From 4647ec28c8450ee96f4709626617763712efd77e Mon Sep 17 00:00:00 2001 From: binmakeswell Date: Thu, 23 May 2024 17:44:06 +0800 Subject: [PATCH] [inference] release (#5747) * [inference] release * [inference] release * [inference] release * [inference] release * [inference] release * [inference] release * [inference] release --- README.md | 37 +++++++++++++++------------------- colossalai/inference/README.md | 9 ++++++++- docs/README-zh-Hans.md | 37 ++++++++++++++-------------------- 3 files changed, 39 insertions(+), 44 deletions(-) diff --git a/README.md b/README.md index 3157d74c91c5..e41b75c46987 100644 --- a/README.md +++ b/README.md @@ -25,6 +25,7 @@ ## 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] [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) @@ -75,11 +76,9 @@
  • Inference
  • @@ -377,6 +376,19 @@ Please visit our [documentation](https://www.colossalai.org/) and [examples](htt ## Inference +### Colossal-Inference +

    + +

    + +

    + +

    + + - 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

    @@ -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) [[ModelScope Grok-1 PyTorch model weights]](https://www.modelscope.cn/models/colossalai/grok-1-pytorch/summary) +### SwiftInfer

    - [SwiftInfer](https://github.com/hpcaitech/SwiftInfer): Inference performance improved by 46%, open source solution breaks the length limit of LLM for multi-round conversations -

    - -

    - -- [Energon-AI](https://github.com/hpcaitech/EnergonAI): 50% inference acceleration on the same hardware - -

    - -

    - -- [OPT Serving](https://colossalai.org/docs/advanced_tutorials/opt_service): Try 175-billion-parameter OPT online services - -

    - -

    - -- [BLOOM](https://github.com/hpcaitech/EnergonAI/tree/main/examples/bloom): Reduce hardware deployment costs of 176-billion-parameter BLOOM by more than 10 times. -

    (back to top)

    ## Installation diff --git a/colossalai/inference/README.md b/colossalai/inference/README.md index cd130a463e7e..cdb32a0f89a2 100644 --- a/colossalai/inference/README.md +++ b/colossalai/inference/README.md @@ -18,8 +18,15 @@ ## 📌 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) +

    + +

    + +

    + +

    ## 🕹 Usage diff --git a/docs/README-zh-Hans.md b/docs/README-zh-Hans.md index 41110612c307..5878abbaaa19 100644 --- a/docs/README-zh-Hans.md +++ b/docs/README-zh-Hans.md @@ -24,6 +24,7 @@ ## 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] [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) @@ -74,11 +75,9 @@
  • 推理
  • @@ -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 +### Colossal-Inference +

    + +

    + +

    + +

    + + - 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

    @@ -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% - -

    - -

    - -- [Energon-AI](https://github.com/hpcaitech/EnergonAI) ：用相同的硬件推理加速50% - -

    - -

    - -- [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 - -

    - -

    - -- [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 -

    (返回顶端)

    ## 安装