Add library name and pipeline tag

#66
by nielsr HF staff - opened
Files changed (1) hide show
  1. README.md +48 -29
README.md CHANGED
@@ -1,3 +1,10 @@
 
 
 
 
 
 
 
1
  <!-- markdownlint-disable first-line-h1 -->
2
  <!-- markdownlint-disable html -->
3
  <!-- markdownlint-disable no-duplicate-header -->
@@ -98,7 +105,8 @@ Throughout the entire training process, we did not experience any irrecoverable
98
 
99
  </div>
100
 
101
- **NOTE: The total size of DeepSeek-V3 models on HuggingFace is 685B, which includes 671B of the Main Model weights and 14B of the Multi-Token Prediction (MTP) Module weights.**
 
102
 
103
  To ensure optimal performance and flexibility, we have partnered with open-source communities and hardware vendors to provide multiple ways to run the model locally. For step-by-step guidance, check out Section 6: [How_to Run_Locally](#6-how-to-run-locally).
104
 
@@ -151,8 +159,9 @@ For developers looking to dive deeper, we recommend exploring [README_WEIGHTS.md
151
 
152
  </div>
153
 
154
- Note: Best results are shown in bold. Scores with a gap not exceeding 0.3 are considered to be at the same level. DeepSeek-V3 achieves the best performance on most benchmarks, especially on math and code tasks.
155
- For more evaluation details, please check our paper.
 
156
 
157
  #### Context Window
158
  <p align="center">
@@ -193,17 +202,16 @@ Evaluation results on the ``Needle In A Haystack`` (NIAH) tests. DeepSeek-V3 pe
193
  | | C-Eval (EM) | 78.6 | 79.5 | 86.1 | 61.5 | 76.7 | 76.0 | **86.5** |
194
  | | C-SimpleQA (Correct) | 48.5 | 54.1 | 48.4 | 50.4 | 51.3 | 59.3 | **64.8** |
195
 
196
- Note: All models are evaluated in a configuration that limits the output length to 8K. Benchmarks containing fewer than 1000 samples are tested multiple times using varying temperature settings to derive robust final results. DeepSeek-V3 stands as the best-performing open-source model, and also exhibits competitive performance against frontier closed-source models.
197
-
198
  </div>
199
 
 
 
 
200
 
201
  #### Open Ended Generation Evaluation
202
 
203
  <div align="center">
204
 
205
-
206
-
207
  | Model | Arena-Hard | AlpacaEval 2.0 |
208
  |-------|------------|----------------|
209
  | DeepSeek-V2.5-0905 | 76.2 | 50.5 |
@@ -213,9 +221,11 @@ Note: All models are evaluated in a configuration that limits the output length
213
  | Claude-Sonnet-3.5-1022 | 85.2 | 52.0 |
214
  | DeepSeek-V3 | **85.5** | **70.0** |
215
 
216
- Note: English open-ended conversation evaluations. For AlpacaEval 2.0, we use the length-controlled win rate as the metric.
217
  </div>
218
 
 
 
 
219
 
220
  ## 5. Chat Website & API Platform
221
  You can chat with DeepSeek-V3 on DeepSeek's official website: [chat.deepseek.com](https://chat.deepseek.com/sign_in)
@@ -227,10 +237,10 @@ We also provide OpenAI-Compatible API at DeepSeek Platform: [platform.deepseek.c
227
  DeepSeek-V3 can be deployed locally using the following hardware and open-source community software:
228
 
229
  1. **DeepSeek-Infer Demo**: We provide a simple and lightweight demo for FP8 and BF16 inference.
230
- 2. **SGLang**: Fully support the DeepSeek-V3 model in both BF16 and FP8 inference modes.
231
  3. **LMDeploy**: Enables efficient FP8 and BF16 inference for local and cloud deployment.
232
- 4. **TensorRT-LLM**: Currently supports BF16 inference and INT4/8 quantization, with FP8 support coming soon.
233
- 5. **vLLM**: Support DeekSeek-V3 model with FP8 and BF16 modes for tensor parallelism and pipeline parallelism.
234
  6. **AMD GPU**: Enables running the DeepSeek-V3 model on AMD GPUs via SGLang in both BF16 and FP8 modes.
235
  7. **Huawei Ascend NPU**: Supports running DeepSeek-V3 on Huawei Ascend devices.
236
 
@@ -243,10 +253,23 @@ cd inference
243
  python fp8_cast_bf16.py --input-fp8-hf-path /path/to/fp8_weights --output-bf16-hf-path /path/to/bf16_weights
244
  ```
245
 
246
- **NOTE: Huggingface's Transformers has not been directly supported yet.**
 
247
 
248
  ### 6.1 Inference with DeepSeek-Infer Demo (example only)
249
 
 
 
 
 
 
 
 
 
 
 
 
 
250
  #### Model Weights & Demo Code Preparation
251
 
252
  First, clone our DeepSeek-V3 GitHub repository:
@@ -255,18 +278,18 @@ First, clone our DeepSeek-V3 GitHub repository:
255
  git clone https://github.com/deepseek-ai/DeepSeek-V3.git
256
  ```
257
 
258
- Navigate to the `inference` folder and install dependencies listed in `requirements.txt`.
259
 
260
  ```shell
261
  cd DeepSeek-V3/inference
262
  pip install -r requirements.txt
263
  ```
264
 
265
- Download the model weights from HuggingFace, and put them into `/path/to/DeepSeek-V3` folder.
266
 
267
  #### Model Weights Conversion
268
 
269
- Convert HuggingFace model weights to a specific format:
270
 
271
  ```shell
272
  python convert.py --hf-ckpt-path /path/to/DeepSeek-V3 --save-path /path/to/DeepSeek-V3-Demo --n-experts 256 --model-parallel 16
@@ -277,21 +300,25 @@ python convert.py --hf-ckpt-path /path/to/DeepSeek-V3 --save-path /path/to/DeepS
277
  Then you can chat with DeepSeek-V3:
278
 
279
  ```shell
280
- torchrun --nnodes 2 --nproc-per-node 8 generate.py --node-rank $RANK --master-addr $ADDR --ckpt-path /path/to/DeepSeek-V3-Demo --config configs/config_671B.json --interactive --temperature 0.7 --max-new-tokens 200
281
  ```
282
 
283
  Or batch inference on a given file:
284
 
285
  ```shell
286
- torchrun --nnodes 2 --nproc-per-node 8 generate.py --node-rank $RANK --master-addr $ADDR --ckpt-path /path/to/DeepSeek-V3-Demo --config configs/config_671B.json --input-file $FILE
287
  ```
288
 
289
  ### 6.2 Inference with SGLang (recommended)
290
 
291
- [SGLang](https://github.com/sgl-project/sglang) currently supports MLA optimizations, FP8 (W8A8), FP8 KV Cache, and Torch Compile, delivering state-of-the-art latency and throughput performance among open-source frameworks.
292
 
293
  Notably, [SGLang v0.4.1](https://github.com/sgl-project/sglang/releases/tag/v0.4.1) fully supports running DeepSeek-V3 on both **NVIDIA and AMD GPUs**, making it a highly versatile and robust solution.
294
 
 
 
 
 
295
  Here are the launch instructions from the SGLang team: https://github.com/sgl-project/sglang/tree/main/benchmark/deepseek_v3
296
 
297
  ### 6.3 Inference with LMDeploy (recommended)
@@ -304,6 +331,7 @@ For comprehensive step-by-step instructions on running DeepSeek-V3 with LMDeploy
304
 
305
  [TensorRT-LLM](https://github.com/NVIDIA/TensorRT-LLM) now supports the DeepSeek-V3 model, offering precision options such as BF16 and INT4/INT8 weight-only. Support for FP8 is currently in progress and will be released soon. You can access the custom branch of TRTLLM specifically for DeepSeek-V3 support through the following link to experience the new features directly: https://github.com/NVIDIA/TensorRT-LLM/tree/deepseek/examples/deepseek_v3.
306
 
 
307
  ### 6.5 Inference with vLLM (recommended)
308
 
309
  [vLLM](https://github.com/vllm-project/vllm) v0.6.6 supports DeepSeek-V3 inference for FP8 and BF16 modes on both NVIDIA and AMD GPUs. Aside from standard techniques, vLLM offers _pipeline parallelism_ allowing you to run this model on multiple machines connected by networks. For detailed guidance, please refer to the [vLLM instructions](https://docs.vllm.ai/en/latest/serving/distributed_serving.html). Please feel free to follow [the enhancement plan](https://github.com/vllm-project/vllm/issues/11539) as well.
@@ -323,14 +351,5 @@ This code repository is licensed under [the MIT License](LICENSE-CODE). The use
323
  ```
324
  @misc{deepseekai2024deepseekv3technicalreport,
325
  title={DeepSeek-V3 Technical Report},
326
- author={DeepSeek-AI and Aixin Liu and Bei Feng and Bing Xue and Bingxuan Wang and Bochao Wu and Chengda Lu and Chenggang Zhao and Chengqi Deng and Chenyu Zhang and Chong Ruan and Damai Dai and Daya Guo and Dejian Yang and Deli Chen and Dongjie Ji and Erhang Li and Fangyun Lin and Fucong Dai and Fuli Luo and Guangbo Hao and Guanting Chen and Guowei Li and H. Zhang and Han Bao and Hanwei Xu and Haocheng Wang and Haowei Zhang and Honghui Ding and Huajian Xin and Huazuo Gao and Hui Li and Hui Qu and J. L. Cai and Jian Liang and Jianzhong Guo and Jiaqi Ni and Jiashi Li and Jiawei Wang and Jin Chen and Jingchang Chen and Jingyang Yuan and Junjie Qiu and Junlong Li and Junxiao Song and Kai Dong and Kai Hu and Kaige Gao and Kang Guan and Kexin Huang and Kuai Yu and Lean Wang and Lecong Zhang and Lei Xu and Leyi Xia and Liang Zhao and Litong Wang and Liyue Zhang and Meng Li and Miaojun Wang and Mingchuan Zhang and Minghua Zhang and Minghui Tang and Mingming Li and Ning Tian and Panpan Huang and Peiyi Wang and Peng Zhang and Qiancheng Wang and Qihao Zhu and Qinyu Chen and Qiushi Du and R. J. Chen and R. L. Jin and Ruiqi Ge and Ruisong Zhang and Ruizhe Pan and Runji Wang and Runxin Xu and Ruoyu Zhang and Ruyi Chen and S. S. Li and Shanghao Lu and Shangyan Zhou and Shanhuang Chen and Shaoqing Wu and Shengfeng Ye and Shengfeng Ye and Shirong Ma and Shiyu Wang and Shuang Zhou and Shuiping Yu and Shunfeng Zhou and Shuting Pan and T. Wang and Tao Yun and Tian Pei and Tianyu Sun and W. L. Xiao and Wangding Zeng and Wanjia Zhao and Wei An and Wen Liu and Wenfeng Liang and Wenjun Gao and Wenqin Yu and Wentao Zhang and X. Q. Li and Xiangyue Jin and Xianzu Wang and Xiao Bi and Xiaodong Liu and Xiaohan Wang and Xiaojin Shen and Xiaokang Chen and Xiaokang Zhang and Xiaosha Chen and Xiaotao Nie and Xiaowen Sun and Xiaoxiang Wang and Xin Cheng and Xin Liu and Xin Xie and Xingchao Liu and Xingkai Yu and Xinnan Song and Xinxia Shan and Xinyi Zhou and Xinyu Yang and Xinyuan Li and Xuecheng Su and Xuheng Lin and Y. K. Li and Y. Q. Wang and Y. X. Wei and Y. X. Zhu and Yang Zhang and Yanhong Xu and Yanhong Xu and Yanping Huang and Yao Li and Yao Zhao and Yaofeng Sun and Yaohui Li and Yaohui Wang and Yi Yu and Yi Zheng and Yichao Zhang and Yifan Shi and Yiliang Xiong and Ying He and Ying Tang and Yishi Piao and Yisong Wang and Yixuan Tan and Yiyang Ma and Yiyuan Liu and Yongqiang Guo and Yu Wu and Yuan Ou and Yuchen Zhu and Yuduan Wang and Yue Gong and Yuheng Zou and Yujia He and Yukun Zha and Yunfan Xiong and Yunxian Ma and Yuting Yan and Yuxiang Luo and Yuxiang You and Yuxuan Liu and Yuyang Zhou and Z. F. Wu and Z. Z. Ren and Zehui Ren and Zhangli Sha and Zhe Fu and Zhean Xu and Zhen Huang and Zhen Zhang and Zhenda Xie and Zhengyan Zhang and Zhewen Hao and Zhibin Gou and Zhicheng Ma and Zhigang Yan and Zhihong Shao and Zhipeng Xu and Zhiyu Wu and Zhongyu Zhang and Zhuoshu Li and Zihui Gu and Zijia Zhu and Zijun Liu and Zilin Li and Ziwei Xie and Ziyang Song and Ziyi Gao and Zizheng Pan},
327
- year={2024},
328
- eprint={2412.19437},
329
- archivePrefix={arXiv},
330
- primaryClass={cs.CL},
331
- url={https://arxiv.org/abs/2412.19437},
332
- }
333
- ```
334
-
335
- ## 9. Contact
336
- If you have any questions, please raise an issue or contact us at [service@deepseek.com](service@deepseek.com).
 
1
+ ---
2
+ license: other
3
+ library_name: transformers
4
+ pipeline_tag: text-generation
5
+ ---
6
+
7
+ ```markdown
8
  <!-- markdownlint-disable first-line-h1 -->
9
  <!-- markdownlint-disable html -->
10
  <!-- markdownlint-disable no-duplicate-header -->
 
105
 
106
  </div>
107
 
108
+ > [!NOTE]
109
+ > The total size of DeepSeek-V3 models on HuggingFace is 685B, which includes 671B of the Main Model weights and 14B of the Multi-Token Prediction (MTP) Module weights.
110
 
111
  To ensure optimal performance and flexibility, we have partnered with open-source communities and hardware vendors to provide multiple ways to run the model locally. For step-by-step guidance, check out Section 6: [How_to Run_Locally](#6-how-to-run-locally).
112
 
 
159
 
160
  </div>
161
 
162
+ > [!NOTE]
163
+ > Best results are shown in bold. Scores with a gap not exceeding 0.3 are considered to be at the same level. DeepSeek-V3 achieves the best performance on most benchmarks, especially on math and code tasks.
164
+ > For more evaluation details, please check our paper.
165
 
166
  #### Context Window
167
  <p align="center">
 
202
  | | C-Eval (EM) | 78.6 | 79.5 | 86.1 | 61.5 | 76.7 | 76.0 | **86.5** |
203
  | | C-SimpleQA (Correct) | 48.5 | 54.1 | 48.4 | 50.4 | 51.3 | 59.3 | **64.8** |
204
 
 
 
205
  </div>
206
 
207
+ > [!NOTE]
208
+ > All models are evaluated in a configuration that limits the output length to 8K. Benchmarks containing fewer than 1000 samples are tested multiple times using varying temperature settings to derive robust final results. DeepSeek-V3 stands as the best-performing open-source model, and also exhibits competitive performance against frontier closed-source models.
209
+
210
 
211
  #### Open Ended Generation Evaluation
212
 
213
  <div align="center">
214
 
 
 
215
  | Model | Arena-Hard | AlpacaEval 2.0 |
216
  |-------|------------|----------------|
217
  | DeepSeek-V2.5-0905 | 76.2 | 50.5 |
 
221
  | Claude-Sonnet-3.5-1022 | 85.2 | 52.0 |
222
  | DeepSeek-V3 | **85.5** | **70.0** |
223
 
 
224
  </div>
225
 
226
+ > [!NOTE]
227
+ > English open-ended conversation evaluations. For AlpacaEval 2.0, we use the length-controlled win rate as the metric.
228
+
229
 
230
  ## 5. Chat Website & API Platform
231
  You can chat with DeepSeek-V3 on DeepSeek's official website: [chat.deepseek.com](https://chat.deepseek.com/sign_in)
 
237
  DeepSeek-V3 can be deployed locally using the following hardware and open-source community software:
238
 
239
  1. **DeepSeek-Infer Demo**: We provide a simple and lightweight demo for FP8 and BF16 inference.
240
+ 2. **SGLang**: Fully support the DeepSeek-V3 model in both BF16 and FP8 inference modes, with Multi-Token Prediction [coming soon](https://github.com/sgl-project/sglang/issues/2591).
241
  3. **LMDeploy**: Enables efficient FP8 and BF16 inference for local and cloud deployment.
242
+ 4. **TensorRT-LLM**: Currently supports BF16 inference and INT4/INT8 weight-only. Support for FP8 is currently in progress and will be released soon.
243
+ 5. **vLLM**: Support DeepSeek-V3 model with FP8 and BF16 modes for tensor parallelism and pipeline parallelism.
244
  6. **AMD GPU**: Enables running the DeepSeek-V3 model on AMD GPUs via SGLang in both BF16 and FP8 modes.
245
  7. **Huawei Ascend NPU**: Supports running DeepSeek-V3 on Huawei Ascend devices.
246
 
 
253
  python fp8_cast_bf16.py --input-fp8-hf-path /path/to/fp8_weights --output-bf16-hf-path /path/to/bf16_weights
254
  ```
255
 
256
+ > [!NOTE]
257
+ > Hugging Face's Transformers has not been directly supported yet.
258
 
259
  ### 6.1 Inference with DeepSeek-Infer Demo (example only)
260
 
261
+ #### System Requirements
262
+
263
+ > [!NOTE]
264
+ > Linux with Python 3.10 only. Mac and Windows are not supported.
265
+
266
+ Dependencies:
267
+ ```pip-requirements
268
+ torch==2.4.1
269
+ triton==3.0.0
270
+ transformers==4.46.3
271
+ safetensors==0.4.5
272
+ ```
273
  #### Model Weights & Demo Code Preparation
274
 
275
  First, clone our DeepSeek-V3 GitHub repository:
 
278
  git clone https://github.com/deepseek-ai/DeepSeek-V3.git
279
  ```
280
 
281
+ Navigate to the `inference` folder and install dependencies listed in `requirements.txt`. Easiest way is to use a package manager like `conda` or `uv` to create a new virtual environment and install the dependencies.
282
 
283
  ```shell
284
  cd DeepSeek-V3/inference
285
  pip install -r requirements.txt
286
  ```
287
 
288
+ Download the model weights from Hugging Face, and put them into `/path/to/DeepSeek-V3` folder.
289
 
290
  #### Model Weights Conversion
291
 
292
+ Convert Hugging Face model weights to a specific format:
293
 
294
  ```shell
295
  python convert.py --hf-ckpt-path /path/to/DeepSeek-V3 --save-path /path/to/DeepSeek-V3-Demo --n-experts 256 --model-parallel 16
 
300
  Then you can chat with DeepSeek-V3:
301
 
302
  ```shell
303
+ torchrun --nnodes 2 --nproc-per-node 8 --node-rank $RANK --master-addr $ADDR generate.py --ckpt-path /path/to/DeepSeek-V3-Demo --config configs/config_671B.json --interactive --temperature 0.7 --max-new-tokens 200
304
  ```
305
 
306
  Or batch inference on a given file:
307
 
308
  ```shell
309
+ torchrun --nnodes 2 --nproc-per-node 8 --node-rank $RANK --master-addr $ADDR generate.py --ckpt-path /path/to/DeepSeek-V3-Demo --config configs/config_671B.json --input-file $FILE
310
  ```
311
 
312
  ### 6.2 Inference with SGLang (recommended)
313
 
314
+ [SGLang](https://github.com/sgl-project/sglang) currently supports [MLA optimizations](https://lmsys.org/blog/2024-09-04-sglang-v0-3/#deepseek-multi-head-latent-attention-mla-throughput-optimizations), [DP Attention](https://lmsys.org/blog/2024-12-04-sglang-v0-4/#data-parallelism-attention-for-deepseek-models), FP8 (W8A8), FP8 KV Cache, and Torch Compile, delivering state-of-the-art latency and throughput performance among open-source frameworks.
315
 
316
  Notably, [SGLang v0.4.1](https://github.com/sgl-project/sglang/releases/tag/v0.4.1) fully supports running DeepSeek-V3 on both **NVIDIA and AMD GPUs**, making it a highly versatile and robust solution.
317
 
318
+ SGLang also supports [multi-node tensor parallelism](https://github.com/sgl-project/sglang/tree/main/benchmark/deepseek_v3#example-serving-with-2-h208), enabling you to run this model on multiple network-connected machines.
319
+
320
+ Multi-Token Prediction (MTP) is in development, and progress can be tracked in the [optimization plan](https://github.com/sgl-project/sglang/issues/2591).
321
+
322
  Here are the launch instructions from the SGLang team: https://github.com/sgl-project/sglang/tree/main/benchmark/deepseek_v3
323
 
324
  ### 6.3 Inference with LMDeploy (recommended)
 
331
 
332
  [TensorRT-LLM](https://github.com/NVIDIA/TensorRT-LLM) now supports the DeepSeek-V3 model, offering precision options such as BF16 and INT4/INT8 weight-only. Support for FP8 is currently in progress and will be released soon. You can access the custom branch of TRTLLM specifically for DeepSeek-V3 support through the following link to experience the new features directly: https://github.com/NVIDIA/TensorRT-LLM/tree/deepseek/examples/deepseek_v3.
333
 
334
+
335
  ### 6.5 Inference with vLLM (recommended)
336
 
337
  [vLLM](https://github.com/vllm-project/vllm) v0.6.6 supports DeepSeek-V3 inference for FP8 and BF16 modes on both NVIDIA and AMD GPUs. Aside from standard techniques, vLLM offers _pipeline parallelism_ allowing you to run this model on multiple machines connected by networks. For detailed guidance, please refer to the [vLLM instructions](https://docs.vllm.ai/en/latest/serving/distributed_serving.html). Please feel free to follow [the enhancement plan](https://github.com/vllm-project/vllm/issues/11539) as well.
 
351
  ```
352
  @misc{deepseekai2024deepseekv3technicalreport,
353
  title={DeepSeek-V3 Technical Report},
354
+ author={DeepSeek-AI and Aixin Liu and Bei Feng and Bing Xue and Bingxuan Wang and Bochao Wu and Chengda Lu and Chenggang Zhao and Chengqi Deng and Chenyu Zhang and Chong Ruan and Damai Dai and Daya Guo and Dejian Yang and Deli Chen and Dongjie Ji and Erhang Li and Fangyun Lin and Fucong Dai and Fuli Luo and Guangbo Hao and Guanting Chen and Guowei Li and H. Zhang and Han Bao and Hanwei Xu and Haocheng Wang and Haowei Zhang and Honghui Ding and Huajian Xin and Huazuo Gao and Hui Li and Hui Qu and J. L. Cai and Jian Liang and Jianzhong Guo and Jiaqi Ni and Jiashi Li and Jiawei Wang and Jin Chen and Jingchang Chen and Jingyang Yuan and Junjie Qiu and Junlong Li and Junxiao Song and Kai Dong and Kai Hu and Kaige Gao and Kang Guan and Kexin Huang and Kuai Yu and Lean Wang and Lecong Zhang and Lei Xu and Leyi Xia and Liang Zhao and Litong Wang and Liyue Zhang and Meng Li and Miaojun Wang and Mingchuan Zhang and Minghua Zhang and Minghui Tang and Mingming Li and Ning Tian and Panpan Huang and Peiyi Wang and Peng Zhang and Qiancheng Wang and Qihao Zhu and Qinyu Chen and Qiushi Du and R. J. Chen and R. L. Jin and Ruiqi Ge and Ruisong Zhang and Ruizhe Pan and Runji Wang and Runxin Xu and Ruoyu Zhang and Ruyi Chen and S. S. Li and Shanghao Lu and Shangyan Zhou and Shanhuang Chen and Shaoqing Wu and Shengfeng Ye and Shengfeng Ye and Shirong Ma and Shiyu Wang and Shuang Zhou and Shuiping Yu and Shunfeng Zhou and Shuting Pan and T. Wang and Tao Yun and Tian Pei and Tianyu Sun and W. L. Xiao and Wangding Zeng and Wanjia Zhao and Wei An and Wen Liu and Wenfeng Liang and Wenjun Gao and Wenqin Yu and Wentao Zhang and X. Q. Li and Xiangyue Jin and Xianzu Wang and Xiao Bi and Xiaodong Liu and Xiaohan Wang and Xiaojin Shen and Xiaokang Chen and Xiaokang Zhang and Xiaosha Chen and Xiaotao Nie and Xiaowen Sun and Xiaoxiang Wang and Xin Cheng and Xin Liu and Xin Xie and Xingchao Liu and Xingkai Yu and Xinnan Song and Xinxia Shan and Xinyi Zhou and
355
+ ```