Enable gptqmodel (huggingface#35012)

* gptqmodel Signed-off-by: jiqing-feng <[email protected]> * fix format Signed-off-by: jiqing-feng <[email protected]> * update readme Signed-off-by: jiqing-feng <[email protected]> * gptqmodel need use checkpoint_format (#1) * gptqmodel need use checkpoint_format * fix quantize * Update quantization_config.py * Update quantization_config.py * Update quantization_config.py --------- Co-authored-by: ZX-ModelCloud <[email protected]> Co-authored-by: Qubitium-ModelCloud <[email protected]> * Revert quantizer_gptq.py (#2) * revert quantizer_gptq.py change * pass **kwargs * limit gptqmodel and optimum version Signed-off-by: jiqing-feng <[email protected]> * fix format Signed-off-by: jiqing-feng <[email protected]> * fix warning Signed-off-by: jiqing-feng <[email protected]> * fix version check Signed-off-by: jiqing-feng <[email protected]> * revert unrelated changes Signed-off-by: jiqing-feng <[email protected]> * enable gptqmodel tests Signed-off-by: jiqing-feng <[email protected]> * fix requires gptq Signed-off-by: jiqing-feng <[email protected]> * Fix Transformer compat (#3) * revert quantizer_gptq.py change * pass **kwargs * add meta info * cleanup * cleanup * Update quantization_config.py * hf_select_quant_linear pass checkpoint_format and meta * fix GPTQTestCUDA * Update test_gptq.py * gptqmodel.hf_select_quant_linear() now does not select ExllamaV2 * cleanup * add backend * cleanup * cleanup * no need check exllama version * Update quantization_config.py * lower checkpoint_format and backend * check none * cleanup * Update quantization_config.py * fix self.use_exllama == False * spell * fix unittest * fix unittest --------- Co-authored-by: LRL <[email protected]> Co-authored-by: Qubitium-ModelCloud <[email protected]> * fix format Signed-off-by: jiqing-feng <[email protected]> * fix format again Signed-off-by: jiqing-feng <[email protected]> * update gptqmodel version (huggingface#6) * update gptqmodel version * update gptqmodel version * fix unit test (huggingface#5) * update gptqmodel version * update gptqmodel version * "not self.use_exllama" is not equivalent to "self.use_exllama==False" * fix unittest * update gptqmodel version * backend is loading_attibutes (huggingface#7) * fix format and tests Signed-off-by: jiqing-feng <[email protected]> * fix memory check Signed-off-by: jiqing-feng <[email protected]> * fix device mismatch Signed-off-by: jiqing-feng <[email protected]> * fix result check Signed-off-by: jiqing-feng <[email protected]> * Update src/transformers/quantizers/quantizer_gptq.py Co-authored-by: Marc Sun <[email protected]> * Update src/transformers/quantizers/quantizer_gptq.py Co-authored-by: Marc Sun <[email protected]> * Update src/transformers/quantizers/quantizer_gptq.py Co-authored-by: Marc Sun <[email protected]> * update tests Signed-off-by: jiqing-feng <[email protected]> * review: update docs (huggingface#10) * review: update docs (huggingface#12) * review: update docs * fix typo * update tests for gptqmodel Signed-off-by: jiqing-feng <[email protected]> * update document (huggingface#9) * update overview.md * cleanup * Update overview.md * Update overview.md * Update overview.md * update gptq.md * Update gptq.md * Update gptq.md * Update gptq.md * Update gptq.md * Update gptq.md * Update gptq.md --------- Co-authored-by: Qubitium-ModelCloud <[email protected]> * typo * doc note for asymmetric quant * typo with apple silicon(e) * typo for marlin * column name revert: review * doc rocm support * Update docs/source/en/quantization/gptq.md Co-authored-by: Steven Liu <[email protected]> * Update docs/source/en/quantization/gptq.md Co-authored-by: Steven Liu <[email protected]> * Update docs/source/en/quantization/gptq.md Co-authored-by: Steven Liu <[email protected]> * Update docs/source/en/quantization/gptq.md Co-authored-by: Steven Liu <[email protected]> * Update docs/source/en/quantization/overview.md Co-authored-by: Steven Liu <[email protected]> * Update docs/source/en/quantization/overview.md Co-authored-by: Steven Liu <[email protected]> --------- Signed-off-by: jiqing-feng <[email protected]> Co-authored-by: LRL-ModelCloud <[email protected]> Co-authored-by: ZX-ModelCloud <[email protected]> Co-authored-by: Qubitium-ModelCloud <[email protected]> Co-authored-by: ZX-ModelCloud <[email protected]> Co-authored-by: LRL <[email protected]> Co-authored-by: Marc Sun <[email protected]> Co-authored-by: Mohamed Mekkouri <[email protected]> Co-authored-by: Steven Liu <[email protected]>
ariG23498 · Jan 15, 2025 · 387663e · 387663e
1 parent 615bf9c
commit 387663e
Show file tree

Hide file tree

Showing 9 changed files with 266 additions and 86 deletions.
diff --git a/docs/source/en/quantization/gptq.md b/docs/source/en/quantization/gptq.md
@@ -22,15 +22,42 @@ Try GPTQ quantization with PEFT in this [notebook](https://colab.research.google
 
 </Tip>
 
-The [AutoGPTQ](https://github.com/PanQiWei/AutoGPTQ) library implements the GPTQ algorithm, a post-training quantization technique where each row of the weight matrix is quantized independently to find a version of the weights that minimizes the error. These weights are quantized to int4, but they're restored to fp16 on the fly during inference. This can save your memory-usage by 4x because the int4 weights are dequantized in a fused kernel rather than a GPU's global memory, and you can also expect a speedup in inference because using a lower bitwidth takes less time to communicate.
+Both [GPTQModel](https://github.com/ModelCloud/GPTQModel) and [AutoGPTQ](https://github.com/PanQiWei/AutoGPTQ) libraries implement the GPTQ algorithm, a post-training quantization technique where each row of the weight matrix is quantized independently to find a version of the weights that minimizes error. These weights are quantized to int4, stored as int32 (int4 x 8) and dequantized (restored) to fp16 on the fly during inference. This can save memory by almost 4x because the int4 weights are often dequantized in a fused kernel. You can also expect a substantial speedup in inference due to lower bandwidth requirements for lower bitwidth.
 
-Before you begin, make sure the following libraries are installed:
+[GPTQModel](https://github.com/ModelCloud/GPTQModel) started as a maintained fork of AutoGPTQ but has since differentiated itself with the following major differences.
+
+* Model support: GPTQModel continues to support all of the latest LLM models.
+* Multimodal support: GPTQModel supports accurate quantization of Qwen 2-VL and Ovis 1.6-VL image-to-text models. 
+* Platform support: Linux, macOS (Apple Silicon), and Windows 11.
+* Hardware support: NVIDIA CUDA, AMD ROCm, Apple Silicon M1/MPS /CPU, Intel/AMD CPU, and Intel Datacenter Max/Arc GPUs.
+* Asymmetric support: Asymmetric quantization can potentially introduce lower quantization errors compared to symmetric quantization. However, it is not backward compatible with AutoGPTQ, and not all kernels, such as Marlin, support asymmetric quantization.
+* IPEX kernel for Intel/AMD accelerated CPU and Intel GPU (Datacenter Max/Arc GPUs) support.
+* Updated Marlin kernel from Neural Magic optimized for A100 (Ampere).
+* Updated kernels with auto-padding for legacy model support and models with non-uniform in/out-features. 
+* Faster quantization, lower memory usage, and more accurate default quantization via GPTQModel quantization APIs.
+* User and developer friendly APIs. 
+
+
+[AutoGPTQ](https://github.com/PanQiWei/AutoGPTQ) will likely be deprecated in the future due the lack of continued support for new models and features. 
+
+Before you begin, make sure the following libraries are installed and updated to the latest release:
 
 ```bash
-pip install auto-gptq
 pip install --upgrade accelerate optimum transformers
 ```
 
+Then install either GPTQModel or AutoGPTQ.
+
+```bash
+pip install gptqmodel --no-build-isolation
+```
+
+or
+
+```bash
+pip install auto-gptq --no-build-isolation
+```
+
 To quantize a model (currently only supported for text models), you need to create a [`GPTQConfig`] class and set the number of bits to quantize to, a dataset to calibrate the weights for quantization, and a tokenizer to prepare the dataset.
 
 ```py
@@ -92,9 +119,22 @@ from transformers import AutoModelForCausalLM
 model = AutoModelForCausalLM.from_pretrained("{your_username}/opt-125m-gptq", device_map="auto")
 ```
 
+## Marlin
+
+[Marlin](https://github.com/IST-DASLab/marlin) is a 4-bit only CUDA GPTQ kernel, highly optimized for the NVIDIA A100 GPU (Ampere) architecture. Loading, dequantization, and execution of post-dequantized weights are highly parallelized, offering a substantial inference improvement versus the original CUDA GPTQ kernel. Marlin is only available for quantized inference and does not support model quantization.
+
+Marlin inference can be activated with the `backend` parameter in [`GPTQConfig`].
+
+```py
+
+from transformers import AutoModelForCausalLM, GPTQConfig
+
+model = AutoModelForCausalLM.from_pretrained("{your_username}/opt-125m-gptq", device_map="auto", quantization_config=GPTQConfig(bits=4, backend="marlin"))
+```
+
 ## ExLlama
 
-[ExLlama](https://github.com/turboderp/exllama) is a Python/C++/CUDA implementation of the [Llama](model_doc/llama) model that is designed for faster inference with 4-bit GPTQ weights (check out these [benchmarks](https://github.com/huggingface/optimum/tree/main/tests/benchmark#gptq-benchmark)). The ExLlama kernel is activated by default when you create a [`GPTQConfig`] object. To boost inference speed even further, use the [ExLlamaV2](https://github.com/turboderp/exllamav2) kernels by configuring the `exllama_config` parameter:
+[ExLlama](https://github.com/turboderp/exllama) is a CUDA implementation of the [Llama](model_doc/llama) model that is designed for faster inference with 4-bit GPTQ weights (check out these [benchmarks](https://github.com/huggingface/optimum/tree/main/tests/benchmark#gptq-benchmark)). The ExLlama kernel is activated by default when you create a [`GPTQConfig`] object. To boost inference speed even further, use the [ExLlamaV2](https://github.com/turboderp/exllamav2) kernels by configuring the `exllama_config` parameter:
 
 ```py
 import torch
@@ -110,11 +150,11 @@ Only 4-bit models are supported, and we recommend deactivating the ExLlama kerne
 
 </Tip>
 
-The ExLlama kernels are only supported when the entire model is on the GPU. If you're doing inference on a CPU with AutoGPTQ (version > 0.4.2), then you'll need to disable the ExLlama kernel. This overwrites the attributes related to the ExLlama kernels in the quantization config of the config.json file.
+The ExLlama kernels are only supported when the entire model is on the GPU. If you're doing inference on a CPU with AutoGPTQ or GPTQModel, then you'll need to disable the ExLlama kernel. This overwrites the attributes related to the ExLlama kernels in the quantization config of the config.json file.
 
 ```py
 import torch
 from transformers import AutoModelForCausalLM, GPTQConfig
 gptq_config = GPTQConfig(bits=4, use_exllama=False)
 model = AutoModelForCausalLM.from_pretrained("{your_username}/opt-125m-gptq", device_map="cpu", quantization_config=gptq_config)
-```
+```
diff --git a/docs/source/en/quantization/overview.md b/docs/source/en/quantization/overview.md
@@ -45,32 +45,50 @@ In short, supporting a wide range of quantization methods allows you to pick the
 
 Use the table below to help you decide which quantization method to use.
 
-| Quantization method                 | On the fly quantization | CPU | CUDA GPU | RoCm GPU (AMD) | Metal (Apple Silicon) | Intel GPU | torch.compile() support | Number of bits | Supports fine-tuning (through PEFT) | Serializable with 🤗 transformers | 🤗 transformers support | Link to library                             |
-|-------------------------------------|-------------------------|-----|----------|----------------|-----------------------|-----------|-------------------------|----------------|-------------------------------------|--------------|------------------------|---------------------------------------------|
-| [AQLM](./aqlm)                                | 🔴                       |  🟢   |     🟢     | 🔴              | 🔴                     | 🔴         | 🟢                      | 1 / 2          | 🟢                                   | 🟢            | 🟢                      | https://github.com/Vahe1994/AQLM            |
-| [AWQ](./awq) | 🔴                       | 🟢   | 🟢        | 🟢              | 🔴                     | 🟢         | ?                       | 4              | 🟢                                   | 🟢            | 🟢                      | https://github.com/casper-hansen/AutoAWQ    |
-| [bitsandbytes](./bitsandbytes)     | 🟢            | 🟡 *   |     🟢     | 🟡 *            | 🔴 **    | 🟡 *       | 🔴    (soon!)          | 4 / 8          | 🟢                                   | 🟢            | 🟢                      | https://github.com/bitsandbytes-foundation/bitsandbytes |
-| [compressed-tensors](./compressed_tensors)                        | 🔴                       | 🟢   |     🟢     | 🟢              | 🔴                     | 🔴         | 🔴                       | 1 - 8          | 🟢                                   | 🟢            | 🟢                      | https://github.com/neuralmagic/compressed-tensors |
-| [EETQ](./eetq)                                | 🟢                       | 🔴   | 🟢        | 🔴              | 🔴         | 🔴                     | ?                       | 8              | 🟢                                   | 🟢            | 🟢                      | https://github.com/NetEase-FuXi/EETQ        |
-| GGUF / GGML (llama.cpp)             | 🟢                       | 🟢   | 🟢        | 🔴              | 🟢                     | 🔴         | 🔴                       | 1 - 8          | 🔴                                   | [See GGUF section](../gguf)                | [See GGUF section](../gguf)                      | https://github.com/ggerganov/llama.cpp      |
-| [GPTQ](./gptq)                                | 🔴                       | 🔴   | 🟢        | 🟢              | 🔴                     | 🔴         | 🔴                       | 2 - 3 - 4 - 8          | 🟢                                   | 🟢            | 🟢                      | https://github.com/AutoGPTQ/AutoGPTQ        |
-| [HIGGS](./higgs)                             | 🟢                       | 🔴    | 🟢        | 🔴              | 🔴                     | 🔴         | 🟢                       | 2 - 4          | 🔴                                   | 🟢            | 🟢                      | https://github.com/HanGuo97/flute           |       
-| [HQQ](./hqq)                                 | 🟢                       | 🟢    | 🟢        | 🔴              | 🔴                     | 🔴         | 🟢                       | 1 - 8          | 🟢                                   | 🔴            | 🟢                      | https://github.com/mobiusml/hqq/            |
-| [optimum-quanto](./quanto)                              | 🟢                       | 🟢   | 🟢        | 🔴              | 🟢                     | 🔴         | 🟢                       | 2 / 4 / 8      | 🔴                                   | 🔴            | 🟢                      | https://github.com/huggingface/optimum-quanto       |
-| [FBGEMM_FP8](./fbgemm_fp8.md)                              | 🟢                       | 🔴    | 🟢        | 🔴              | 🔴                      | 🔴         | 🔴                        | 8      | 🔴                                   | 🟢            | 🟢                      | https://github.com/pytorch/FBGEMM       |
-| [torchao](./torchao.md)                              | 🟢                       |     | 🟢        | 🔴              | partial support (int4 weight only)       | 🔴         |                       | 4 / 8      |                                   | 🟢🔴           | 🟢                      | https://github.com/pytorch/ao       |
-| [VPTQ](./vptq)                      | 🔴                       |  🔴   |     🟢     | 🟡              | 🔴      | 🔴                | 🟢                      | 1 - 8          | 🔴                                   | 🟢            | 🟢                      | https://github.com/microsoft/VPTQ            |
+| Quantization Method                           | On the fly quantization | CPU             | CUDA GPU | ROCm GPU  | Metal (Apple Silicon)              | Intel GPU       | Torch compile() | Bits          | PEFT Fine Tuning | Serializable with 🤗Transformers | 🤗Transformers Support  | Link to library                             |
+|-----------------------------------------------|----------------------|-----------------|----------|-----------|------------------------------------|-----------------|-----------------|---------------|------------------|-----------------------------|-------------------------|---------------------------------------------|
+| [AQLM](./aqlm.md)                             | 🔴                   | 🟢              |     🟢     | 🔴        | 🔴                                 | 🔴              | 🟢              | 1/2         | 🟢               | 🟢                          | 🟢                      | https://github.com/Vahe1994/AQLM            |
+| [AWQ](./awq.md)                               | 🔴                   | 🟢              | 🟢        | 🟢        | 🔴                                 | 🟢              | ?               | 4             | 🟢               | 🟢                          | 🟢                      | https://github.com/casper-hansen/AutoAWQ    |
+| [bitsandbytes](./bitsandbytes.md)             | 🟢                   | 🟡 <sub>1</sub> |     🟢     | 🟡 <sub>1</sub> | 🔴 <sub>2</sub>                    | 🟡 <sub>1</sub> | 🔴 <sub>1</sub> | 4/8         | 🟢               | 🟢                          | 🟢                      | https://github.com/bitsandbytes-foundation/bitsandbytes |
+| [compressed-tensors](./compressed_tensors.md) | 🔴                   | 🟢              |     🟢     | 🟢        | 🔴                                 | 🔴              | 🔴              | 1/8         | 🟢               | 🟢                          | 🟢                      | https://github.com/neuralmagic/compressed-tensors |
+| [EETQ](./eetq.md)                             | 🟢                   | 🔴              | 🟢        | 🔴        | 🔴                                 | 🔴              | ?               | 8             | 🟢               | 🟢                          | 🟢                      | https://github.com/NetEase-FuXi/EETQ        |
+| [GGUF / GGML (llama.cpp)](../gguf.md)         | 🟢                   | 🟢              | 🟢        | 🔴        | 🟢                                 | 🔴              | 🔴              | 1/8         | 🔴               | [See Notes](../gguf.md)     | [See Notes](../gguf.md) | https://github.com/ggerganov/llama.cpp      |
+| [GPTQModel](./gptq.md)                        | 🔴                   | 🟢 <sub>3</sub> | 🟢        | 🟢        | 🟢                                 | 🟢 <sub>4</sub> | 🔴              | 2/3/4/8 | 🟢               | 🟢                          | 🟢                      | https://github.com/ModelCloud/GPTQModel        |
+| [AutoGPTQ](./gptq.md)                         | 🔴                   | 🔴              | 🟢        | 🟢        | 🔴                                 | 🔴              | 🔴              | 2/3/4/8 | 🟢               | 🟢                          | 🟢                      | https://github.com/AutoGPTQ/AutoGPTQ        |
+| [HIGGS](./higgs.md)                           | 🟢                   | 🔴              | 🟢        | 🔴        | 🔴                                 | 🔴              | 🟢              | 2/4         | 🔴               | 🟢                          | 🟢                      | https://github.com/HanGuo97/flute           |       
+| [HQQ](./hqq.md)                               | 🟢                   | 🟢              | 🟢        | 🔴        | 🔴                                 | 🔴              | 🟢              | 1/8         | 🟢               | 🔴                          | 🟢                      | https://github.com/mobiusml/hqq/            |
+| [optimum-quanto](./quanto.md)                 | 🟢                   | 🟢              | 🟢        | 🔴        | 🟢                                 | 🔴              | 🟢              | 2/4/8     | 🔴               | 🔴                          | 🟢                      | https://github.com/huggingface/optimum-quanto       |
+| [FBGEMM_FP8](./fbgemm_fp8.md)                 | 🟢                   | 🔴              | 🟢        | 🔴        | 🔴                                 | 🔴              | 🔴              | 8             | 🔴               | 🟢                          | 🟢                      | https://github.com/pytorch/FBGEMM       |
+| [torchao](./torchao.md)                       | 🟢                   |                 | 🟢        | 🔴        | 🟡 <sub>5</sub> | 🔴              |                 | 4/8         |                  | 🟢🔴                        | 🟢                      | https://github.com/pytorch/ao       |
+| [VPTQ](./vptq.md)                             | 🔴                   | 🔴              |     🟢     | 🟡        | 🔴                                 | 🔴              | 🟢              | 1/8         | 🔴               | 🟢                          | 🟢                      | https://github.com/microsoft/VPTQ            |
 
 <Tip>
+
+**1:** bitsandbytes is being refactored to support multiple backends beyond CUDA. Currently, ROCm (AMD GPU) and Intel CPU implementations are mature, with Intel XPU in progress and Apple Silicon support expected by Q4/Q1. For installation instructions and the latest backend updates, visit [this link](https://huggingface.co/docs/bitsandbytes/main/en/installation#multi-backend). Check out [these docs](https://huggingface.co/docs/bitsandbytes/main/en/non_cuda_backends) for more details and feedback links.
 
-\* bitsandbytes is being refactored to support multiple backends beyond CUDA. Currently, ROCm (AMD GPU) and Intel CPU implementations are mature, with Intel XPU in progress and Apple Silicon support expected by Q4/Q1. For installation instructions and the latest backend updates, visit [this link](https://huggingface.co/docs/bitsandbytes/main/en/installation#multi-backend).
+</Tip>
+
+<Tip>
+
+**2:** bitsandbytes is seeking contributors to help develop and lead the Apple Silicon backend. Interested? Contact them directly via their repo. Stipends may be available through sponsorships.
+
+</Tip>
+
+<Tip>
 
-We value your feedback to help identify bugs before the full release! Check out [these docs](https://huggingface.co/docs/bitsandbytes/main/en/non_cuda_backends) for more details and feedback links.
+**3:** GPTQModel[CPU] supports 4-bit via IPEX on Intel/AMD and full bit range via Torch on Intel/AMD/Apple Silicon.
 
 </Tip>
 
 <Tip>
 
-\** bitsandbytes is seeking contributors to help develop and lead the Apple Silicon backend. Interested? Contact them directly via their repo. Stipends may be available through sponsorships.
+**4:** GPTQModel[Intel GPU] via IPEX only supports 4-bit for Intel Datacenter Max/Arc GPUs.
+
+</Tip>
+
+<Tip>
+
+**5:** torchao only supports int4 weight on Metal (Apple Silicon).
+
+</Tip>
 
-</Tip>