github
/
ColossalAI
mirror of https://github.com/hpcaitech/ColossalAI
1
0
Fork 0
Code Issues Projects Releases Wiki Activity

[inference] release (#5747)

Browse Source 
* [inference] release

* [inference] release

* [inference] release

* [inference] release

* [inference] release

* [inference] release

* [inference] release
pull/5713/head^2
binmakeswell 6 months ago committed by GitHub
parent df6747603f
commit 4647ec28c8
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
3 changed files with 39 additions and 44 deletions
Show Stats Download Patch File Download Diff File

37
README.md
Unescape View File

@ -25,6 +25,7 @@
</div> </div>
## Latest News ## Latest News
* [2024/05] [Large AI Models Inference Speed Doubled, Colossal-Inference Open Source Release](https://hpc-ai.com/blog/colossal-inference)
* [2024/04] [Open-Sora Unveils Major Upgrade: Embracing Open Source with Single-Shot 16-Second Video Generation and 720p Resolution](https://hpc-ai.com/blog/open-soras-comprehensive-upgrade-unveiled-embracing-16-second-video-generation-and-720p-resolution-in-open-source) * [2024/04] [Open-Sora Unveils Major Upgrade: Embracing Open Source with Single-Shot 16-Second Video Generation and 720p Resolution](https://hpc-ai.com/blog/open-soras-comprehensive-upgrade-unveiled-embracing-16-second-video-generation-and-720p-resolution-in-open-source)
* [2024/04] [Most cost-effective solutions for inference, fine-tuning and pretraining, tailored to LLaMA3 series](https://hpc-ai.com/blog/most-cost-effective-solutions-for-inference-fine-tuning-and-pretraining-tailored-to-llama3-series) * [2024/04] [Most cost-effective solutions for inference, fine-tuning and pretraining, tailored to LLaMA3 series](https://hpc-ai.com/blog/most-cost-effective-solutions-for-inference-fine-tuning-and-pretraining-tailored-to-llama3-series)
* [2024/03] [314 Billion Parameter Grok-1 Inference Accelerated by 3.8x, Efficient and Easy-to-Use PyTorch+HuggingFace version is Here](https://hpc-ai.com/blog/314-billion-parameter-grok-1-inference-accelerated-by-3.8x-efficient-and-easy-to-use-pytorchhuggingface-version-is-here) * [2024/03] [314 Billion Parameter Grok-1 Inference Accelerated by 3.8x, Efficient and Easy-to-Use PyTorch+HuggingFace version is Here](https://hpc-ai.com/blog/314-billion-parameter-grok-1-inference-accelerated-by-3.8x-efficient-and-easy-to-use-pytorchhuggingface-version-is-here)
@ -75,11 +76,9 @@
<li> <li>
<a href="#Inference">Inference</a> <a href="#Inference">Inference</a>
<ul> <ul>
<li><a href="#Colossal-Inference">Colossal-Inference: Large AI Models Inference Speed Doubled</a></li>
<li><a href="#Grok-1">Grok-1: 314B model of PyTorch + HuggingFace Inference</a></li> <li><a href="#Grok-1">Grok-1: 314B model of PyTorch + HuggingFace Inference</a></li>
<li><a href="#SwiftInfer">SwiftInfer:Breaks the Length Limit of LLM for Multi-Round Conversations with 46% Acceleration</a></li> <li><a href="#SwiftInfer">SwiftInfer:Breaks the Length Limit of LLM for Multi-Round Conversations with 46% Acceleration</a></li>
<li><a href="#GPT-3-Inference">GPT-3</a></li>
<li><a href="#OPT-Serving">OPT-175B Online Serving for Text Generation</a></li>
<li><a href="#BLOOM-Inference">176B BLOOM</a></li>
</ul> </ul>
</li> </li>
<li> <li>
@ -377,6 +376,19 @@ Please visit our [documentation](https://www.colossalai.org/) and [examples](htt
## Inference ## Inference
### Colossal-Inference
<p align="center">
<img src="https://raw.githubusercontent.com/hpcaitech/public_assets/main/colossalai/img/inference/colossal-inference-v1-1.png" width=1000/>
</p>
<p align="center">
<img src="https://raw.githubusercontent.com/hpcaitech/public_assets/main/colossalai/img/inference/colossal-inference-v1-2.png" width=1000/>
</p>
- Large AI models inference speed doubled, compared to the offline inference performance of vLLM in some cases.
[[code]](https://github.com/hpcaitech/ColossalAI/tree/main/colossalai/inference)
[[blog]](https://hpc-ai.com/blog/colossal-inference)
### Grok-1 ### Grok-1
<p id="Grok-1" align="center"> <p id="Grok-1" align="center">
<img src="https://raw.githubusercontent.com/hpcaitech/public_assets/main/examples/images/grok-1-inference.jpg" width=600/> <img src="https://raw.githubusercontent.com/hpcaitech/public_assets/main/examples/images/grok-1-inference.jpg" width=600/>
@ -389,30 +401,13 @@ Please visit our [documentation](https://www.colossalai.org/) and [examples](htt
[[HuggingFace Grok-1 PyTorch model weights]](https://huggingface.co/hpcai-tech/grok-1) [[HuggingFace Grok-1 PyTorch model weights]](https://huggingface.co/hpcai-tech/grok-1)
[[ModelScope Grok-1 PyTorch model weights]](https://www.modelscope.cn/models/colossalai/grok-1-pytorch/summary) [[ModelScope Grok-1 PyTorch model weights]](https://www.modelscope.cn/models/colossalai/grok-1-pytorch/summary)
### SwiftInfer
<p id="SwiftInfer" align="center"> <p id="SwiftInfer" align="center">
<img src="https://raw.githubusercontent.com/hpcaitech/public_assets/main/colossalai/img/SwiftInfer.jpg" width=800/> <img src="https://raw.githubusercontent.com/hpcaitech/public_assets/main/colossalai/img/SwiftInfer.jpg" width=800/>
</p> </p>
- [SwiftInfer](https://github.com/hpcaitech/SwiftInfer): Inference performance improved by 46%, open source solution breaks the length limit of LLM for multi-round conversations - [SwiftInfer](https://github.com/hpcaitech/SwiftInfer): Inference performance improved by 46%, open source solution breaks the length limit of LLM for multi-round conversations
<p id="GPT-3-Inference" align="center">
<img src="https://raw.githubusercontent.com/hpcaitech/public_assets/main/colossalai/img/inference_GPT-3.jpg" width=800/>
</p>
- [Energon-AI](https://github.com/hpcaitech/EnergonAI): 50% inference acceleration on the same hardware
<p id="OPT-Serving" align="center">
<img src="https://raw.githubusercontent.com/hpcaitech/public_assets/main/colossalai/img/BLOOM%20serving.png" width=600/>
</p>
- [OPT Serving](https://colossalai.org/docs/advanced_tutorials/opt_service): Try 175-billion-parameter OPT online services
<p id="BLOOM-Inference" align="center">
<img src="https://raw.githubusercontent.com/hpcaitech/public_assets/main/colossalai/img/BLOOM%20Inference.PNG" width=800/>
</p>
- [BLOOM](https://github.com/hpcaitech/EnergonAI/tree/main/examples/bloom): Reduce hardware deployment costs of 176-billion-parameter BLOOM by more than 10 times.
<p align="right">(<a href="#top">back to top</a>)</p> <p align="right">(<a href="#top">back to top</a>)</p>
## Installation ## Installation

9
colossalai/inference/README.md
Unescape View File

@ -18,8 +18,15 @@
## 📌 Introduction ## 📌 Introduction
ColossalAI-Inference is a module which offers acceleration to the inference execution of Transformers models, especially LLMs. In ColossalAI-Inference, we leverage high-performance kernels, KV cache, paged attention, continous batching and other techniques to accelerate the inference of LLMs. We also provide simple and unified APIs for the sake of user-friendliness. ColossalAI-Inference is a module which offers acceleration to the inference execution of Transformers models, especially LLMs. In ColossalAI-Inference, we leverage high-performance kernels, KV cache, paged attention, continous batching and other techniques to accelerate the inference of LLMs. We also provide simple and unified APIs for the sake of user-friendliness. [[blog]](https://hpc-ai.com/blog/colossal-inference)
<p align="center">
<img src="https://raw.githubusercontent.com/hpcaitech/public_assets/main/colossalai/img/inference/colossal-inference-v1-1.png" width=1000/>
</p>
<p align="center">
<img src="https://raw.githubusercontent.com/hpcaitech/public_assets/main/colossalai/img/inference/colossal-inference-v1-2.png" width=1000/>
</p>
## 🕹 Usage ## 🕹 Usage

37
docs/README-zh-Hans.md
Unescape View File

@ -24,6 +24,7 @@
</div> </div>
## 新闻 ## 新闻
* [2024/05] [Large AI Models Inference Speed Doubled, Colossal-Inference Open Source Release](https://hpc-ai.com/blog/colossal-inference)
* [2024/04] [Open-Sora Unveils Major Upgrade: Embracing Open Source with Single-Shot 16-Second Video Generation and 720p Resolution](https://hpc-ai.com/blog/open-soras-comprehensive-upgrade-unveiled-embracing-16-second-video-generation-and-720p-resolution-in-open-source) * [2024/04] [Open-Sora Unveils Major Upgrade: Embracing Open Source with Single-Shot 16-Second Video Generation and 720p Resolution](https://hpc-ai.com/blog/open-soras-comprehensive-upgrade-unveiled-embracing-16-second-video-generation-and-720p-resolution-in-open-source)
* [2024/04] [Most cost-effective solutions for inference, fine-tuning and pretraining, tailored to LLaMA3 series](https://hpc-ai.com/blog/most-cost-effective-solutions-for-inference-fine-tuning-and-pretraining-tailored-to-llama3-series) * [2024/04] [Most cost-effective solutions for inference, fine-tuning and pretraining, tailored to LLaMA3 series](https://hpc-ai.com/blog/most-cost-effective-solutions-for-inference-fine-tuning-and-pretraining-tailored-to-llama3-series)
* [2024/03] [314 Billion Parameter Grok-1 Inference Accelerated by 3.8x, Efficient and Easy-to-Use PyTorch+HuggingFace version is Here](https://hpc-ai.com/blog/314-billion-parameter-grok-1-inference-accelerated-by-3.8x-efficient-and-easy-to-use-pytorchhuggingface-version-is-here) * [2024/03] [314 Billion Parameter Grok-1 Inference Accelerated by 3.8x, Efficient and Easy-to-Use PyTorch+HuggingFace version is Here](https://hpc-ai.com/blog/314-billion-parameter-grok-1-inference-accelerated-by-3.8x-efficient-and-easy-to-use-pytorchhuggingface-version-is-here)
@ -74,11 +75,9 @@
<li> <li>
<a href="#推理">推理</a> <a href="#推理">推理</a>
<ul> <ul>
<li><a href="#Colossal-Inference">Colossal-Inference: AI大模型推理速度翻倍</a></li>
<li><a href="#Grok-1">Grok-1: 3140亿参数PyTorch + HuggingFace推理</a></li> <li><a href="#Grok-1">Grok-1: 3140亿参数PyTorch + HuggingFace推理</a></li>
<li><a href="#SwiftInfer">SwiftInfer:打破LLM多轮对话的长度限制，推理加速46%</a></li> <li><a href="#SwiftInfer">SwiftInfer:打破LLM多轮对话的长度限制，推理加速46%</a></li>
<li><a href="#GPT-3-Inference">GPT-3</a></li>
<li><a href="#OPT-Serving">1750亿参数OPT在线推理服务</a></li>
<li><a href="#BLOOM-Inference">1760亿参数 BLOOM</a></li>
</ul> </ul>
</li> </li>
<li> <li>
@ -370,6 +369,19 @@ Colossal-AI 为您提供了一系列并行组件。我们的目标是让您的
## 推理 ## 推理
### Colossal-Inference
<p align="center">
<img src="https://raw.githubusercontent.com/hpcaitech/public_assets/main/colossalai/img/inference/colossal-inference-v1-1.png" width=1000/>
</p>
<p align="center">
<img src="https://raw.githubusercontent.com/hpcaitech/public_assets/main/colossalai/img/inference/colossal-inference-v1-2.png" width=1000/>
</p>
- AI大模型推理速度部分接近翻倍，与vLLM的离线推理性能相比
[[代码]](https://github.com/hpcaitech/ColossalAI/tree/main/colossalai/inference)
[[博客]](https://hpc-ai.com/blog/colossal-inference)
### Grok-1 ### Grok-1
<p id="Grok-1" align="center"> <p id="Grok-1" align="center">
<img src="https://raw.githubusercontent.com/hpcaitech/public_assets/main/examples/images/grok-1-inference.jpg" width=600/> <img src="https://raw.githubusercontent.com/hpcaitech/public_assets/main/examples/images/grok-1-inference.jpg" width=600/>
@ -388,25 +400,6 @@ Colossal-AI 为您提供了一系列并行组件。我们的目标是让您的
- [SwiftInfer](https://github.com/hpcaitech/SwiftInfer): 开源解决方案打破了多轮对话的 LLM 长度限制，推理性能提高了46% - [SwiftInfer](https://github.com/hpcaitech/SwiftInfer): 开源解决方案打破了多轮对话的 LLM 长度限制，推理性能提高了46%
<p id="GPT-3-Inference" align="center">
<img src="https://raw.githubusercontent.com/hpcaitech/public_assets/main/colossalai/img/inference_GPT-3.jpg" width=800/>
</p>
- [Energon-AI](https://github.com/hpcaitech/EnergonAI) ：用相同的硬件推理加速50%
<p id="OPT-Serving" align="center">
<img src="https://raw.githubusercontent.com/hpcaitech/public_assets/main/colossalai/img/BLOOM%20serving.png" width=600/>
</p>
- [OPT推理服务](https://colossalai.org/docs/advanced_tutorials/opt_service): 体验1750亿参数OPT在线推理服务
<p id="BLOOM-Inference" align="center">
<img src="https://raw.githubusercontent.com/hpcaitech/public_assets/main/colossalai/img/BLOOM%20Inference.PNG" width=800/>
</p>
- [BLOOM](https://github.com/hpcaitech/EnergonAI/tree/main/examples/bloom): 降低1760亿参数BLOOM模型部署推理成本超10倍
<p align="right">(<a href="#top">返回顶端</a>)</p> <p align="right">(<a href="#top">返回顶端</a>)</p>
## 安装 ## 安装

Write Preview
Loading…
Cancel
Save