     title: Serving Private & Gated Models
   - local: basic_tutorials/using_cli
     title: Using TGI CLI
+  - local: basic_tutorials/non_core_models
+    title: Non-core Model Serving
   title: Tutorials
 - sections:
   - local: conceptual/streaming
     title: Streaming
+  - local: conceptual/quantization
+    title: Quantization
+  - local: conceptual/tensor_parallelism
+    title: Tensor Parallelism
   - local: conceptual/paged_attention
     title: PagedAttention
   - local: conceptual/safetensors

docs/source/basic_tutorials/non_core_models.md

@@ -0,0 +1,24 @@
+# Non-core Model Serving
+
+TGI supports various LLM architectures (see full list [here](../supported_models)). If you wish to serve a model that is not one of the supported models, TGI will fallback to the `transformers` implementation of that model. This means you will be unable to use some of the features introduced by TGI, such as tensor-parallel sharding or flash attention. However, you can still get many benefits of TGI, such as continuous batching or streaming outputs.
+
+You can serve these models using the same Docker command-line invocation as with fully supported models 👇 
+
+```bash
+docker run --gpus all --shm-size 1g -p 8080:80 -v $volume:/data ghcr.io/huggingface/text-generation-inference:latest --model-id gpt2
+```
+
+If the model you wish to serve is a custom transformers model, and its weights and implementation are available in the Hub, you can still serve the model by passing the `--trust-remote-code` flag to the `docker run` command like below 👇 
+
+```bash
+docker run --gpus all --shm-size 1g -p 8080:80 -v $volume:/data ghcr.io/huggingface/text-generation-inference:latest --model-id <CUSTOM_MODEL_ID> --trust-remote-code
+```
+
+Finally, if the model is not on Hugging Face Hub but on your local, you can pass the path to the folder that contains your model like below 👇 
+
+```bash
+# Make sure your model is in the $volume directory
+docker run --shm-size 1g -p 8080:80 -v $volume:/data  ghcr.io/huggingface/text-generation-inference:latest --model-id /data/<PATH-TO-FOLDER>
+```
+
+You can refer to [transformers docs on custom models](https://huggingface.co/docs/transformers/main/en/custom_models) for more information.

docs/source/basic_tutorials/preparing_model.md

@@ -4,7 +4,7 @@ Text Generation Inference improves the model in several aspects.
 
 ## Quantization
 
-TGI supports [bits-and-bytes](https://github.com/TimDettmers/bitsandbytes#bitsandbytes) and [GPT-Q](https://arxiv.org/abs/2210.17323) quantization. To speed up inference with quantization, simply set `quantize` flag to `bitsandbytes` or `gptq` depending on the quantization technique you wish to use. When using GPT-Q quantization, you need to point to one of the models [here](https://huggingface.co/models?search=gptq).
+TGI supports [bits-and-bytes](https://github.com/TimDettmers/bitsandbytes#bitsandbytes) and [GPT-Q](https://arxiv.org/abs/2210.17323) quantization. To speed up inference with quantization, simply set `quantize` flag to `bitsandbytes` or `gptq` depending on the quantization technique you wish to use. When using GPT-Q quantization, you need to point to one of the models [here](https://huggingface.co/models?search=gptq). To get more information about quantization, please refer to (./conceptual/quantization.md)
 
 
 ## RoPE Scaling

docs/source/conceptual/quantization.md

@@ -0,0 +1,59 @@
+# Quantization
+
+TGI offers GPTQ and bits-and-bytes quantization to quantize large language models.
+
+## Quantization with GPTQ
+
+GPTQ is a post-training quantization method to make the model smaller. It quantizes the layers by finding a compressed version of that weight, that will yield a minimum mean squared error like below 👇 
+
+Given a layer \\(l\\) with weight matrix \\(W_{l}\\) and layer input \\(X_{l}\\), find quantized weight \\(\\hat{W}_{l}\\):
+
+$$({\hat{W}_{l}}^{*} = argmin_{\hat{W_{l}}} ||W_{l}X-\hat{W}_{l}X||^{2}_{2})$$
+
+
+TGI allows you to both run an already GPTQ quantized model (see available models [here](https://huggingface.co/models?search=gptq)) or quantize a model of your choice using quantization script. You can run a quantized model by simply passing --quantize like below 👇 
+
+```bash
+docker run --gpus all --shm-size 1g -p 8080:80 -v $volume:/data ghcr.io/huggingface/text-generation-inference:latest --model-id $model --quantize gptq
+```
+
+Note that TGI's GPTQ implementation doesn't use [AutoGPTQ](https://github.com/PanQiWei/AutoGPTQ) under the hood. However, models quantized using AutoGPTQ or Optimum can still be served by TGI. 
+
+To quantize a given model using GPTQ with a calibration dataset, simply run
+
+```bash
+text-generation-server quantize tiiuae/falcon-40b /data/falcon-40b-gptq
+# Add --upload-to-model-id MYUSERNAME/falcon-40b to push the created model to the hub directly
+```
+
+This will create a new directory with the quantized files which you can use with,
+
+```bash
+text-generation-launcher --model-id /data/falcon-40b-gptq/ --sharded true --num-shard 2 --quantize gptq
+```
+
+You can learn more about the quantization options by running `text-generation-server quantize --help`.
+
+If you wish to do more with GPTQ models (e.g. train an adapter on top), you can read about transformers GPTQ integration [here](https://huggingface.co/blog/gptq-integration).
+You can learn more about GPTQ from the [paper](https://arxiv.org/pdf/2210.17323.pdf).
+
+## Quantization with bitsandbytes
+
+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.
+
+8-bit quantization enables multi-billion parameter scale models to fit in smaller hardware without degrading performance too much. 
+In TGI, you can use 8-bit quantization by adding `--quantize bitsandbytes` like below 👇
+
+```bash
+docker run --gpus all --shm-size 1g -p 8080:80 -v $volume:/data ghcr.io/huggingface/text-generation-inference:latest --model-id $model --quantize --bitsandbytes
+```
+
+4-bit quantization is also possible with bitsandbytes. You can choose one of the following 4-bit data types: 4-bit float (`fp4`), or 4-bit `NormalFloat` (`nf4`). These data types were introduced in the context of parameter-efficient fine-tuning, but you can apply them for inference by automatically converting the model weights on load.
+
+In TGI, you can use 4-bit quantization by adding `--quantize bitsandbytes-nf4` or `--quantize bitsandbytes-fp4` like below 👇 
+
+```bash
+docker run --gpus all --shm-size 1g -p 8080:80 -v $volume:/data ghcr.io/huggingface/text-generation-inference:latest --model-id $model --quantize --bitsandbytes-nf4
+```
+
+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).

docs/source/conceptual/tensor_parallelism.md

@@ -0,0 +1,14 @@
+# Tensor Parallelism
+
+Tensor parallelism is a technique used to fit a large model in multiple GPUs. For example, when multiplying the input tensors with the first weight tensor, the matrix multiplication is equivalent to splitting the weight tensor column-wise, multiplying each column with the input separately, and then concatenating the separate outputs. These outputs are then transferred from the GPUs and concatenated together to get the final result, like below 👇 
+
+![Image courtesy of Anton Lozkhov](https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/tgi/TP.png)
+
+
+<Tip warning={true}>
+
+Tensor Parallelism only works for [models officially supported](../supported_models), it will not work when falling back to `transformers`. You can get more information about unsupported models [here](../basic_tutorials/non_core_models).
+
+</Tip>
+
+You can learn a lot more details about tensor-parallelism from [the `transformers` docs](https://huggingface.co/docs/transformers/main/en/perf_train_gpu_many#tensor-parallelism).