simcop2387/AutoGPTQ
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages 1 Wiki Activity

update README

Browse source
This commit is contained in:
PanQiWei 2023-06-05 22:53:17 +08:00
parent 923fc87a11
commit b132d774e3
3 changed files with 27 additions and 30 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

28
README.md
View file

@ -17,9 +17,7 @@
## News or Update ## News or Update
**To experience adapter training using `auto_gptq` quantized model in advance, you can try [this branch](https://github.com/PanQiWei/AutoGPTQ/tree/peft_integration) and discuss [in here](https://github.com/PanQiWei/AutoGPTQ/issues/103), examples are [in here](https://github.com/PanQiWei/AutoGPTQ/tree/peft_integration/examples/peft).** - 2023-06-05 - (Update) - Integrate with 🤗 peft to use gptq quantized model to train adapters, support LoRA, AdaLoRA, AdaptionPrompt, etc.
- 2023-05-25 - (In Progress) - Integrate with 🤗 peft to use gptq quantized model to train adapters, support LoRA, AdaLoRA, AdaptionPrompt, etc.
- 2023-05-30 - (Update) - Support download/upload quantized model from/to 🤗 Hub. - 2023-05-30 - (Update) - Support download/upload quantized model from/to 🤗 Hub.
- 2023-05-27 - (Update) - Support quantization and inference for `gpt_bigcode`, `codegen` and `RefineWeb/RefineWebModel`(falcon) model types. - 2023-05-27 - (Update) - Support quantization and inference for `gpt_bigcode`, `codegen` and `RefineWeb/RefineWebModel`(falcon) model types.
- 2023-05-04 - (Update) - Support using faster cuda kernel when `not desc_act or group_size == -1`. - 2023-05-04 - (Update) - Support using faster cuda kernel when `not desc_act or group_size == -1`.
@ -298,18 +296,18 @@ print(
> >
> for example, model_type of `WizardLM`, `vicuna` and `gpt4all` are all `llama`, hence they are all 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-adaption_prompt | | model type | quantization | inference | peft-lora | peft-ada-lora | peft-adaption_prompt |
|------------------------------------|--------------|-----------|-----------|----------------------| |------------------------------------|--------------|-----------|-----------|---------------|-------------------------------------------------------------------------------------------------|
| bloom | ✅ | ✅ | | | | bloom | ✅ | ✅ | ✅ | ✅ | |
| gpt2 | ✅ | ✅ | | | | gpt2 | ✅ | ✅ | ✅ | ✅ | |
| gpt_neox | ✅ | ✅ | | | | gpt_neox | ✅ | ✅ | ✅ | ✅ | ✅[requires this peft branch](https://github.com/PanQiWei/peft/tree/multi_modal_adaption_prompt) |
| gptj | ✅ | ✅ | | | | gptj | ✅ | ✅ | ✅ | ✅ | ✅[requires this peft branch](https://github.com/PanQiWei/peft/tree/multi_modal_adaption_prompt) |
| llama | ✅ | ✅ | | ✅ | | llama | ✅ | ✅ | ✅ | ✅ | ✅ |
| moss | ✅ | ✅ | | | | moss | ✅ | ✅ | ✅ | ✅ | ✅[requires this peft branch](https://github.com/PanQiWei/peft/tree/multi_modal_adaption_prompt) |
| opt | ✅ | ✅ | | | | opt | ✅ | ✅ | ✅ | ✅ | |
| gpt_bigcode | ✅ | ✅ | | | | gpt_bigcode | ✅ | ✅ | ✅ | ✅ | |
| codegen | ✅ | ✅ | | | | codegen | ✅ | ✅ | ✅ | ✅ | |
| falcon(RefinedWebModel/RefinedWeb) | ✅ | ✅ | | | | falcon(RefinedWebModel/RefinedWeb) | ✅ | ✅ | ✅ | ✅ | |
## Supported Evaluation Tasks ## Supported Evaluation Tasks
Currently, `auto_gptq` supports: `LanguageModelingTask`, `SequenceClassificationTask` and `TextSummarizationTask`; more Tasks will come soon! Currently, `auto_gptq` supports: `LanguageModelingTask`, `SequenceClassificationTask` and `TextSummarizationTask`; more Tasks will come soon!

28
README_zh.md
View file

@ -17,9 +17,7 @@
## 新闻或更新 ## 新闻或更新
**提前体验使用 `auto_gptq` 量化过的模型来训练适应层，你可以尝试[这个分支](https://github.com/PanQiWei/AutoGPTQ/tree/peft_integration) 并在[这里](https://github.com/PanQiWei/AutoGPTQ/issues/103)进行讨论，你也可以参考[这里](https://github.com/PanQiWei/AutoGPTQ/tree/peft_integration/examples/peft)所提供的示例脚本。** - 2023-06-05 - (更新) - 集成 🤗 peft 来使用 gptq 量化过的模型训练适应层，支持 LoRA，AdaLoRA，AdaptionPrompt 等。
- 2023-05-25 - (开发中) - 集成 🤗 peft 来使用 gptq 量化过的模型训练适应层，支持 LoRA，AdaLoRA，AdaptionPrompt 等。
- 2023-05-30 - (更新) - 支持从 🤗 Hub 下载量化好的模型或上次量化好的模型到 🤗 Hub。 - 2023-05-30 - (更新) - 支持从 🤗 Hub 下载量化好的模型或上次量化好的模型到 🤗 Hub。
- 2023-05-27 - (更新) - 支持以下模型的量化和推理： `gpt_bigcode`， `codegen` 以及 `RefineWeb/RefineWebModel`（falcon）。 - 2023-05-27 - (更新) - 支持以下模型的量化和推理： `gpt_bigcode`， `codegen` 以及 `RefineWeb/RefineWebModel`（falcon）。
- 2023-05-04 - (更新) - 支持在 `not desc_act or group_size == -1` 的情况下使用更快的 cuda 算子。 - 2023-05-04 - (更新) - 支持在 `not desc_act or group_size == -1` 的情况下使用更快的 cuda 算子。
@ -297,18 +295,18 @@ print(
> >
> 比如， `WizardLM`，`vicuna` 和 `gpt4all` 模型的 `model_type` 皆为 `llama`， 因此这些模型皆被 `auto_gptq` 所支持。 > 比如， `WizardLM`，`vicuna` 和 `gpt4all` 模型的 `model_type` 皆为 `llama`， 因此这些模型皆被 `auto_gptq` 所支持。
| model type | quantization | inference | peft-lora | peft-adaption_prompt | | model type | quantization | inference | peft-lora | peft-ada-lora | peft-adaption_prompt |
|------------------------------------|--------------|-----------|-----------|----------------------| |------------------------------------|--------------|-----------|-----------|---------------|-----------------------------------------------------------------------------------|
| bloom | ✅ | ✅ | | | | bloom | ✅ | ✅ | ✅ | ✅ | |
| gpt2 | ✅ | ✅ | | | | gpt2 | ✅ | ✅ | ✅ | ✅ | |
| gpt_neox | ✅ | ✅ | | | | gpt_neox | ✅ | ✅ | ✅ | ✅ | ✅[要求该分支的 peft](https://github.com/PanQiWei/peft/tree/multi_modal_adaption_prompt) |
| gptj | ✅ | ✅ | | | | gptj | ✅ | ✅ | ✅ | ✅ | ✅[要求该分支的 peft](https://github.com/PanQiWei/peft/tree/multi_modal_adaption_prompt) |
| llama | ✅ | ✅ | | ✅ | | llama | ✅ | ✅ | ✅ | ✅ | ✅ |
| moss | ✅ | ✅ | | | | moss | ✅ | ✅ | ✅ | ✅ | ✅[要求该分支的 peft](https://github.com/PanQiWei/peft/tree/multi_modal_adaption_prompt) |
| opt | ✅ | ✅ | | | | opt | ✅ | ✅ | ✅ | ✅ | |
| gpt_bigcode | ✅ | ✅ | | | | gpt_bigcode | ✅ | ✅ | ✅ | ✅ | |
| codegen | ✅ | ✅ | | | | codegen | ✅ | ✅ | ✅ | ✅ | |
| falcon(RefinedWebModel/RefinedWeb) | ✅ | ✅ | | | | falcon(RefinedWebModel/RefinedWeb) | ✅ | ✅ | ✅ | ✅ | |
## 支持的评估任务 ## 支持的评估任务
目前， `auto_gptq` 支持以下评估任务： `LanguageModelingTask`, `SequenceClassificationTask` 和 `TextSummarizationTask`；更多的评估任务即将到来！ 目前， `auto_gptq` 支持以下评估任务： `LanguageModelingTask`, `SequenceClassificationTask` 和 `TextSummarizationTask`；更多的评估任务即将到来！

1
docs/NEWS_OR_UPDATE.md
View file

@ -1,4 +1,5 @@
## <center>News or Update</center> ## <center>News or Update</center>
- 2023-06-05 - (Update) - Integrate with 🤗 peft to use gptq quantized model to train adapters, support LoRA, AdaLoRA, AdaptionPrompt, etc.
- 2023-05-30 - (Update) - support download/upload quantized model from/to 🤗 Hub. - 2023-05-30 - (Update) - support download/upload quantized model from/to 🤗 Hub.
- 2023-05-27 - (Update) - Support quantization and inference for `gpt_bigcode`, `codegen` and `RefineWeb/RefineWebModel`(falcon) model types. - 2023-05-27 - (Update) - Support quantization and inference for `gpt_bigcode`, `codegen` and `RefineWeb/RefineWebModel`(falcon) model types.
- 2023-05-04 - (Update) - Support using faster cuda kernel when `not desc_act or group_size == -1` - 2023-05-04 - (Update) - Support using faster cuda kernel when `not desc_act or group_size == -1`