Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/panqiwei/autogptq
An easy-to-use LLMs quantization package with user-friendly apis, based on GPTQ algorithm.
https://github.com/panqiwei/autogptq
deep-learning inference large-language-models llms nlp pytorch quantization transformer transformers
Last synced: 9 days ago
JSON representation
An easy-to-use LLMs quantization package with user-friendly apis, based on GPTQ algorithm.
- Host: GitHub
- URL: https://github.com/panqiwei/autogptq
- Owner: AutoGPTQ
- License: mit
- Created: 2023-04-13T02:18:11.000Z (over 1 year ago)
- Default Branch: main
- Last Pushed: 2024-11-27T04:51:41.000Z (25 days ago)
- Last Synced: 2024-12-11T17:53:14.506Z (10 days ago)
- Topics: deep-learning, inference, large-language-models, llms, nlp, pytorch, quantization, transformer, transformers
- Language: Python
- Homepage:
- Size: 7.98 MB
- Stars: 4,557
- Watchers: 31
- Forks: 492
- Open Issues: 266
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
- awesome-ChatGPT-repositories - AutoGPTQ - An easy-to-use LLMs quantization package with user-friendly apis, based on GPTQ algorithm. (NLP)
README
AutoGPTQ
An easy-to-use LLM quantization package with user-friendly APIs, based on GPTQ algorithm (weight-only quantization).
English |
ไธญๆ
## News or Update
- 2024-02-15 - (News) - AutoGPTQ 0.7.0 is released, with [Marlin](https://github.com/IST-DASLab/marlin) int4*fp16 matrix multiplication kernel support, with the argument `use_marlin=True` when loading models.
- 2023-08-23 - (News) - ๐ค Transformers, optimum and peft have integrated `auto-gptq`, so now running and training GPTQ models can be more available to everyone! See [this blog](https://huggingface.co/blog/gptq-integration) and it's resources for more details!
*For more histories please turn to [here](docs/NEWS_OR_UPDATE.md)*
## Performance Comparison
### Inference Speed
> The result is generated using [this script](examples/benchmark/generation_speed.py), batch size of input is 1, decode strategy is beam search and enforce the model to generate 512 tokens, speed metric is tokens/s (the larger, the better).
>
> The quantized model is loaded using the setup that can gain the fastest inference speed.
| model | GPU | num_beams | fp16 | gptq-int4 |
|---------------|---------------|-----------|-------|-----------|
| llama-7b | 1xA100-40G | 1 | 18.87 | 25.53 |
| llama-7b | 1xA100-40G | 4 | 68.79 | 91.30 |
| moss-moon 16b | 1xA100-40G | 1 | 12.48 | 15.25 |
| moss-moon 16b | 1xA100-40G | 4 | OOM | 42.67 |
| moss-moon 16b | 2xA100-40G | 1 | 06.83 | 06.78 |
| moss-moon 16b | 2xA100-40G | 4 | 13.10 | 10.80 |
| gpt-j 6b | 1xRTX3060-12G | 1 | OOM | 29.55 |
| gpt-j 6b | 1xRTX3060-12G | 4 | OOM | 47.36 |
### Perplexity
For perplexity comparison, you can turn to [here](https://github.com/qwopqwop200/GPTQ-for-LLaMa#result) and [here](https://github.com/qwopqwop200/GPTQ-for-LLaMa#gptq-vs-bitsandbytes)
## Installation
AutoGPTQ is available on Linux and Windows only. You can install the latest stable release of AutoGPTQ from pip with pre-built wheels:
| Platform version | Installation | Built against PyTorch |
|-------------------|---------------------------------------------------------------------------------------------------|-----------------------|
| CUDA 11.8 | `pip install auto-gptq --no-build-isolation --extra-index-url https://huggingface.github.io/autogptq-index/whl/cu118/` | 2.2.1+cu118 |
| CUDA 12.1 | `pip install auto-gptq --no-build-isolation` | 2.2.1+cu121 |
| ROCm 5.7 | `pip install auto-gptq --no-build-isolation --extra-index-url https://huggingface.github.io/autogptq-index/whl/rocm573/` | 2.2.1+rocm5.7
AutoGPTQ can be installed with the Triton dependency with `pip install auto-gptq[triton] --no-build-isolation` in order to be able to use the Triton backend (currently only supports linux, no 3-bits quantization).
For older AutoGPTQ, please refer to [the previous releases installation table](docs/INSTALLATION.md).
On NVIDIA systems, AutoGPTQ does not support [Maxwell or lower](https://qiita.com/uyuni/items/733a93b975b524f89f46) GPUs.
### Install from source
Clone the source code:
```bash
git clone https://github.com/PanQiWei/AutoGPTQ.git && cd AutoGPTQ
```
A few packages are required in order to build from source: `pip install numpy gekko pandas`.
Then, install locally from source:
```bash
pip install -vvv --no-build-isolation -e .
```
You can set `BUILD_CUDA_EXT=0` to disable pytorch extension building, but this is **strongly discouraged** as AutoGPTQ then falls back on a slow python implementation.
As a last resort, if the above command fails, you can try `python setup.py install`.
#### On ROCm systems
To install from source for AMD GPUs supporting ROCm, please specify the `ROCM_VERSION` environment variable. Example:
```bash
ROCM_VERSION=5.6 pip install -vvv --no-build-isolation -e .
```
The compilation can be speeded up by specifying the `PYTORCH_ROCM_ARCH` variable ([reference](https://github.com/pytorch/pytorch/blob/7b73b1e8a73a1777ebe8d2cd4487eb13da55b3ba/setup.py#L132)) in order to build for a single target device, for example `gfx90a` for MI200 series devices.
For ROCm systems, the packages `rocsparse-dev`, `hipsparse-dev`, `rocthrust-dev`, `rocblas-dev` and `hipblas-dev` are required to build.
#### On Intelยฎ Gaudiยฎ 2 systems
>Notice: make sure you're in commit 65c2e15 or later
To install from source for Intel Gaudi 2 HPUs, set the `BUILD_CUDA_EXT=0` environment variable to disable building the CUDA PyTorch extension. Example:
```bash
BUILD_CUDA_EXT=0 pip install -vvv --no-build-isolation -e .
```
>Notice that Intel Gaudi 2 uses an optimized kernel upon inference, and requires `BUILD_CUDA_EXT=0` on non-CUDA machines.
## Quick Tour
### Quantization and Inference
> warning: this is just a showcase of the usage of basic apis in AutoGPTQ, which uses only one sample to quantize a much small model, quality of quantized model using such little samples may not good.
Below is an example for the simplest use of `auto_gptq` to quantize a model and inference after quantization:
```python
from transformers import AutoTokenizer, TextGenerationPipeline
from auto_gptq import AutoGPTQForCausalLM, BaseQuantizeConfig
import logging
logging.basicConfig(
format="%(asctime)s %(levelname)s [%(name)s] %(message)s", level=logging.INFO, datefmt="%Y-%m-%d %H:%M:%S"
)
pretrained_model_dir = "facebook/opt-125m"
quantized_model_dir = "opt-125m-4bit"
tokenizer = AutoTokenizer.from_pretrained(pretrained_model_dir, use_fast=True)
examples = [
tokenizer(
"auto-gptq is an easy-to-use model quantization library with user-friendly apis, based on GPTQ algorithm."
)
]
quantize_config = BaseQuantizeConfig(
bits=4, # quantize model to 4-bit
group_size=128, # it is recommended to set the value to 128
desc_act=False, # set to False can significantly speed up inference but the perplexity may slightly bad
)
# load un-quantized model, by default, the model will always be loaded into CPU memory
model = AutoGPTQForCausalLM.from_pretrained(pretrained_model_dir, quantize_config)
# quantize model, the examples should be list of dict whose keys can only be "input_ids" and "attention_mask"
model.quantize(examples)
# save quantized model
model.save_quantized(quantized_model_dir)
# save quantized model using safetensors
model.save_quantized(quantized_model_dir, use_safetensors=True)
# push quantized model to Hugging Face Hub.
# to use use_auth_token=True, Login first via huggingface-cli login.
# or pass explcit token with: use_auth_token="hf_xxxxxxx"
# (uncomment the following three lines to enable this feature)
# repo_id = f"YourUserName/{quantized_model_dir}"
# commit_message = f"AutoGPTQ model for {pretrained_model_dir}: {quantize_config.bits}bits, gr{quantize_config.group_size}, desc_act={quantize_config.desc_act}"
# model.push_to_hub(repo_id, commit_message=commit_message, use_auth_token=True)
# alternatively you can save and push at the same time
# (uncomment the following three lines to enable this feature)
# repo_id = f"YourUserName/{quantized_model_dir}"
# commit_message = f"AutoGPTQ model for {pretrained_model_dir}: {quantize_config.bits}bits, gr{quantize_config.group_size}, desc_act={quantize_config.desc_act}"
# model.push_to_hub(repo_id, save_dir=quantized_model_dir, use_safetensors=True, commit_message=commit_message, use_auth_token=True)
# load quantized model to the first GPU
model = AutoGPTQForCausalLM.from_quantized(quantized_model_dir, device="cuda:0")
# download quantized model from Hugging Face Hub and load to the first GPU
# model = AutoGPTQForCausalLM.from_quantized(repo_id, device="cuda:0", use_safetensors=True, use_triton=False)
# inference with model.generate
print(tokenizer.decode(model.generate(**tokenizer("auto_gptq is", return_tensors="pt").to(model.device))[0]))
# or you can also use pipeline
pipeline = TextGenerationPipeline(model=model, tokenizer=tokenizer)
print(pipeline("auto-gptq is")[0]["generated_text"])
```
For more advanced features of model quantization, please reference to [this script](examples/quantization/quant_with_alpaca.py)
### Customize Model
Below is an example to extend `auto_gptq` to support `OPT` model, as you will see, it's very easy:
```python
from auto_gptq.modeling import BaseGPTQForCausalLM
class OPTGPTQForCausalLM(BaseGPTQForCausalLM):
# chained attribute name of transformer layer block
layers_block_name = "model.decoder.layers"
# chained attribute names of other nn modules that in the same level as the transformer layer block
outside_layer_modules = [
"model.decoder.embed_tokens", "model.decoder.embed_positions", "model.decoder.project_out",
"model.decoder.project_in", "model.decoder.final_layer_norm"
]
# chained attribute names of linear layers in transformer layer module
# normally, there are four sub lists, for each one the modules in it can be seen as one operation,
# and the order should be the order when they are truly executed, in this case (and usually in most cases),
# they are: attention q_k_v projection, attention output projection, MLP project input, MLP project output
inside_layer_modules = [
["self_attn.k_proj", "self_attn.v_proj", "self_attn.q_proj"],
["self_attn.out_proj"],
["fc1"],
["fc2"]
]
```
After this, you can use `OPTGPTQForCausalLM.from_pretrained` and other methods as shown in Basic.
### Evaluation on Downstream Tasks
You can use tasks defined in `auto_gptq.eval_tasks` to evaluate model's performance on specific down-stream task before and after quantization.
The predefined tasks support all causal-language-models implemented in [๐ค transformers](https://github.com/huggingface/transformers) and in this project.
Below is an example to evaluate `EleutherAI/gpt-j-6b` on sequence-classification task using `cardiffnlp/tweet_sentiment_multilingual` dataset:
```python
from functools import partial
import datasets
from transformers import AutoTokenizer, AutoModelForCausalLM, GenerationConfig
from auto_gptq import AutoGPTQForCausalLM, BaseQuantizeConfig
from auto_gptq.eval_tasks import SequenceClassificationTask
MODEL = "EleutherAI/gpt-j-6b"
DATASET = "cardiffnlp/tweet_sentiment_multilingual"
TEMPLATE = "Question:What's the sentiment of the given text? Choices are {labels}.\nText: {text}\nAnswer:"
ID2LABEL = {
0: "negative",
1: "neutral",
2: "positive"
}
LABELS = list(ID2LABEL.values())
def ds_refactor_fn(samples):
text_data = samples["text"]
label_data = samples["label"]
new_samples = {"prompt": [], "label": []}
for text, label in zip(text_data, label_data):
prompt = TEMPLATE.format(labels=LABELS, text=text)
new_samples["prompt"].append(prompt)
new_samples["label"].append(ID2LABEL[label])
return new_samples
# model = AutoModelForCausalLM.from_pretrained(MODEL).eval().half().to("cuda:0")
model = AutoGPTQForCausalLM.from_pretrained(MODEL, BaseQuantizeConfig())
tokenizer = AutoTokenizer.from_pretrained(MODEL)
task = SequenceClassificationTask(
model=model,
tokenizer=tokenizer,
classes=LABELS,
data_name_or_path=DATASET,
prompt_col_name="prompt",
label_col_name="label",
**{
"num_samples": 1000, # how many samples will be sampled to evaluation
"sample_max_len": 1024, # max tokens for each sample
"block_max_len": 2048, # max tokens for each data block
# function to load dataset, one must only accept data_name_or_path as input
# and return datasets.Dataset
"load_fn": partial(datasets.load_dataset, name="english"),
# function to preprocess dataset, which is used for datasets.Dataset.map,
# must return Dict[str, list] with only two keys: [prompt_col_name, label_col_name]
"preprocess_fn": ds_refactor_fn,
# truncate label when sample's length exceed sample_max_len
"truncate_prompt": False
}
)
# note that max_new_tokens will be automatically specified internally based on given classes
print(task.run())
# self-consistency
print(
task.run(
generation_config=GenerationConfig(
num_beams=3,
num_return_sequences=3,
do_sample=True
)
)
)
```
## Learn More
[tutorials](docs/tutorial) provide step-by-step guidance to integrate `auto_gptq` with your own project and some best practice principles.
[examples](examples/README.md) provide plenty of example scripts to use `auto_gptq` in different ways.
## Supported Models
> you can use `model.config.model_type` to compare with the table below to check whether the model you use is supported by `auto_gptq`.
>
> for example, model_type of `WizardLM`, `vicuna` and `gpt4all` are all `llama`, hence they are all supported by `auto_gptq`.
| model type | quantization | inference | peft-lora | peft-ada-lora | peft-adaption_prompt |
|------------------------------------|--------------|-----------|-----------|---------------|-------------------------------------------------------------------------------------------------|
| bloom | โ
| โ
| โ
| โ
| |
| gpt2 | โ
| โ
| โ
| โ
| |
| gpt_neox | โ
| โ
| โ
| โ
| โ
[requires this peft branch](https://github.com/PanQiWei/peft/tree/multi_modal_adaption_prompt) |
| gptj | โ
| โ
| โ
| โ
| โ
[requires this peft branch](https://github.com/PanQiWei/peft/tree/multi_modal_adaption_prompt) |
| llama | โ
| โ
| โ
| โ
| โ
|
| moss | โ
| โ
| โ
| โ
| โ
[requires this peft branch](https://github.com/PanQiWei/peft/tree/multi_modal_adaption_prompt) |
| opt | โ
| โ
| โ
| โ
| |
| gpt_bigcode | โ
| โ
| โ
| โ
| |
| codegen | โ
| โ
| โ
| โ
| |
| falcon(RefinedWebModel/RefinedWeb) | โ
| โ
| โ
| โ
| |
## Supported Evaluation Tasks
Currently, `auto_gptq` supports: `LanguageModelingTask`, `SequenceClassificationTask` and `TextSummarizationTask`; more Tasks will come soon!
## Running tests
Tests can be run with:
```
pytest tests/ -s
```
## FAQ
### Which kernel is used by default?
AutoGPTQ defaults to using exllamav2 int4*fp16 kernel for matrix multiplication.
### How to use Marlin kernel?
Marlin is an optimized int4 * fp16 kernel was recently proposed at https://github.com/IST-DASLab/marlin. This is integrated in AutoGPTQ when loading a model with `use_marlin=True`. This kernel is available only on devices with compute capability 8.0 or 8.6 (Ampere GPUs).
## Acknowledgement
- Special thanks **Elias Frantar**, **Saleh Ashkboos**, **Torsten Hoefler** and **Dan Alistarh** for proposing **GPTQ** algorithm and open source the [code](https://github.com/IST-DASLab/gptq), and for releasing [Marlin kernel](https://github.com/IST-DASLab/marlin) for mixed precision computation.
- Special thanks **qwopqwop200**, for code in this project that relevant to quantization are mainly referenced from [GPTQ-for-LLaMa](https://github.com/qwopqwop200/GPTQ-for-LLaMa/tree/cuda).
- Special thanks to **turboderp**, for releasing [Exllama](https://github.com/turboderp/exllama) and [Exllama v2](https://github.com/turboderp/exllamav2) libraries with efficient mixed precision kernels.