Upload folder using huggingface_hub
Browse filesThis view is limited to 50 files because it contains too many changes.
See raw diff
- .gitattributes +52 -0
- LICENSE +34 -0
- OPEN SOURCE SOFTWARE NOTICE +635 -0
- README.md +85 -5
- README_EN.md +85 -0
- config.json +45 -0
- configuration_pangu_moe.py +96 -0
- doc/omniinfer_for_openpangu_r_72b_2512.md +118 -0
- doc/omniinfer_for_openpangu_r_72b_2512_EN.md +119 -0
- examples/start_serving_openpangu_r_72b_2512.sh +57 -0
- generation_config.json +11 -0
- model-00001-of-000051.safetensors +3 -0
- model-00002-of-000051.safetensors +3 -0
- model-00003-of-000051.safetensors +3 -0
- model-00004-of-000051.safetensors +3 -0
- model-00005-of-000051.safetensors +3 -0
- model-00006-of-000051.safetensors +3 -0
- model-00007-of-000051.safetensors +3 -0
- model-00008-of-000051.safetensors +3 -0
- model-00009-of-000051.safetensors +3 -0
- model-00010-of-000051.safetensors +3 -0
- model-00011-of-000051.safetensors +3 -0
- model-00012-of-000051.safetensors +3 -0
- model-00013-of-000051.safetensors +3 -0
- model-00014-of-000051.safetensors +3 -0
- model-00015-of-000051.safetensors +3 -0
- model-00016-of-000051.safetensors +3 -0
- model-00017-of-000051.safetensors +3 -0
- model-00018-of-000051.safetensors +3 -0
- model-00019-of-000051.safetensors +3 -0
- model-00020-of-000051.safetensors +3 -0
- model-00021-of-000051.safetensors +3 -0
- model-00022-of-000051.safetensors +3 -0
- model-00023-of-000051.safetensors +3 -0
- model-00024-of-000051.safetensors +3 -0
- model-00025-of-000051.safetensors +3 -0
- model-00026-of-000051.safetensors +3 -0
- model-00027-of-000051.safetensors +3 -0
- model-00028-of-000051.safetensors +3 -0
- model-00029-of-000051.safetensors +3 -0
- model-00030-of-000051.safetensors +3 -0
- model-00031-of-000051.safetensors +3 -0
- model-00032-of-000051.safetensors +3 -0
- model-00033-of-000051.safetensors +3 -0
- model-00034-of-000051.safetensors +3 -0
- model-00035-of-000051.safetensors +3 -0
- model-00036-of-000051.safetensors +3 -0
- model-00037-of-000051.safetensors +3 -0
- model-00038-of-000051.safetensors +3 -0
- model-00039-of-000051.safetensors +3 -0
.gitattributes
CHANGED
|
@@ -33,3 +33,55 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
|
| 33 |
*.zip filter=lfs diff=lfs merge=lfs -text
|
| 34 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
| 35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 33 |
*.zip filter=lfs diff=lfs merge=lfs -text
|
| 34 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
| 35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
| 36 |
+
model-00016-of-000051.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 37 |
+
model-00051-of-000051.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 38 |
+
model-00045-of-000051.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 39 |
+
model-00013-of-000051.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 40 |
+
model-00010-of-000051.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 41 |
+
model-00015-of-000051.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 42 |
+
model-00041-of-000051.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 43 |
+
model-00047-of-000051.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 44 |
+
model-00021-of-000051.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 45 |
+
tokenizer.model filter=lfs diff=lfs merge=lfs -text
|
| 46 |
+
model-00046-of-000051.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 47 |
+
model-00032-of-000051.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 48 |
+
model-00042-of-000051.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 49 |
+
model-00025-of-000051.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 50 |
+
model-00024-of-000051.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 51 |
+
model-00004-of-000051.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 52 |
+
model-00037-of-000051.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 53 |
+
model-00049-of-000051.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 54 |
+
model-00023-of-000051.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 55 |
+
model-00033-of-000051.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 56 |
+
model-00040-of-000051.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 57 |
+
model-00027-of-000051.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 58 |
+
model-00009-of-000051.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 59 |
+
model-00031-of-000051.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 60 |
+
model-00036-of-000051.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 61 |
+
model-00018-of-000051.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 62 |
+
model-00019-of-000051.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 63 |
+
model-00034-of-000051.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 64 |
+
model-00022-of-000051.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 65 |
+
model-00026-of-000051.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 66 |
+
model-00017-of-000051.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 67 |
+
model-00020-of-000051.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 68 |
+
model-00001-of-000051.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 69 |
+
model-00039-of-000051.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 70 |
+
model-00038-of-000051.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 71 |
+
model-00029-of-000051.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 72 |
+
model-00012-of-000051.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 73 |
+
model-00035-of-000051.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 74 |
+
model-00002-of-000051.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 75 |
+
model-00006-of-000051.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 76 |
+
model-00044-of-000051.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 77 |
+
model-00011-of-000051.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 78 |
+
model-00003-of-000051.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 79 |
+
model-00005-of-000051.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 80 |
+
model-00014-of-000051.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 81 |
+
model-00007-of-000051.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 82 |
+
model-00043-of-000051.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 83 |
+
model-00048-of-000051.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 84 |
+
model-00008-of-000051.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 85 |
+
model-00030-of-000051.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 86 |
+
model-00050-of-000051.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 87 |
+
model-00028-of-000051.safetensors filter=lfs diff=lfs merge=lfs -text
|
LICENSE
ADDED
|
@@ -0,0 +1,34 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
OPENPANGU MODEL LICENSE AGREEMENT VERSION 1.0
|
| 2 |
+
|
| 3 |
+
This OPENPANGU MODEL LICENSE AGREEMENT VERSION 1.0 (the "Agreement") is a legal agreement between You and Huawei Technologies Co., Ltd. ("Huawei", "We" or "Us"), and it governs Your reproducing, use, modification, and distribution of openPangu as made available by Huawei under this Agreement.
|
| 4 |
+
|
| 5 |
+
By using, reproducing, modifying, distributing, performing or displaying any portion or element of openPangu, or otherwise accepting the terms of this Agreement, You agree to be bound by this Agreement.
|
| 6 |
+
|
| 7 |
+
1. Definitions.
|
| 8 |
+
1.1. “openPangu” or “Model” means openPangu large language models and software, including trained model weights, parameters (including optimizer states), accompanying source code and scripts released under this Agreement.
|
| 9 |
+
1.2. “Derivative Model” means all (1) modifications to the Model, (2) works based on the Model, and (3) any other derivative works of the Model. For clarity, information or content results from operating or otherwise using the Model is not a Derivative Model.
|
| 10 |
+
1.3. “You” or “Your” means an individual or Legal Entity exercising permissions granted by this Agreement and/or using the Model for any purpose.
|
| 11 |
+
1.4. “Third Party” or “Third Parties” means individuals or legal entities that are not under common control with Us or You.
|
| 12 |
+
|
| 13 |
+
2. License Grant. Subject to Your full compliance with the terms and conditions of this Agreement, We hereby grant to You a perpetual, worldwide, non-exclusive, non-transferable, no-charge, royalty-free license (except as stated in Section 3) to use, reproduce, modify, and distribute the Model.
|
| 14 |
+
|
| 15 |
+
3. Conditions for License Grant. You represent and warrant that You will not, access, download, install, run, deploy, integrate, modify, or otherwise use the Model, directly or indirectly, within the European Union.
|
| 16 |
+
|
| 17 |
+
|
| 18 |
+
4. Redistribution.
|
| 19 |
+
4.1. If You distribute the Model or Derivative Model, You shall retain in Your distribution (1) a copy of this agreement, and (2) all copyright notices and other notices of origin included in the Model that are applicable to Your distribution.
|
| 20 |
+
4.2. Further, if You distribute or make available to Third Parties a product or service (including another AI model) based on the Model, You are required to (1) display the acknowledgement “Powered by openPangu” and (2) include a trademark notice “openPangu is a trademark of Huawei Technologies Co., Ltd.” on related webpages, user manuals, product documentations or other advertising materials mentioning features of the Model.
|
| 21 |
+
4.3. You may add Your own copyright statement to Your modifications and may provide additional or different license terms and conditions for use, reproduction, or distribution of Your modifications, or for Derivative Model made by You as a whole, provided Your use, reproduction, and distribution of the Model otherwise complies with the terms and conditions of this Agreement.
|
| 22 |
+
|
| 23 |
+
5. Ownership. We do not claim ownership to any information or content generated using the Model or Derivative Model that are made by You. You are solely responsible for evaluating the accuracy and appropriateness of such information or content for Your use case.
|
| 24 |
+
|
| 25 |
+
6. Trademarks. This License does not grant permission to use the trade names, trademarks, service marks, or product names of Huawei, except as required for complying with Section 4.2.
|
| 26 |
+
|
| 27 |
+
7. Indemnity. You will indemnify and hold harmless Huawei from and against any claim by any third party arising out of or related to Your use or distribution of the Model or Derivative Model made by You (e.g. a violation against Section 3). For avoidance of doubt, “third party” in this clause include supervisory authorities.
|
| 28 |
+
|
| 29 |
+
8. THE MODEL IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE, TITLE, NONINFRINGEMENT, ACCURACY, OR THE ABSENCE OF LATENT OR OTHER DEFECTS OR ERRORS, WHETHER OR NOT DISCOVERABLE, ALL TO THE GREATEST EXTENT PERMISSIBLE UNDER APPLICABLE LAW.
|
| 30 |
+
|
| 31 |
+
9. IN NO EVENT SHALL WE BE LIABLE TO YOU FOR ANY DAMAGES, INCLUDING, BUT NOT LIMITED TO ANY DIRECT, OR INDIRECT, SPECIAL OR CONSEQUENTIAL DAMAGES ARISING FROM YOUR USE OR INABILITY TO USE THE MODEL, IN WHOLE OR IN PART, NO MATTER HOW IT’S CAUSED OR THE LEGAL THEORY IT IS BASED ON, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGES.
|
| 32 |
+
|
| 33 |
+
|
| 34 |
+
END OF THE TERMS AND CONDITIONS
|
OPEN SOURCE SOFTWARE NOTICE
ADDED
|
@@ -0,0 +1,635 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
OPEN SOURCE SOFTWARE NOTICE
|
| 2 |
+
|
| 3 |
+
Please note we provide an open source software notice along with this product and/or this product firmware (in the following just “this product”). The open source software licenses are granted by the respective right holders. And the open source licenses prevail all other license information with regard to the respective open source software contained in the product, including but not limited to End User Software Licensing Agreement. This notice is provided on behalf of Huawei Technologies Co. Ltd. and any of its local subsidiaries which may have provided this product to you in your local country.
|
| 4 |
+
|
| 5 |
+
Warranty Disclaimer
|
| 6 |
+
THE OPEN SOURCE SOFTWARE IN THIS PRODUCT IS DISTRIBUTED IN THE HOPE THAT IT WILL BE USEFUL, BUT WITHOUT ANY WARRANTY, WITHOUT EVEN THE IMPLIED WARRANTY OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. SEE THE APPLICABLE LICENSES FOR MORE DETAILS.
|
| 7 |
+
|
| 8 |
+
Copyright Notice and License Texts
|
| 9 |
+
|
| 10 |
+
Software: transformers 4.48.2
|
| 11 |
+
Copyright notice:
|
| 12 |
+
Copyright 2022 EleutherAI and the HuggingFace Inc. team. All rights reserved.
|
| 13 |
+
|
| 14 |
+
License Text:
|
| 15 |
+
----------------------------------------
|
| 16 |
+
|
| 17 |
+
Apache License
|
| 18 |
+
Version 2.0, January 2004
|
| 19 |
+
http://www.apache.org/licenses/
|
| 20 |
+
|
| 21 |
+
TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
|
| 22 |
+
|
| 23 |
+
1. Definitions.
|
| 24 |
+
|
| 25 |
+
"License" shall mean the terms and conditions for use, reproduction,
|
| 26 |
+
and distribution as defined by Sections 1 through 9 of this document.
|
| 27 |
+
|
| 28 |
+
"Licensor" shall mean the copyright owner or entity authorized by
|
| 29 |
+
the copyright owner that is granting the License.
|
| 30 |
+
|
| 31 |
+
"Legal Entity" shall mean the union of the acting entity and all
|
| 32 |
+
other entities that control, are controlled by, or are under common
|
| 33 |
+
control with that entity. For the purposes of this definition,
|
| 34 |
+
"control" means (i) the power, direct or indirect, to cause the
|
| 35 |
+
direction or management of such entity, whether by contract or
|
| 36 |
+
otherwise, or (ii) ownership of fifty percent (50%) or more of the
|
| 37 |
+
outstanding shares, or (iii) beneficial ownership of such entity.
|
| 38 |
+
|
| 39 |
+
"You" (or "Your") shall mean an individual or Legal Entity
|
| 40 |
+
exercising permissions granted by this License.
|
| 41 |
+
|
| 42 |
+
"Source" form shall mean the preferred form for making modifications,
|
| 43 |
+
including but not limited to software source code, documentation
|
| 44 |
+
source, and configuration files.
|
| 45 |
+
|
| 46 |
+
"Object" form shall mean any form resulting from mechanical
|
| 47 |
+
transformation or translation of a Source form, including but
|
| 48 |
+
not limited to compiled object code, generated documentation,
|
| 49 |
+
and conversions to other media types.
|
| 50 |
+
|
| 51 |
+
"Work" shall mean the work of authorship, whether in Source or
|
| 52 |
+
Object form, made available under the License, as indicated by a
|
| 53 |
+
copyright notice that is included in or attached to the work
|
| 54 |
+
(an example is provided in the Appendix below).
|
| 55 |
+
|
| 56 |
+
"Derivative Works" shall mean any work, whether in Source or Object
|
| 57 |
+
form, that is based on (or derived from) the Work and for which the
|
| 58 |
+
editorial revisions, annotations, elaborations, or other modifications
|
| 59 |
+
represent, as a whole, an original work of authorship. For the purposes
|
| 60 |
+
of this License, Derivative Works shall not include works that remain
|
| 61 |
+
separable from, or merely link (or bind by name) to the interfaces of,
|
| 62 |
+
the Work and Derivative Works thereof.
|
| 63 |
+
|
| 64 |
+
"Contribution" shall mean any work of authorship, including
|
| 65 |
+
the original version of the Work and any modifications or additions
|
| 66 |
+
to that Work or Derivative Works thereof, that is intentionally
|
| 67 |
+
submitted to Licensor for inclusion in the Work by the copyright owner
|
| 68 |
+
or by an individual or Legal Entity authorized to submit on behalf of
|
| 69 |
+
the copyright owner. For the purposes of this definition, "submitted"
|
| 70 |
+
means any form of electronic, verbal, or written communication sent
|
| 71 |
+
to the Licensor or its representatives, including but not limited to
|
| 72 |
+
communication on electronic mailing lists, source code control systems,
|
| 73 |
+
and issue tracking systems that are managed by, or on behalf of, the
|
| 74 |
+
Licensor for the purpose of discussing and improving the Work, but
|
| 75 |
+
excluding communication that is conspicuously marked or otherwise
|
| 76 |
+
designated in writing by the copyright owner as "Not a Contribution."
|
| 77 |
+
|
| 78 |
+
"Contributor" shall mean Licensor and any individual or Legal Entity
|
| 79 |
+
on behalf of whom a Contribution has been received by Licensor and
|
| 80 |
+
subsequently incorporated within the Work.
|
| 81 |
+
|
| 82 |
+
2. Grant of Copyright License. Subject to the terms and conditions of
|
| 83 |
+
this License, each Contributor hereby grants to You a perpetual,
|
| 84 |
+
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
|
| 85 |
+
copyright license to reproduce, prepare Derivative Works of,
|
| 86 |
+
publicly display, publicly perform, sublicense, and distribute the
|
| 87 |
+
Work and such Derivative Works in Source or Object form.
|
| 88 |
+
|
| 89 |
+
3. Grant of Patent License. Subject to the terms and conditions of
|
| 90 |
+
this License, each Contributor hereby grants to You a perpetual,
|
| 91 |
+
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
|
| 92 |
+
(except as stated in this section) patent license to make, have made,
|
| 93 |
+
use, offer to sell, sell, import, and otherwise transfer the Work,
|
| 94 |
+
where such license applies only to those patent claims licensable
|
| 95 |
+
by such Contributor that are necessarily infringed by their
|
| 96 |
+
Contribution(s) alone or by combination of their Contribution(s)
|
| 97 |
+
with the Work to which such Contribution(s) was submitted. If You
|
| 98 |
+
institute patent litigation against any entity (including a
|
| 99 |
+
cross-claim or counterclaim in a lawsuit) alleging that the Work
|
| 100 |
+
or a Contribution incorporated within the Work constitutes direct
|
| 101 |
+
or contributory patent infringement, then any patent licenses
|
| 102 |
+
granted to You under this License for that Work shall terminate
|
| 103 |
+
as of the date such litigation is filed.
|
| 104 |
+
|
| 105 |
+
4. Redistribution. You may reproduce and distribute copies of the
|
| 106 |
+
Work or Derivative Works thereof in any medium, with or without
|
| 107 |
+
modifications, and in Source or Object form, provided that You
|
| 108 |
+
meet the following conditions:
|
| 109 |
+
|
| 110 |
+
(a) You must give any other recipients of the Work or
|
| 111 |
+
Derivative Works a copy of this License; and
|
| 112 |
+
|
| 113 |
+
(b) You must cause any modified files to carry prominent notices
|
| 114 |
+
stating that You changed the files; and
|
| 115 |
+
|
| 116 |
+
(c) You must retain, in the Source form of any Derivative Works
|
| 117 |
+
that You distribute, all copyright, patent, trademark, and
|
| 118 |
+
attribution notices from the Source form of the Work,
|
| 119 |
+
excluding those notices that do not pertain to any part of
|
| 120 |
+
the Derivative Works; and
|
| 121 |
+
|
| 122 |
+
(d) If the Work includes a "NOTICE" text file as part of its
|
| 123 |
+
distribution, then any Derivative Works that You distribute must
|
| 124 |
+
include a readable copy of the attribution notices contained
|
| 125 |
+
within such NOTICE file, excluding those notices that do not
|
| 126 |
+
pertain to any part of the Derivative Works, in at least one
|
| 127 |
+
of the following places: within a NOTICE text file distributed
|
| 128 |
+
as part of the Derivative Works; within the Source form or
|
| 129 |
+
documentation, if provided along with the Derivative Works; or,
|
| 130 |
+
within a display generated by the Derivative Works, if and
|
| 131 |
+
wherever such third-party notices normally appear. The contents
|
| 132 |
+
of the NOTICE file are for informational purposes only and
|
| 133 |
+
do not modify the License. You may add Your own attribution
|
| 134 |
+
notices within Derivative Works that You distribute, alongside
|
| 135 |
+
or as an addendum to the NOTICE text from the Work, provided
|
| 136 |
+
that such additional attribution notices cannot be construed
|
| 137 |
+
as modifying the License.
|
| 138 |
+
|
| 139 |
+
You may add Your own copyright statement to Your modifications and
|
| 140 |
+
may provide additional or different license terms and conditions
|
| 141 |
+
for use, reproduction, or distribution of Your modifications, or
|
| 142 |
+
for any such Derivative Works as a whole, provided Your use,
|
| 143 |
+
reproduction, and distribution of the Work otherwise complies with
|
| 144 |
+
the conditions stated in this License.
|
| 145 |
+
|
| 146 |
+
5. Submission of Contributions. Unless You explicitly state otherwise,
|
| 147 |
+
any Contribution intentionally submitted for inclusion in the Work
|
| 148 |
+
by You to the Licensor shall be under the terms and conditions of
|
| 149 |
+
this License, without any additional terms or conditions.
|
| 150 |
+
Notwithstanding the above, nothing herein shall supersede or modify
|
| 151 |
+
the terms of any separate license agreement you may have executed
|
| 152 |
+
with Licensor regarding such Contributions.
|
| 153 |
+
|
| 154 |
+
6. Trademarks. This License does not grant permission to use the trade
|
| 155 |
+
names, trademarks, service marks, or product names of the Licensor,
|
| 156 |
+
except as required for reasonable and customary use in describing the
|
| 157 |
+
origin of the Work and reproducing the content of the NOTICE file.
|
| 158 |
+
|
| 159 |
+
7. Disclaimer of Warranty. Unless required by applicable law or
|
| 160 |
+
agreed to in writing, Licensor provides the Work (and each
|
| 161 |
+
Contributor provides its Contributions) on an "AS IS" BASIS,
|
| 162 |
+
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
|
| 163 |
+
implied, including, without limitation, any warranties or conditions
|
| 164 |
+
of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
|
| 165 |
+
PARTICULAR PURPOSE. You are solely responsible for determining the
|
| 166 |
+
appropriateness of using or redistributing the Work and assume any
|
| 167 |
+
risks associated with Your exercise of permissions under this License.
|
| 168 |
+
|
| 169 |
+
8. Limitation of Liability. In no event and under no legal theory,
|
| 170 |
+
whether in tort (including negligence), contract, or otherwise,
|
| 171 |
+
unless required by applicable law (such as deliberate and grossly
|
| 172 |
+
negligent acts) or agreed to in writing, shall any Contributor be
|
| 173 |
+
liable to You for damages, including any direct, indirect, special,
|
| 174 |
+
incidental, or consequential damages of any character arising as a
|
| 175 |
+
result of this License or out of the use or inability to use the
|
| 176 |
+
Work (including but not limited to damages for loss of goodwill,
|
| 177 |
+
work stoppage, computer failure or malfunction, or any and all
|
| 178 |
+
other commercial damages or losses), even if such Contributor
|
| 179 |
+
has been advised of the possibility of such damages.
|
| 180 |
+
|
| 181 |
+
9. Accepting Warranty or Additional Liability. While redistributing
|
| 182 |
+
the Work or Derivative Works thereof, You may choose to offer,
|
| 183 |
+
and charge a fee for, acceptance of support, warranty, indemnity,
|
| 184 |
+
or other liability obligations and/or rights consistent with this
|
| 185 |
+
License. However, in accepting such obligations, You may act only
|
| 186 |
+
on Your own behalf and on Your sole responsibility, not on behalf
|
| 187 |
+
of any other Contributor, and only if You agree to indemnify,
|
| 188 |
+
defend, and hold each Contributor harmless for any liability
|
| 189 |
+
incurred by, or claims asserted against, such Contributor by reason
|
| 190 |
+
of your accepting any such warranty or additional liability.
|
| 191 |
+
|
| 192 |
+
END OF TERMS AND CONDITIONS
|
| 193 |
+
|
| 194 |
+
APPENDIX: How to apply the Apache License to your work.
|
| 195 |
+
|
| 196 |
+
To apply the Apache License to your work, attach the following
|
| 197 |
+
boilerplate notice, with the fields enclosed by brackets "[]"
|
| 198 |
+
replaced with your own identifying information. (Don't include
|
| 199 |
+
the brackets!) The text should be enclosed in the appropriate
|
| 200 |
+
comment syntax for the file format. We also recommend that a
|
| 201 |
+
file or class name and description of purpose be included on the
|
| 202 |
+
same "printed page" as the copyright notice for easier
|
| 203 |
+
identification within third-party archives.
|
| 204 |
+
|
| 205 |
+
Copyright [yyyy] [name of copyright owner]
|
| 206 |
+
|
| 207 |
+
Licensed under the Apache License, Version 2.0 (the "License");
|
| 208 |
+
you may not use this file except in compliance with the License.
|
| 209 |
+
You may obtain a copy of the License at
|
| 210 |
+
|
| 211 |
+
http://www.apache.org/licenses/LICENSE-2.0
|
| 212 |
+
|
| 213 |
+
Unless required by applicable law or agreed to in writing, software
|
| 214 |
+
distributed under the License is distributed on an "AS IS" BASIS,
|
| 215 |
+
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
| 216 |
+
See the License for the specific language governing permissions and
|
| 217 |
+
limitations under the License.
|
| 218 |
+
|
| 219 |
+
Software: vllm 0.9.1
|
| 220 |
+
Copyright notice:
|
| 221 |
+
Copyright 2025 The vLLM team.
|
| 222 |
+
|
| 223 |
+
License Text:
|
| 224 |
+
----------------------------------------
|
| 225 |
+
|
| 226 |
+
Apache License
|
| 227 |
+
Version 2.0, January 2004
|
| 228 |
+
http://www.apache.org/licenses/
|
| 229 |
+
|
| 230 |
+
TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
|
| 231 |
+
|
| 232 |
+
1. Definitions.
|
| 233 |
+
|
| 234 |
+
"License" shall mean the terms and conditions for use, reproduction,
|
| 235 |
+
and distribution as defined by Sections 1 through 9 of this document.
|
| 236 |
+
|
| 237 |
+
"Licensor" shall mean the copyright owner or entity authorized by
|
| 238 |
+
the copyright owner that is granting the License.
|
| 239 |
+
|
| 240 |
+
"Legal Entity" shall mean the union of the acting entity and all
|
| 241 |
+
other entities that control, are controlled by, or are under common
|
| 242 |
+
control with that entity. For the purposes of this definition,
|
| 243 |
+
"control" means (i) the power, direct or indirect, to cause the
|
| 244 |
+
direction or management of such entity, whether by contract or
|
| 245 |
+
otherwise, or (ii) ownership of fifty percent (50%) or more of the
|
| 246 |
+
outstanding shares, or (iii) beneficial ownership of such entity.
|
| 247 |
+
|
| 248 |
+
"You" (or "Your") shall mean an individual or Legal Entity
|
| 249 |
+
exercising permissions granted by this License.
|
| 250 |
+
|
| 251 |
+
"Source" form shall mean the preferred form for making modifications,
|
| 252 |
+
including but not limited to software source code, documentation
|
| 253 |
+
source, and configuration files.
|
| 254 |
+
|
| 255 |
+
"Object" form shall mean any form resulting from mechanical
|
| 256 |
+
transformation or translation of a Source form, including but
|
| 257 |
+
not limited to compiled object code, generated documentation,
|
| 258 |
+
and conversions to other media types.
|
| 259 |
+
|
| 260 |
+
"Work" shall mean the work of authorship, whether in Source or
|
| 261 |
+
Object form, made available under the License, as indicated by a
|
| 262 |
+
copyright notice that is included in or attached to the work
|
| 263 |
+
(an example is provided in the Appendix below).
|
| 264 |
+
|
| 265 |
+
"Derivative Works" shall mean any work, whether in Source or Object
|
| 266 |
+
form, that is based on (or derived from) the Work and for which the
|
| 267 |
+
editorial revisions, annotations, elaborations, or other modifications
|
| 268 |
+
represent, as a whole, an original work of authorship. For the purposes
|
| 269 |
+
of this License, Derivative Works shall not include works that remain
|
| 270 |
+
separable from, or merely link (or bind by name) to the interfaces of,
|
| 271 |
+
the Work and Derivative Works thereof.
|
| 272 |
+
|
| 273 |
+
"Contribution" shall mean any work of authorship, including
|
| 274 |
+
the original version of the Work and any modifications or additions
|
| 275 |
+
to that Work or Derivative Works thereof, that is intentionally
|
| 276 |
+
submitted to Licensor for inclusion in the Work by the copyright owner
|
| 277 |
+
or by an individual or Legal Entity authorized to submit on behalf of
|
| 278 |
+
the copyright owner. For the purposes of this definition, "submitted"
|
| 279 |
+
means any form of electronic, verbal, or written communication sent
|
| 280 |
+
to the Licensor or its representatives, including but not limited to
|
| 281 |
+
communication on electronic mailing lists, source code control systems,
|
| 282 |
+
and issue tracking systems that are managed by, or on behalf of, the
|
| 283 |
+
Licensor for the purpose of discussing and improving the Work, but
|
| 284 |
+
excluding communication that is conspicuously marked or otherwise
|
| 285 |
+
designated in writing by the copyright owner as "Not a Contribution."
|
| 286 |
+
|
| 287 |
+
"Contributor" shall mean Licensor and any individual or Legal Entity
|
| 288 |
+
on behalf of whom a Contribution has been received by Licensor and
|
| 289 |
+
subsequently incorporated within the Work.
|
| 290 |
+
|
| 291 |
+
2. Grant of Copyright License. Subject to the terms and conditions of
|
| 292 |
+
this License, each Contributor hereby grants to You a perpetual,
|
| 293 |
+
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
|
| 294 |
+
copyright license to reproduce, prepare Derivative Works of,
|
| 295 |
+
publicly display, publicly perform, sublicense, and distribute the
|
| 296 |
+
Work and such Derivative Works in Source or Object form.
|
| 297 |
+
|
| 298 |
+
3. Grant of Patent License. Subject to the terms and conditions of
|
| 299 |
+
this License, each Contributor hereby grants to You a perpetual,
|
| 300 |
+
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
|
| 301 |
+
(except as stated in this section) patent license to make, have made,
|
| 302 |
+
use, offer to sell, sell, import, and otherwise transfer the Work,
|
| 303 |
+
where such license applies only to those patent claims licensable
|
| 304 |
+
by such Contributor that are necessarily infringed by their
|
| 305 |
+
Contribution(s) alone or by combination of their Contribution(s)
|
| 306 |
+
with the Work to which such Contribution(s) was submitted. If You
|
| 307 |
+
institute patent litigation against any entity (including a
|
| 308 |
+
cross-claim or counterclaim in a lawsuit) alleging that the Work
|
| 309 |
+
or a Contribution incorporated within the Work constitutes direct
|
| 310 |
+
or contributory patent infringement, then any patent licenses
|
| 311 |
+
granted to You under this License for that Work shall terminate
|
| 312 |
+
as of the date such litigation is filed.
|
| 313 |
+
|
| 314 |
+
4. Redistribution. You may reproduce and distribute copies of the
|
| 315 |
+
Work or Derivative Works thereof in any medium, with or without
|
| 316 |
+
modifications, and in Source or Object form, provided that You
|
| 317 |
+
meet the following conditions:
|
| 318 |
+
|
| 319 |
+
(a) You must give any other recipients of the Work or
|
| 320 |
+
Derivative Works a copy of this License; and
|
| 321 |
+
|
| 322 |
+
(b) You must cause any modified files to carry prominent notices
|
| 323 |
+
stating that You changed the files; and
|
| 324 |
+
|
| 325 |
+
(c) You must retain, in the Source form of any Derivative Works
|
| 326 |
+
that You distribute, all copyright, patent, trademark, and
|
| 327 |
+
attribution notices from the Source form of the Work,
|
| 328 |
+
excluding those notices that do not pertain to any part of
|
| 329 |
+
the Derivative Works; and
|
| 330 |
+
|
| 331 |
+
(d) If the Work includes a "NOTICE" text file as part of its
|
| 332 |
+
distribution, then any Derivative Works that You distribute must
|
| 333 |
+
include a readable copy of the attribution notices contained
|
| 334 |
+
within such NOTICE file, excluding those notices that do not
|
| 335 |
+
pertain to any part of the Derivative Works, in at least one
|
| 336 |
+
of the following places: within a NOTICE text file distributed
|
| 337 |
+
as part of the Derivative Works; within the Source form or
|
| 338 |
+
documentation, if provided along with the Derivative Works; or,
|
| 339 |
+
within a display generated by the Derivative Works, if and
|
| 340 |
+
wherever such third-party notices normally appear. The contents
|
| 341 |
+
of the NOTICE file are for informational purposes only and
|
| 342 |
+
do not modify the License. You may add Your own attribution
|
| 343 |
+
notices within Derivative Works that You distribute, alongside
|
| 344 |
+
or as an addendum to the NOTICE text from the Work, provided
|
| 345 |
+
that such additional attribution notices cannot be construed
|
| 346 |
+
as modifying the License.
|
| 347 |
+
|
| 348 |
+
You may add Your own copyright statement to Your modifications and
|
| 349 |
+
may provide additional or different license terms and conditions
|
| 350 |
+
for use, reproduction, or distribution of Your modifications, or
|
| 351 |
+
for any such Derivative Works as a whole, provided Your use,
|
| 352 |
+
reproduction, and distribution of the Work otherwise complies with
|
| 353 |
+
the conditions stated in this License.
|
| 354 |
+
|
| 355 |
+
5. Submission of Contributions. Unless You explicitly state otherwise,
|
| 356 |
+
any Contribution intentionally submitted for inclusion in the Work
|
| 357 |
+
by You to the Licensor shall be under the terms and conditions of
|
| 358 |
+
this License, without any additional terms or conditions.
|
| 359 |
+
Notwithstanding the above, nothing herein shall supersede or modify
|
| 360 |
+
the terms of any separate license agreement you may have executed
|
| 361 |
+
with Licensor regarding such Contributions.
|
| 362 |
+
|
| 363 |
+
6. Trademarks. This License does not grant permission to use the trade
|
| 364 |
+
names, trademarks, service marks, or product names of the Licensor,
|
| 365 |
+
except as required for reasonable and customary use in describing the
|
| 366 |
+
origin of the Work and reproducing the content of the NOTICE file.
|
| 367 |
+
|
| 368 |
+
7. Disclaimer of Warranty. Unless required by applicable law or
|
| 369 |
+
agreed to in writing, Licensor provides the Work (and each
|
| 370 |
+
Contributor provides its Contributions) on an "AS IS" BASIS,
|
| 371 |
+
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
|
| 372 |
+
implied, including, without limitation, any warranties or conditions
|
| 373 |
+
of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
|
| 374 |
+
PARTICULAR PURPOSE. You are solely responsible for determining the
|
| 375 |
+
appropriateness of using or redistributing the Work and assume any
|
| 376 |
+
risks associated with Your exercise of permissions under this License.
|
| 377 |
+
|
| 378 |
+
8. Limitation of Liability. In no event and under no legal theory,
|
| 379 |
+
whether in tort (including negligence), contract, or otherwise,
|
| 380 |
+
unless required by applicable law (such as deliberate and grossly
|
| 381 |
+
negligent acts) or agreed to in writing, shall any Contributor be
|
| 382 |
+
liable to You for damages, including any direct, indirect, special,
|
| 383 |
+
incidental, or consequential damages of any character arising as a
|
| 384 |
+
result of this License or out of the use or inability to use the
|
| 385 |
+
Work (including but not limited to damages for loss of goodwill,
|
| 386 |
+
work stoppage, computer failure or malfunction, or any and all
|
| 387 |
+
other commercial damages or losses), even if such Contributor
|
| 388 |
+
has been advised of the possibility of such damages.
|
| 389 |
+
|
| 390 |
+
9. Accepting Warranty or Additional Liability. While redistributing
|
| 391 |
+
the Work or Derivative Works thereof, You may choose to offer,
|
| 392 |
+
and charge a fee for, acceptance of support, warranty, indemnity,
|
| 393 |
+
or other liability obligations and/or rights consistent with this
|
| 394 |
+
License. However, in accepting such obligations, You may act only
|
| 395 |
+
on Your own behalf and on Your sole responsibility, not on behalf
|
| 396 |
+
of any other Contributor, and only if You agree to indemnify,
|
| 397 |
+
defend, and hold each Contributor harmless for any liability
|
| 398 |
+
incurred by, or claims asserted against, such Contributor by reason
|
| 399 |
+
of your accepting any such warranty or additional liability.
|
| 400 |
+
|
| 401 |
+
END OF TERMS AND CONDITIONS
|
| 402 |
+
|
| 403 |
+
APPENDIX: How to apply the Apache License to your work.
|
| 404 |
+
|
| 405 |
+
To apply the Apache License to your work, attach the following
|
| 406 |
+
boilerplate notice, with the fields enclosed by brackets "[]"
|
| 407 |
+
replaced with your own identifying information. (Don't include
|
| 408 |
+
the brackets!) The text should be enclosed in the appropriate
|
| 409 |
+
comment syntax for the file format. We also recommend that a
|
| 410 |
+
file or class name and description of purpose be included on the
|
| 411 |
+
same "printed page" as the copyright notice for easier
|
| 412 |
+
identification within third-party archives.
|
| 413 |
+
|
| 414 |
+
Copyright [yyyy] [name of copyright owner]
|
| 415 |
+
|
| 416 |
+
Licensed under the Apache License, Version 2.0 (the "License");
|
| 417 |
+
you may not use this file except in compliance with the License.
|
| 418 |
+
You may obtain a copy of the License at
|
| 419 |
+
|
| 420 |
+
http://www.apache.org/licenses/LICENSE-2.0
|
| 421 |
+
|
| 422 |
+
Unless required by applicable law or agreed to in writing, software
|
| 423 |
+
distributed under the License is distributed on an "AS IS" BASIS,
|
| 424 |
+
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
| 425 |
+
See the License for the specific language governing permissions and
|
| 426 |
+
limitations under the License.
|
| 427 |
+
|
| 428 |
+
Software: vllm-ascend 0.9.1
|
| 429 |
+
Copyright notice:
|
| 430 |
+
Copyright (c) 2025 Huawei Technologies Co., Ltd. All Rights Reserved.
|
| 431 |
+
|
| 432 |
+
License Text:
|
| 433 |
+
----------------------------------------
|
| 434 |
+
|
| 435 |
+
Apache License
|
| 436 |
+
Version 2.0, January 2004
|
| 437 |
+
http://www.apache.org/licenses/
|
| 438 |
+
|
| 439 |
+
TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
|
| 440 |
+
|
| 441 |
+
1. Definitions.
|
| 442 |
+
|
| 443 |
+
"License" shall mean the terms and conditions for use, reproduction,
|
| 444 |
+
and distribution as defined by Sections 1 through 9 of this document.
|
| 445 |
+
|
| 446 |
+
"Licensor" shall mean the copyright owner or entity authorized by
|
| 447 |
+
the copyright owner that is granting the License.
|
| 448 |
+
|
| 449 |
+
"Legal Entity" shall mean the union of the acting entity and all
|
| 450 |
+
other entities that control, are controlled by, or are under common
|
| 451 |
+
control with that entity. For the purposes of this definition,
|
| 452 |
+
"control" means (i) the power, direct or indirect, to cause the
|
| 453 |
+
direction or management of such entity, whether by contract or
|
| 454 |
+
otherwise, or (ii) ownership of fifty percent (50%) or more of the
|
| 455 |
+
outstanding shares, or (iii) beneficial ownership of such entity.
|
| 456 |
+
|
| 457 |
+
"You" (or "Your") shall mean an individual or Legal Entity
|
| 458 |
+
exercising permissions granted by this License.
|
| 459 |
+
|
| 460 |
+
"Source" form shall mean the preferred form for making modifications,
|
| 461 |
+
including but not limited to software source code, documentation
|
| 462 |
+
source, and configuration files.
|
| 463 |
+
|
| 464 |
+
"Object" form shall mean any form resulting from mechanical
|
| 465 |
+
transformation or translation of a Source form, including but
|
| 466 |
+
not limited to compiled object code, generated documentation,
|
| 467 |
+
and conversions to other media types.
|
| 468 |
+
|
| 469 |
+
"Work" shall mean the work of authorship, whether in Source or
|
| 470 |
+
Object form, made available under the License, as indicated by a
|
| 471 |
+
copyright notice that is included in or attached to the work
|
| 472 |
+
(an example is provided in the Appendix below).
|
| 473 |
+
|
| 474 |
+
"Derivative Works" shall mean any work, whether in Source or Object
|
| 475 |
+
form, that is based on (or derived from) the Work and for which the
|
| 476 |
+
editorial revisions, annotations, elaborations, or other modifications
|
| 477 |
+
represent, as a whole, an original work of authorship. For the purposes
|
| 478 |
+
of this License, Derivative Works shall not include works that remain
|
| 479 |
+
separable from, or merely link (or bind by name) to the interfaces of,
|
| 480 |
+
the Work and Derivative Works thereof.
|
| 481 |
+
|
| 482 |
+
"Contribution" shall mean any work of authorship, including
|
| 483 |
+
the original version of the Work and any modifications or additions
|
| 484 |
+
to that Work or Derivative Works thereof, that is intentionally
|
| 485 |
+
submitted to Licensor for inclusion in the Work by the copyright owner
|
| 486 |
+
or by an individual or Legal Entity authorized to submit on behalf of
|
| 487 |
+
the copyright owner. For the purposes of this definition, "submitted"
|
| 488 |
+
means any form of electronic, verbal, or written communication sent
|
| 489 |
+
to the Licensor or its representatives, including but not limited to
|
| 490 |
+
communication on electronic mailing lists, source code control systems,
|
| 491 |
+
and issue tracking systems that are managed by, or on behalf of, the
|
| 492 |
+
Licensor for the purpose of discussing and improving the Work, but
|
| 493 |
+
excluding communication that is conspicuously marked or otherwise
|
| 494 |
+
designated in writing by the copyright owner as "Not a Contribution."
|
| 495 |
+
|
| 496 |
+
"Contributor" shall mean Licensor and any individual or Legal Entity
|
| 497 |
+
on behalf of whom a Contribution has been received by Licensor and
|
| 498 |
+
subsequently incorporated within the Work.
|
| 499 |
+
|
| 500 |
+
2. Grant of Copyright License. Subject to the terms and conditions of
|
| 501 |
+
this License, each Contributor hereby grants to You a perpetual,
|
| 502 |
+
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
|
| 503 |
+
copyright license to reproduce, prepare Derivative Works of,
|
| 504 |
+
publicly display, publicly perform, sublicense, and distribute the
|
| 505 |
+
Work and such Derivative Works in Source or Object form.
|
| 506 |
+
|
| 507 |
+
3. Grant of Patent License. Subject to the terms and conditions of
|
| 508 |
+
this License, each Contributor hereby grants to You a perpetual,
|
| 509 |
+
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
|
| 510 |
+
(except as stated in this section) patent license to make, have made,
|
| 511 |
+
use, offer to sell, sell, import, and otherwise transfer the Work,
|
| 512 |
+
where such license applies only to those patent claims licensable
|
| 513 |
+
by such Contributor that are necessarily infringed by their
|
| 514 |
+
Contribution(s) alone or by combination of their Contribution(s)
|
| 515 |
+
with the Work to which such Contribution(s) was submitted. If You
|
| 516 |
+
institute patent litigation against any entity (including a
|
| 517 |
+
cross-claim or counterclaim in a lawsuit) alleging that the Work
|
| 518 |
+
or a Contribution incorporated within the Work constitutes direct
|
| 519 |
+
or contributory patent infringement, then any patent licenses
|
| 520 |
+
granted to You under this License for that Work shall terminate
|
| 521 |
+
as of the date such litigation is filed.
|
| 522 |
+
|
| 523 |
+
4. Redistribution. You may reproduce and distribute copies of the
|
| 524 |
+
Work or Derivative Works thereof in any medium, with or without
|
| 525 |
+
modifications, and in Source or Object form, provided that You
|
| 526 |
+
meet the following conditions:
|
| 527 |
+
|
| 528 |
+
(a) You must give any other recipients of the Work or
|
| 529 |
+
Derivative Works a copy of this License; and
|
| 530 |
+
|
| 531 |
+
(b) You must cause any modified files to carry prominent notices
|
| 532 |
+
stating that You changed the files; and
|
| 533 |
+
|
| 534 |
+
(c) You must retain, in the Source form of any Derivative Works
|
| 535 |
+
that You distribute, all copyright, patent, trademark, and
|
| 536 |
+
attribution notices from the Source form of the Work,
|
| 537 |
+
excluding those notices that do not pertain to any part of
|
| 538 |
+
the Derivative Works; and
|
| 539 |
+
|
| 540 |
+
(d) If the Work includes a "NOTICE" text file as part of its
|
| 541 |
+
distribution, then any Derivative Works that You distribute must
|
| 542 |
+
include a readable copy of the attribution notices contained
|
| 543 |
+
within such NOTICE file, excluding those notices that do not
|
| 544 |
+
pertain to any part of the Derivative Works, in at least one
|
| 545 |
+
of the following places: within a NOTICE text file distributed
|
| 546 |
+
as part of the Derivative Works; within the Source form or
|
| 547 |
+
documentation, if provided along with the Derivative Works; or,
|
| 548 |
+
within a display generated by the Derivative Works, if and
|
| 549 |
+
wherever such third-party notices normally appear. The contents
|
| 550 |
+
of the NOTICE file are for informational purposes only and
|
| 551 |
+
do not modify the License. You may add Your own attribution
|
| 552 |
+
notices within Derivative Works that You distribute, alongside
|
| 553 |
+
or as an addendum to the NOTICE text from the Work, provided
|
| 554 |
+
that such additional attribution notices cannot be construed
|
| 555 |
+
as modifying the License.
|
| 556 |
+
|
| 557 |
+
You may add Your own copyright statement to Your modifications and
|
| 558 |
+
may provide additional or different license terms and conditions
|
| 559 |
+
for use, reproduction, or distribution of Your modifications, or
|
| 560 |
+
for any such Derivative Works as a whole, provided Your use,
|
| 561 |
+
reproduction, and distribution of the Work otherwise complies with
|
| 562 |
+
the conditions stated in this License.
|
| 563 |
+
|
| 564 |
+
5. Submission of Contributions. Unless You explicitly state otherwise,
|
| 565 |
+
any Contribution intentionally submitted for inclusion in the Work
|
| 566 |
+
by You to the Licensor shall be under the terms and conditions of
|
| 567 |
+
this License, without any additional terms or conditions.
|
| 568 |
+
Notwithstanding the above, nothing herein shall supersede or modify
|
| 569 |
+
the terms of any separate license agreement you may have executed
|
| 570 |
+
with Licensor regarding such Contributions.
|
| 571 |
+
|
| 572 |
+
6. Trademarks. This License does not grant permission to use the trade
|
| 573 |
+
names, trademarks, service marks, or product names of the Licensor,
|
| 574 |
+
except as required for reasonable and customary use in describing the
|
| 575 |
+
origin of the Work and reproducing the content of the NOTICE file.
|
| 576 |
+
|
| 577 |
+
7. Disclaimer of Warranty. Unless required by applicable law or
|
| 578 |
+
agreed to in writing, Licensor provides the Work (and each
|
| 579 |
+
Contributor provides its Contributions) on an "AS IS" BASIS,
|
| 580 |
+
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
|
| 581 |
+
implied, including, without limitation, any warranties or conditions
|
| 582 |
+
of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
|
| 583 |
+
PARTICULAR PURPOSE. You are solely responsible for determining the
|
| 584 |
+
appropriateness of using or redistributing the Work and assume any
|
| 585 |
+
risks associated with Your exercise of permissions under this License.
|
| 586 |
+
|
| 587 |
+
8. Limitation of Liability. In no event and under no legal theory,
|
| 588 |
+
whether in tort (including negligence), contract, or otherwise,
|
| 589 |
+
unless required by applicable law (such as deliberate and grossly
|
| 590 |
+
negligent acts) or agreed to in writing, shall any Contributor be
|
| 591 |
+
liable to You for damages, including any direct, indirect, special,
|
| 592 |
+
incidental, or consequential damages of any character arising as a
|
| 593 |
+
result of this License or out of the use or inability to use the
|
| 594 |
+
Work (including but not limited to damages for loss of goodwill,
|
| 595 |
+
work stoppage, computer failure or malfunction, or any and all
|
| 596 |
+
other commercial damages or losses), even if such Contributor
|
| 597 |
+
has been advised of the possibility of such damages.
|
| 598 |
+
|
| 599 |
+
9. Accepting Warranty or Additional Liability. While redistributing
|
| 600 |
+
the Work or Derivative Works thereof, You may choose to offer,
|
| 601 |
+
and charge a fee for, acceptance of support, warranty, indemnity,
|
| 602 |
+
or other liability obligations and/or rights consistent with this
|
| 603 |
+
License. However, in accepting such obligations, You may act only
|
| 604 |
+
on Your own behalf and on Your sole responsibility, not on behalf
|
| 605 |
+
of any other Contributor, and only if You agree to indemnify,
|
| 606 |
+
defend, and hold each Contributor harmless for any liability
|
| 607 |
+
incurred by, or claims asserted against, such Contributor by reason
|
| 608 |
+
of your accepting any such warranty or additional liability.
|
| 609 |
+
|
| 610 |
+
END OF TERMS AND CONDITIONS
|
| 611 |
+
|
| 612 |
+
APPENDIX: How to apply the Apache License to your work.
|
| 613 |
+
|
| 614 |
+
To apply the Apache License to your work, attach the following
|
| 615 |
+
boilerplate notice, with the fields enclosed by brackets "[]"
|
| 616 |
+
replaced with your own identifying information. (Don't include
|
| 617 |
+
the brackets!) The text should be enclosed in the appropriate
|
| 618 |
+
comment syntax for the file format. We also recommend that a
|
| 619 |
+
file or class name and description of purpose be included on the
|
| 620 |
+
same "printed page" as the copyright notice for easier
|
| 621 |
+
identification within third-party archives.
|
| 622 |
+
|
| 623 |
+
Copyright [yyyy] [name of copyright owner]
|
| 624 |
+
|
| 625 |
+
Licensed under the Apache License, Version 2.0 (the "License");
|
| 626 |
+
you may not use this file except in compliance with the License.
|
| 627 |
+
You may obtain a copy of the License at
|
| 628 |
+
|
| 629 |
+
http://www.apache.org/licenses/LICENSE-2.0
|
| 630 |
+
|
| 631 |
+
Unless required by applicable law or agreed to in writing, software
|
| 632 |
+
distributed under the License is distributed on an "AS IS" BASIS,
|
| 633 |
+
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
| 634 |
+
See the License for the specific language governing permissions and
|
| 635 |
+
limitations under the License.
|
README.md
CHANGED
|
@@ -1,5 +1,85 @@
|
|
| 1 |
-
|
| 2 |
-
|
| 3 |
-
|
| 4 |
-
|
| 5 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# openPangu-R-72B-2512
|
| 2 |
+
中文 | [English](README_EN.md)
|
| 3 |
+
|
| 4 |
+
## 1. 简介
|
| 5 |
+
openPangu-R-72B-2512 是基于昇腾集群训练的MoE模型。模型总参数量74B，激活参数量15B，采用80选8的专家选择机制，支持128k长序列处理。训练数据总量约24T tokens。同一个模型支持快慢思考两种模式切换，慢思考模式下，支持思维链分档（“低”和“高”两种推理深度）。
|
| 6 |
+
|
| 7 |
+
## 2. 模型架构
|
| 8 |
+
openPangu-R-72B-2512 在模型稳定收敛和效果提升方向进行了以下优化：
|
| 9 |
+
- 在注意力机制中引入参数式Sink Token技术： 有效缓解极大激活值问题，训练中最大激活值从$10^3$降至$10^2$量级， 提升训练稳定性并对后量化亲和。
|
| 10 |
+
|
| 11 |
+
- K-Norm与Depth-Scaled Sandwich-Norm：为保证attention logits的稳定性，我们采用了K-Norm结构。K-Norm与QK-Norm类似，但只对attention的key施加RMS Norm。K-Norm可起到与QK-Norm类似的稳定性效果，但引入的计算开销更小，同时K-Norm不影响Query的scale，带来更灵活的表达能力。为了保证残差连接的稳定性，我们采用了Depth-Scaled Sandwich-Norm方法。
|
| 12 |
+
|
| 13 |
+
- 注意力架构优化：增加Query头数和注意力头维度，使模型能够从更多角度捕获细粒度语义关系。引入Partial RoPE机制，仅对Query和Key中1/3维度应用位置编码。尽管Key头维度有所增加，但通过将KV组数量减半，KV cache仍可减少37.5%，在保持推理阶段显存和速度优化的同时，实现了更低的训练损失和更优的推理性能。
|
| 14 |
+
|
| 15 |
+
- Adaptive Aux Free负载优化技术：能够自适应调整专家bias更新幅度，减少均衡震荡现象，优化专家负载分布均衡性。
|
| 16 |
+
|
| 17 |
+
|
| 18 |
+
详细架构参数如下：
|
| 19 |
+
|
| 20 |
+
| | |
|
| 21 |
+
|:---:|:---:|
|
| 22 |
+
| **Architecture** | Mixture-of-Experts (MoE) |
|
| 23 |
+
| **Total Parameters** | 74B |
|
| 24 |
+
| **Activated Parameters** | 15B |
|
| 25 |
+
| **Number of Layers** (Dense layer included) | 50 |
|
| 26 |
+
| **Number of Dense Layers** | 4 |
|
| 27 |
+
| **Number of MTP Modules** | 1 |
|
| 28 |
+
| **Hidden Dimension** | 4608 |
|
| 29 |
+
| **MoE Hidden Dimension** (per Expert) | 1280 |
|
| 30 |
+
| **Attention Mechanism** | GQA |
|
| 31 |
+
| **Number of Attention Heads** | 64 |
|
| 32 |
+
| **Number of Query Groups** | 4 |
|
| 33 |
+
| **Number of Experts** | 80 |
|
| 34 |
+
| **Selected Experts per Token** | 8 |
|
| 35 |
+
| **Number of Shared Experts** | 2 |
|
| 36 |
+
| **Vocabulary Size** | 153K |
|
| 37 |
+
| **Context Length** | 128K |
|
| 38 |
+
|
| 39 |
+
|
| 40 |
+
## 3. 测评结果
|
| 41 |
+
|
| 42 |
+
| 测评集 | 测评指标 | openPangu-R-72B-2512 快思考 | openPangu-R-72B-2512 慢思考 |
|
| 43 |
+
|:------------------:|:----------------------------:|:-----:|:-----:|
|
| 44 |
+
| **通用能力** | | |
|
| 45 |
+
| LiveBench | Acc (2024-11-25) | 67.3 | 75.2 |
|
| 46 |
+
| MMLU-Pro | Exact Match | 84.2 | 84.8 |
|
| 47 |
+
| MMLU-ProX | Acc | 76.9 | 80.6 |
|
| 48 |
+
| RULER | Acc | 95.6 | 94.7 |
|
| 49 |
+
| LongBench V2 | Acc |45.3 |55.3 |
|
| 50 |
+
| IF-Eval | Prompt Strict | 86.3 | 79.1 |
|
| 51 |
+
| Hallucination-LeaderBoard | 1-HHEM | 96.5 | 97.1 |
|
| 52 |
+
| GPQA-Dimaond | Avg@4 | 76.8 | 83.2 |
|
| 53 |
+
| SuperGPQA | Acc | 58.9 | 64.2 |
|
| 54 |
+
| **数学能力** | | |
|
| 55 |
+
| AIME24 | Avg@16 | 75.6 | 89.0 |
|
| 56 |
+
| AIME25 | Avg@16 | 60.6 | 81.3 |
|
| 57 |
+
| CNMO 2024 | Avg@32 | 77.8 | 82.8 |
|
| 58 |
+
| HMMT 2025 | Avg@16 (February) | 45.4 | 74.8 |
|
| 59 |
+
| **代码能力** | | |
|
| 60 |
+
| LiveCodeBench V6 | Avg@3 (01/25~05/25) | 41.9 | 69.5 |
|
| 61 |
+
| Codeforces | Elo Avg@3 (02/25~09/25) | 1044.5 | 1701.4 |
|
| 62 |
+
| **Agent工具调用** | | |
|
| 63 |
+
| BFCL-V3 | Acc (Prompt) | 74.6 | 76.5 |
|
| 64 |
+
| Tau-Bench (airline) | Avg@3 (FC) | 45.3 | 56.0 |
|
| 65 |
+
| Tau-Bench (retail) | Avg@3 (FC) | 70.1 | 73.0 |
|
| 66 |
+
| Tau2-Bench (airline) | Avg@3 (FC) | 58.0 | 65.3 |
|
| 67 |
+
| Tau2-Bench (retail) | Avg@3 (FC) | 71.4 | 78.7 |
|
| 68 |
+
| Tau2-Bench (telecom) | Avg@3 (FC) | 48.8 | 49.4 |
|
| 69 |
+
| AceBench | Acc (Prompt) | 74.3 | 79.6 |
|
| 70 |
+
|
| 71 |
+
|
| 72 |
+
## 4. 部署和使用
|
| 73 |
+
- 使用omni-infer推理框架，参考[[omniinfer_for_openpangu_r_72b_2512](doc/omniinfer_for_openpangu_r_72b_2512.md)]
|
| 74 |
+
|
| 75 |
+
## 5. 模型许可证
|
| 76 |
+
除文件中对开源许可证另有约定外，openPangu-R-72B-2512 模型根据 OPENPANGU MODEL LICENSE AGREEMENT VERSION 1.0 授权，旨在允许使用并促进人工智能技术的进一步发展。有关详细信息，请参阅模型存储���根目录中的 [LICENSE](LICENSE) 文件。
|
| 77 |
+
|
| 78 |
+
## 6. 免责声明
|
| 79 |
+
由于 openPangu-R-72B-2512 （“模型”）所依赖的技术固有的限制，以及人工智能生成的内容是由盘古自动生成的，华为无法对以下事项做出任何保证：
|
| 80 |
+
- 该模型的输出通过AI算法自动生成，不能排除某些信息可能存在缺陷、不合理或引起不适的可能性，生成的内容不代表华为的态度或立场；
|
| 81 |
+
- 无法保证该模型100%准确、可靠、功能齐全、及时、安全、无错误、不间断、持续稳定或无任何故障；
|
| 82 |
+
- 该模型的输出内容不构成任何建议或决策，也不保证生成的内容的真实性、完整性、准确性、及时性、合法性、功能性或实用性。生成的内容不能替代医疗、法律等领域的专业人士回答您的问题。生成的内容仅供参考，不代表华为的任何态度、立场或观点。您需要根据实际情况做出独立判断，华为不承担任何责任。
|
| 83 |
+
|
| 84 |
+
## 7. 反馈
|
| 85 |
+
如果有任何意见和建议，请提交issue或联系[openPangu@huawei.com](url)。
|
README_EN.md
ADDED
|
@@ -0,0 +1,85 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# openPangu-R-72B-2512
|
| 2 |
+
[中文](README.md) | English
|
| 3 |
+
|
| 4 |
+
## 1. Introduction
|
| 5 |
+
openPangu-R-72B-2512 is an MoE model trained on Ascend. The model has 74B total parameters and 15B activated parameters. It selects top 8 experts out of 80 routed experts. Its context length is 128k. The total pretraining data contains 24T tokens. It supports switching between two modes (fast-thinking and slow-thinking). In slow-thinking mode, we support two types of reasoning effort ('low' and 'high').
|
| 6 |
+
|
| 7 |
+
## 2. Architecture
|
| 8 |
+
openPangu-R-72B-2512 includes several enhancements:
|
| 9 |
+
- Parametric sink token: Effectively mitigates the problem of extremely large activation values, reducing the maximum activation value from the order of $10^3$ to $10^2$ during training, which improves training stability and enhances compatibility with post-quantization.
|
| 10 |
+
|
| 11 |
+
- K-Norm and Depth-Scaled Sandwich-Norm: To ensure the stability of attention logits, we apply K-Norm, a structure analogous to QK-Norm but applies RMS Norm solely to the attention keys. This approach achieves stability effects comparable to QK-Norm while introducing less computational overhead. Moreover, by preserving the original scale of Query, K-Norm offer greater expressive flexibility. To maintain the stability of residual connections, we employ the Depth-Scaled Sandwich-Norm.
|
| 12 |
+
|
| 13 |
+
- Attention design: We increase Query heads and attention head dimensions to enable the model to capture fine-grained semantic relationships from multiple perspectives. The Partial RoPE mechanism applies positional encoding to only 1/3 of the dimensions in Query and Key. Although the Key head dimension increases, halving the number of KV groups still reduces KV cache by 37.5%, achieving lower training loss and improved inference performance while maintaining memory and speed optimizations during the inference stage.
|
| 14 |
+
|
| 15 |
+
- Adaptive Aux-Free Load Balancing Strategy：This approach adaptively adjusts the update magnitude of expert bias, mitigates balancing oscillations, and optimizes the equilibrium of expert load distribution.
|
| 16 |
+
|
| 17 |
+
Hyperparameters related to model architecture are as follows:
|
| 18 |
+
|
| 19 |
+
| | |
|
| 20 |
+
|:---:|:---:|
|
| 21 |
+
| **Architecture** | Mixture-of-Experts (MoE) |
|
| 22 |
+
| **Total Parameters** | 74B |
|
| 23 |
+
| **Activated Parameters** | 15B |
|
| 24 |
+
| **Number of Layers** (Dense layer included) | 50 |
|
| 25 |
+
| **Number of Dense Layers** | 4 |
|
| 26 |
+
| **Number of MTP Modules** | 1 |
|
| 27 |
+
| **Hidden Dimension** | 4608 |
|
| 28 |
+
| **MoE Hidden Dimension** (per Expert) | 1280 |
|
| 29 |
+
| **Attention Mechanism** | GQA |
|
| 30 |
+
| **Number of Attention Heads** | 64 |
|
| 31 |
+
| **Number of Query Groups** | 4 |
|
| 32 |
+
| **Number of Experts** | 80 |
|
| 33 |
+
| **Selected Experts per Token** | 8 |
|
| 34 |
+
| **Number of Shared Experts** | 2 |
|
| 35 |
+
| **Vocabulary Size** | 153K |
|
| 36 |
+
| **Context Length** | 128K |
|
| 37 |
+
|
| 38 |
+
## 3. Results
|
| 39 |
+
| Benchmark | Metric | openPangu-R-72B-2512 Fast-thinking | openPangu-R-72B-2512 Slow-thinking |
|
| 40 |
+
|:------------------:|:----------------------------:|:-----:|:-----:|
|
| 41 |
+
| **General** | | |
|
| 42 |
+
| LiveBench | Acc (2024-11-25) | 67.3 | 75.2 |
|
| 43 |
+
| MMLU-Pro | Exact Match | 84.2 | 84.8 |
|
| 44 |
+
| MMLU-ProX | Acc | 76.9 | 80.6 |
|
| 45 |
+
| RULER | Acc | 95.6 | 94.7 |
|
| 46 |
+
| LongBench V2 | Acc |45.3 |55.3 |
|
| 47 |
+
| IF-Eval | Prompt Strict | 86.3 | 79.1 |
|
| 48 |
+
| Hallucination-LeaderBoard | 1-HHEM | 96.5 | 97.1 |
|
| 49 |
+
| GPQA-Dimaond | Avg@4 | 76.8 | 83.2 |
|
| 50 |
+
| SuperGPQA | Acc | 58.9 | 64.2 |
|
| 51 |
+
| **Math** | | |
|
| 52 |
+
| AIME24 | Avg@16 | 75.6 | 89.0 |
|
| 53 |
+
| AIME25 | Avg@16 | 60.6 | 81.3 |
|
| 54 |
+
| CNMO 2024 | Avg@32 | 77.8 | 82.8 |
|
| 55 |
+
| HMMT 2025 | Avg@16 (February) | 45.4 | 74.8 |
|
| 56 |
+
| **Coding** | | |
|
| 57 |
+
| LiveCodeBench V6 | Avg@3 (01/25~05/25) | 41.9 | 69.5 |
|
| 58 |
+
| Codeforces | Elo Avg@3 (02/25~09/25) | 1044.5 | 1701.4 |
|
| 59 |
+
| **Agentic Tool Use** | | |
|
| 60 |
+
| BFCL-V3 | Acc (Prompt) | 74.6 | 76.5 |
|
| 61 |
+
| Tau-Bench (airline) | Avg@3 (FC) | 45.3 | 56.0 |
|
| 62 |
+
| Tau-Bench (retail) | Avg@3 (FC) | 70.1 | 73.0 |
|
| 63 |
+
| Tau2-Bench (airline) | Avg@3 (FC) | 58.0 | 65.3 |
|
| 64 |
+
| Tau2-Bench (retail) | Avg@3 (FC) | 71.4 | 78.7 |
|
| 65 |
+
| Tau2-Bench (telecom) | Avg@3 (FC) | 48.8 | 49.4 |
|
| 66 |
+
| AceBench | Acc (Prompt) | 74.3 | 79.6 |
|
| 67 |
+
|
| 68 |
+
## 4. Deployment
|
| 69 |
+
- omni-infer：please refer to [[omniinfer_for_openpangu_r_72b_2512](doc/omniinfer_for_openpangu_r_72b_2512_EN.md)]
|
| 70 |
+
|
| 71 |
+
## 5. Model License
|
| 72 |
+
Unless otherwise noted, the openPangu-R-72B-2512 model is licensed under the terms and conditions of OPENPANGU MODEL LICENSE AGREEMENT VERSION 1.0, which is intended to be used permissively and enable the further development of artificial intelligence technologies. Please refer to the [LICENSE](LICENSE) file located in the root directory of the model repository for details.
|
| 73 |
+
|
| 74 |
+
## 6. Disclaimer
|
| 75 |
+
Due to the technical limitations inherent in the technology on which the openPangu-R-72B-2512 model (“Model”) relies and the fact that the artificial intelligence generated content is automatically produced by Model, Huawei cannot make any guarantees regarding the following matters:
|
| 76 |
+
|
| 77 |
+
- The output of this Model is automatically generated via AI algorithms, it does not rule out the possibility that some of the information may be flawed, unreasonable, or cause discomfort, and the generated content does not represent Huawei's attitude or standpoint;
|
| 78 |
+
- There is no guarantee that this Model is 100% accurate, reliable, functional, timely, secure and safety, error-free, uninterrupted, continuously stable, or free of any faults;
|
| 79 |
+
- The output of this Model does not constitute any advices or decisions for you, and it does not guarantee the authenticity, completeness, accuracy, timeliness, legality, functionality, or practicality of the generated content. The generated content cannot replace professionals in medical, legal, and other fields in answering your questions. The generated content is for your reference only and does not represent any attitude, standpoint, or position of Huawei. You need to make independent judgments based on your actual situation, and Huawei does not assume any responsibilities.
|
| 80 |
+
|
| 81 |
+
## 7. Contact
|
| 82 |
+
If you have any question, please raise an issue or contact us at [openPangu@huawei.com](url).
|
| 83 |
+
|
| 84 |
+
|
| 85 |
+
|
config.json
ADDED
|
@@ -0,0 +1,45 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"PanguProMoEV2ForCausalLM"
|
| 4 |
+
],
|
| 5 |
+
"attention_dropout": 0.0,
|
| 6 |
+
"auto_map": {
|
| 7 |
+
"AutoConfig": "configuration_pangu_moe.PanguProMoEConfig",
|
| 8 |
+
"AutoModel": "modeling_pangu_moe.PanguProMoEModel",
|
| 9 |
+
"AutoModelForCausalLM": "modeling_pangu_moe.PanguProMoEForCausalLM"
|
| 10 |
+
},
|
| 11 |
+
"bos_token_id": 1,
|
| 12 |
+
"eos_token_id": 45892,
|
| 13 |
+
"first_k_dense_replace": 4,
|
| 14 |
+
"hidden_act": "silu",
|
| 15 |
+
"hidden_size": 4608,
|
| 16 |
+
"initializer_range": 0.02,
|
| 17 |
+
"intermediate_size": 10240,
|
| 18 |
+
"max_position_embeddings": 4096,
|
| 19 |
+
"model_type": "PanguProMoE",
|
| 20 |
+
"moe_intermediate_size": 1280,
|
| 21 |
+
"n_routed_experts": 80,
|
| 22 |
+
"n_shared_experts": 2,
|
| 23 |
+
"norm_topk_prob": true,
|
| 24 |
+
"num_attention_heads": 64,
|
| 25 |
+
"num_experts_per_tok": 8,
|
| 26 |
+
"num_hidden_layers": 50,
|
| 27 |
+
"num_key_value_heads": 4,
|
| 28 |
+
"num_nextn_predict_layers": 1,
|
| 29 |
+
"output_router_logits": false,
|
| 30 |
+
"rms_norm_eps": 1e-05,
|
| 31 |
+
"rope_theta": 10000,
|
| 32 |
+
"routed_scaling_factor": 2.5,
|
| 33 |
+
"router_enable_expert_bias": true,
|
| 34 |
+
"sandwich_norm": true,
|
| 35 |
+
"tie_word_embeddings": false,
|
| 36 |
+
"torch_dtype": "bfloat16",
|
| 37 |
+
"transformers_version": "4.48.2",
|
| 38 |
+
"use_cache": true,
|
| 39 |
+
"vocab_size": 153600,
|
| 40 |
+
"qk_nope_dim": 128,
|
| 41 |
+
"qk_rope_dim": 64,
|
| 42 |
+
"v_channels": 128,
|
| 43 |
+
"param_sink_number": 128,
|
| 44 |
+
"param_sink_with_value": true
|
| 45 |
+
}
|
configuration_pangu_moe.py
ADDED
|
@@ -0,0 +1,96 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# coding=utf-8
|
| 2 |
+
# Copyright (c) Huawei Technologies Co., Ltd. 2025. All rights reserved.
|
| 3 |
+
# Copyright 2024 The Qwen team, Alibaba Group and the HuggingFace Inc. team. All rights reserved.
|
| 4 |
+
#
|
| 5 |
+
# Licensed under the Apache License, Version 2.0 (the "License");
|
| 6 |
+
# you may not use this file except in compliance with the License.
|
| 7 |
+
# You may obtain a copy of the License at
|
| 8 |
+
#
|
| 9 |
+
# http://www.apache.org/licenses/LICENSE-2.0
|
| 10 |
+
#
|
| 11 |
+
# Unless required by applicable law or agreed to in writing, software
|
| 12 |
+
# distributed under the License is distributed on an "AS IS" BASIS,
|
| 13 |
+
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
| 14 |
+
# See the License for the specific language governing permissions and
|
| 15 |
+
# limitations under the License.
|
| 16 |
+
""" PanguProMoE model configuration"""
|
| 17 |
+
|
| 18 |
+
|
| 19 |
+
from transformers.configuration_utils import PretrainedConfig
|
| 20 |
+
from transformers.utils import logging
|
| 21 |
+
|
| 22 |
+
|
| 23 |
+
logger = logging.get_logger(__name__)
|
| 24 |
+
|
| 25 |
+
|
| 26 |
+
class PanguProMoEConfig(PretrainedConfig):
|
| 27 |
+
|
| 28 |
+
model_type = "PanguProMoE"
|
| 29 |
+
_auto_class = "AutoConfig"
|
| 30 |
+
|
| 31 |
+
def __init__(
|
| 32 |
+
self,
|
| 33 |
+
vocab_size=153376,
|
| 34 |
+
hidden_size=4608,
|
| 35 |
+
intermediate_size=10240,
|
| 36 |
+
num_hidden_layers=50,
|
| 37 |
+
num_attention_heads=64,
|
| 38 |
+
num_key_value_heads=4,
|
| 39 |
+
mlp_only_layers=[0,1,2,3],
|
| 40 |
+
hidden_act="silu",
|
| 41 |
+
max_position_embeddings=8192,
|
| 42 |
+
initializer_range=0.02,
|
| 43 |
+
rms_norm_eps=1e-5,
|
| 44 |
+
use_cache=True,
|
| 45 |
+
tie_word_embeddings=False,
|
| 46 |
+
rope_theta=100000,
|
| 47 |
+
moe_intermediate_size=1280,
|
| 48 |
+
shared_expert_intermediate_size=2560,
|
| 49 |
+
num_experts_per_tok=8,
|
| 50 |
+
num_experts=80,
|
| 51 |
+
norm_topk_prob=True,
|
| 52 |
+
router_enable_expert_bias=True,
|
| 53 |
+
output_router_logits=False,
|
| 54 |
+
routed_scaling_factor=2.5,
|
| 55 |
+
qk_nope_dim = 128,
|
| 56 |
+
qk_rope_dim = 64,
|
| 57 |
+
v_channels = 128,
|
| 58 |
+
sandwich_norm=True,
|
| 59 |
+
param_sink_number = 128,
|
| 60 |
+
param_sink_with_value=True,
|
| 61 |
+
**kwargs,
|
| 62 |
+
):
|
| 63 |
+
self.vocab_size = vocab_size
|
| 64 |
+
self.max_position_embeddings = max_position_embeddings
|
| 65 |
+
self.hidden_size = hidden_size
|
| 66 |
+
self.num_hidden_layers = num_hidden_layers
|
| 67 |
+
self.num_attention_heads = num_attention_heads
|
| 68 |
+
self.num_key_value_heads = num_key_value_heads
|
| 69 |
+
self.hidden_act = hidden_act
|
| 70 |
+
self.initializer_range = initializer_range
|
| 71 |
+
self.rms_norm_eps = rms_norm_eps
|
| 72 |
+
self.use_cache = use_cache
|
| 73 |
+
self.rope_theta = rope_theta
|
| 74 |
+
self.mlp_only_layers = mlp_only_layers
|
| 75 |
+
self.intermediate_size = intermediate_size
|
| 76 |
+
|
| 77 |
+
# MoE arguments
|
| 78 |
+
self.moe_intermediate_size = moe_intermediate_size
|
| 79 |
+
self.shared_expert_intermediate_size = shared_expert_intermediate_size
|
| 80 |
+
self.num_experts_per_tok = num_experts_per_tok
|
| 81 |
+
self.num_experts = num_experts
|
| 82 |
+
self.norm_topk_prob = norm_topk_prob
|
| 83 |
+
self.output_router_logits = output_router_logits
|
| 84 |
+
self.router_enable_expert_bias = router_enable_expert_bias
|
| 85 |
+
self.routed_scaling_factor = routed_scaling_factor
|
| 86 |
+
self.qk_nope_dim = qk_nope_dim
|
| 87 |
+
self.qk_rope_dim = qk_rope_dim
|
| 88 |
+
self.v_channels = v_channels
|
| 89 |
+
self.sandwich_norm = sandwich_norm
|
| 90 |
+
self.param_sink_number = param_sink_number
|
| 91 |
+
self.param_sink_with_value = param_sink_with_value
|
| 92 |
+
|
| 93 |
+
super().__init__(
|
| 94 |
+
tie_word_embeddings=tie_word_embeddings,
|
| 95 |
+
**kwargs,
|
| 96 |
+
)
|
doc/omniinfer_for_openpangu_r_72b_2512.md
ADDED
|
@@ -0,0 +1,118 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# openPangu-R-72B-2512在Omni-Infer部署指导文档
|
| 2 |
+
|
| 3 |
+
## 硬件环境和部署方式
|
| 4 |
+
PD混部，只需要1台Atlas 800T A3机器中的4个die。
|
| 5 |
+
|
| 6 |
+
## 代码和镜像
|
| 7 |
+
- Omni-Infer代码版本：release_v0.7.0
|
| 8 |
+
- 配套镜像：参考 https://gitee.com/omniai/omniinfer/releases 中v0.7.0镜像，以A3硬件和arm架构为例，使用“docker pull swr.cn-east-4.myhuaweicloud.com/omni/omniinfer-a3-arm:release_v0.7.0-vllm”。
|
| 9 |
+
|
| 10 |
+
## 部署
|
| 11 |
+
### 1. 启动镜像
|
| 12 |
+
```bash
|
| 13 |
+
IMAGE=swr.cn-east-4.myhuaweicloud.com/omni/omniinfer-a3-arm:release_v0.7.0-vllm
|
| 14 |
+
NAME=omniinfer-v0.7.0 # Custom docker name
|
| 15 |
+
NPU_NUM=16 # A3节点die数
|
| 16 |
+
DEVICE_ARGS=$(for i in $(seq 0 $((NPU_NUM-1))); do echo -n "--device /dev/davinci${i} "; done)
|
| 17 |
+
|
| 18 |
+
# Run the container using the defined variables
|
| 19 |
+
# Note if you are running bridge network with docker, Please expose available ports for multiple nodes communication in advance
|
| 20 |
+
# To prevent device interference from other docker containers, add the argument "--privileged"
|
| 21 |
+
docker run -itd \
|
| 22 |
+
--name=${NAME} \
|
| 23 |
+
--network host \
|
| 24 |
+
--privileged \
|
| 25 |
+
--ipc=host \
|
| 26 |
+
$DEVICE_ARGS \
|
| 27 |
+
--device=/dev/davinci_manager \
|
| 28 |
+
--device=/dev/devmm_svm \
|
| 29 |
+
--device=/dev/hisi_hdc \
|
| 30 |
+
-v /usr/local/Ascend/driver:/usr/local/Ascend/driver \
|
| 31 |
+
-v /usr/local/Ascend/firmware:/usr/local/Ascend/firmware \
|
| 32 |
+
-v /usr/local/sbin/npu-smi:/usr/local/sbin/npu-smi \
|
| 33 |
+
-v /etc/ascend_install.info:/etc/ascend_install.info \
|
| 34 |
+
-v /mnt/:/mnt/ \
|
| 35 |
+
-v /data:/data \
|
| 36 |
+
-v /home/work:/home/work \
|
| 37 |
+
--entrypoint /bin/bash \
|
| 38 |
+
swr.cn-east-4.myhuaweicloud.com/omni/omniinfer-a3-arm:release_v0.7.0-vllm
|
| 39 |
+
```
|
| 40 |
+
需要保证模型权重和本项目代码可在容器中访问。进入容器:
|
| 41 |
+
```bash
|
| 42 |
+
docker exec -it $NAME /bin/bash
|
| 43 |
+
```
|
| 44 |
+
|
| 45 |
+
### 2. 将examples/start_serving_openpangu_r_72b_2512.sh脚本放入omniinfer/tools/scripts路径并执行
|
| 46 |
+
|
| 47 |
+
```bash
|
| 48 |
+
git clone -b release_v0.7.0 https://gitee.com/omniai/omniinfer.git
|
| 49 |
+
cd omniinfer/tools/scripts
|
| 50 |
+
# 需修改serving脚本中model-path模型路径、master-ip机器IP地址和PYTHONPATH。
|
| 51 |
+
bash start_serving_openpangu_r_72b_2512.sh
|
| 52 |
+
```
|
| 53 |
+
|
| 54 |
+
### 3. 发请求测试
|
| 55 |
+
|
| 56 |
+
服务启动后，可发送测试请求。
|
| 57 |
+
|
| 58 |
+
```bash
|
| 59 |
+
curl http://0.0.0.0:8000/v1/chat/completions \
|
| 60 |
+
-H "Content-Type: application/json" \
|
| 61 |
+
-d '{
|
| 62 |
+
"model": "openpangu_r_72b_2512",
|
| 63 |
+
"messages": [
|
| 64 |
+
{
|
| 65 |
+
"role": "user",
|
| 66 |
+
"content": "Who are you?"
|
| 67 |
+
}
|
| 68 |
+
],
|
| 69 |
+
"temperature": 1.0,
|
| 70 |
+
"top_p": 0.8,
|
| 71 |
+
"top_k": -1,
|
| 72 |
+
"vllm_xargs": {"top_n_sigma": 0.05},
|
| 73 |
+
"chat_template_kwargs": {"think": true, "reasoning_effort": "low"}
|
| 74 |
+
}'
|
| 75 |
+
```
|
| 76 |
+
```bash
|
| 77 |
+
# 工具使用
|
| 78 |
+
curl http://0.0.0.0:8000/v1/chat/completions \
|
| 79 |
+
-H "Content-Type: application/json" \
|
| 80 |
+
-d '{
|
| 81 |
+
"model": "openpangu_r_72b_2512",
|
| 82 |
+
"messages": [
|
| 83 |
+
{"role": "system", "content": "你是华为公司开发的盘古模型。\n现在是2025年7月30日"},
|
| 84 |
+
{"role": "user", "content": "深圳明天的天气如何？"}
|
| 85 |
+
],
|
| 86 |
+
"tools": [
|
| 87 |
+
{
|
| 88 |
+
"type": "function",
|
| 89 |
+
"function": {
|
| 90 |
+
"name": "get_current_weather",
|
| 91 |
+
"description": "获取指定城市的当前天气信息，包括温度、湿度、风速等数据。",
|
| 92 |
+
"parameters": {
|
| 93 |
+
"type": "object",
|
| 94 |
+
"properties": {
|
| 95 |
+
"location": {
|
| 96 |
+
"type": "string",
|
| 97 |
+
"description": "城市名称，例如：北京、深圳。支持中文或拼音输入。"
|
| 98 |
+
},
|
| 99 |
+
"date": {
|
| 100 |
+
"type": "string",
|
| 101 |
+
"description": "查询日期，格式为 YYYY-MM-DD（遵循 ISO 8601 标准）。例如：2023-10-01。"
|
| 102 |
+
}
|
| 103 |
+
},
|
| 104 |
+
"required": ["location", "date"],
|
| 105 |
+
"additionalProperties": "false"
|
| 106 |
+
}
|
| 107 |
+
}
|
| 108 |
+
}
|
| 109 |
+
],
|
| 110 |
+
"temperature": 1.0,
|
| 111 |
+
"top_p": 0.8,
|
| 112 |
+
"top_k": -1,
|
| 113 |
+
"vllm_xargs": {"top_n_sigma": 0.05},
|
| 114 |
+
"chat_template_kwargs": {"think": true, "reasoning_effort": "high"}
|
| 115 |
+
}'
|
| 116 |
+
```
|
| 117 |
+
模型默认是慢思考模式，在慢思考模式下，模型支持思维链分档，可通过请求体字段"chat_template_kwargs": {"think": true, "reasoning_effort": "high"}中"reasoning_effort": "high"和"low"平衡模型精度和效率。
|
| 118 |
+
模型的慢思考模式，可通过请求体字段"chat_template_kwargs": {"think": true/false} 开启和关闭。
|
doc/omniinfer_for_openpangu_r_72b_2512_EN.md
ADDED
|
@@ -0,0 +1,119 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Deployment Guide for openPangu-R-72B-2512 on Omni-Infer
|
| 2 |
+
|
| 3 |
+
## Hardware Environment and Deployment Method
|
| 4 |
+
PD hybrid deployment, requiring only 4 dies of one Atlas 800T A3 machine.
|
| 5 |
+
|
| 6 |
+
## Codes and Image
|
| 7 |
+
- Omni-Infer code version: release_v0.7.0
|
| 8 |
+
- Docker Image: Refer to the v0.7.0 image in https://gitee.com/omniai/omniinfer/releases. For example, for A3 hardware and ARM architecture, use "docker pull swr.cn-east-4.myhuaweicloud.com/omni/omniinfer-a3-arm:release_v0.7.0-vllm".
|
| 9 |
+
|
| 10 |
+
## Deployment
|
| 11 |
+
### 1. Launch the image
|
| 12 |
+
```bash
|
| 13 |
+
IMAGE=swr.cn-east-4.myhuaweicloud.com/omni/omniinfer-a3-arm:release_v0.7.0-vllm
|
| 14 |
+
NAME=omniinfer-v0.7.0 # Custom docker name
|
| 15 |
+
NPU_NUM=16 # 16 dies of A3 node
|
| 16 |
+
DEVICE_ARGS=$(for i in $(seq 0 $((NPU_NUM-1))); do echo -n "--device /dev/davinci${i} "; done)
|
| 17 |
+
|
| 18 |
+
# Run the container using the defined variables
|
| 19 |
+
# Note if you are running bridge network with docker, Please expose available ports for multiple nodes communication in advance
|
| 20 |
+
# To prevent device interference from other docker containers, add the argument "--privileged"
|
| 21 |
+
docker run -itd \
|
| 22 |
+
--name=${NAME} \
|
| 23 |
+
--network host \
|
| 24 |
+
--privileged \
|
| 25 |
+
--ipc=host \
|
| 26 |
+
$DEVICE_ARGS \
|
| 27 |
+
--device=/dev/davinci_manager \
|
| 28 |
+
--device=/dev/devmm_svm \
|
| 29 |
+
--device=/dev/hisi_hdc \
|
| 30 |
+
-v /usr/local/Ascend/driver:/usr/local/Ascend/driver \
|
| 31 |
+
-v /usr/local/Ascend/firmware:/usr/local/Ascend/firmware \
|
| 32 |
+
-v /usr/local/sbin/npu-smi:/usr/local/sbin/npu-smi \
|
| 33 |
+
-v /etc/ascend_install.info:/etc/ascend_install.info \
|
| 34 |
+
-v /mnt/:/mnt/ \
|
| 35 |
+
-v /data:/data \
|
| 36 |
+
-v /home/work:/home/work \
|
| 37 |
+
--entrypoint /bin/bash \
|
| 38 |
+
swr.cn-east-4.myhuaweicloud.com/omni/omniinfer-a3-arm:release_v0.7.0-vllm
|
| 39 |
+
```
|
| 40 |
+
Ensure that the model checkpoint and the project code are accessible within the container. Enter the container:
|
| 41 |
+
```bash
|
| 42 |
+
docker exec -it $NAME /bin/bash
|
| 43 |
+
```
|
| 44 |
+
|
| 45 |
+
### 2. Put examples/start_serving_openpangu_r_72b_2512.sh in the omniinfer/tools/scripts path and start the serving script
|
| 46 |
+
|
| 47 |
+
```bash
|
| 48 |
+
git clone -b release_v0.7.0 https://gitee.com/omniai/omniinfer.git
|
| 49 |
+
cd omniinfer/tools/scripts
|
| 50 |
+
# You need to modify the model-path, master-ip address and PYTHONPATH in the serving script.
|
| 51 |
+
bash start_serving_openpangu_r_72b_2512.sh
|
| 52 |
+
```
|
| 53 |
+
|
| 54 |
+
### 3. Send Testing Requests
|
| 55 |
+
|
| 56 |
+
After the service is started, we can send testing requests.
|
| 57 |
+
|
| 58 |
+
```bash
|
| 59 |
+
curl http://0.0.0.0:8000/v1/chat/completions \
|
| 60 |
+
-H "Content-Type: application/json" \
|
| 61 |
+
-d '{
|
| 62 |
+
"model": "openpangu_r_72b_2512",
|
| 63 |
+
"messages": [
|
| 64 |
+
{
|
| 65 |
+
"role": "user",
|
| 66 |
+
"content": "Who are you?"
|
| 67 |
+
}
|
| 68 |
+
],
|
| 69 |
+
"temperature": 1.0,
|
| 70 |
+
"top_p": 0.8,
|
| 71 |
+
"top_k": -1,
|
| 72 |
+
"vllm_xargs": {"top_n_sigma": 0.05},
|
| 73 |
+
"chat_template_kwargs": {"think": true, "reasoning_effort": "low"}
|
| 74 |
+
}'
|
| 75 |
+
```
|
| 76 |
+
```bash
|
| 77 |
+
# Tool use
|
| 78 |
+
curl http://0.0.0.0:8000/v1/chat/completions \
|
| 79 |
+
-H "Content-Type: application/json" \
|
| 80 |
+
-d '{
|
| 81 |
+
"model": "openpangu_r_72b_2512",
|
| 82 |
+
"messages": [
|
| 83 |
+
{"role": "system", "content": "你是华为公司开发的盘古模型。\n现在是2025年7月30日"},
|
| 84 |
+
{"role": "user", "content": "深圳明天的天气如何？"}
|
| 85 |
+
],
|
| 86 |
+
"tools": [
|
| 87 |
+
{
|
| 88 |
+
"type": "function",
|
| 89 |
+
"function": {
|
| 90 |
+
"name": "get_current_weather",
|
| 91 |
+
"description": "获取指定城市的当前天气信息，包括温度、湿度、风速等数据。",
|
| 92 |
+
"parameters": {
|
| 93 |
+
"type": "object",
|
| 94 |
+
"properties": {
|
| 95 |
+
"location": {
|
| 96 |
+
"type": "string",
|
| 97 |
+
"description": "城市名称，例如：北京、深圳。支持中文或拼音输入。"
|
| 98 |
+
},
|
| 99 |
+
"date": {
|
| 100 |
+
"type": "string",
|
| 101 |
+
"description": "查询日期，格式为 YYYY-MM-DD（遵循 ISO 8601 标准）。例如：2023-10-01。"
|
| 102 |
+
}
|
| 103 |
+
},
|
| 104 |
+
"required": ["location", "date"],
|
| 105 |
+
"additionalProperties": "false"
|
| 106 |
+
}
|
| 107 |
+
}
|
| 108 |
+
}
|
| 109 |
+
],
|
| 110 |
+
"temperature": 1.0,
|
| 111 |
+
"top_p": 0.8,
|
| 112 |
+
"top_k": -1,
|
| 113 |
+
"vllm_xargs": {"top_n_sigma": 0.05},
|
| 114 |
+
"chat_template_kwargs": {"think": true, "reasoning_effort": "high"}
|
| 115 |
+
}'
|
| 116 |
+
```
|
| 117 |
+
|
| 118 |
+
The model is in slow-thinking mode by default. In slow-thinking mode, you can specify different reasoning effort by setting the "reasoning_effort" parameter in "chat_template_kwargs" to "high" or "low" to balance model accuracy and efficiency.
|
| 119 |
+
openPangu-R-72B-2512 supports switching between slow-thinking and fast-thinking mode by setting {"think": true/false} in "chat_template_kwargs".
|
examples/start_serving_openpangu_r_72b_2512.sh
ADDED
|
@@ -0,0 +1,57 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Copyright (c) 2025 Huawei Technologies Co., Ltd. All rights reserved.
|
| 2 |
+
|
| 3 |
+
export ASCEND_RT_VISIBLE_DEVICES=0,1,2,3
|
| 4 |
+
export VLLM_USE_V1=1
|
| 5 |
+
export VLLM_WORKER_MULTIPROC_METHOD=fork
|
| 6 |
+
export VLLM_ENABLE_MC2=0
|
| 7 |
+
export USING_LCCL_COM=0
|
| 8 |
+
|
| 9 |
+
export OMNI_USE_PANGU=1
|
| 10 |
+
export ENABLE_PREFILL_TND=1
|
| 11 |
+
|
| 12 |
+
export HCCL_OP_EXPANSION_MODE="AIV"
|
| 13 |
+
export VLLM_ALLOW_LONG_MAX_MODEL_LEN=1
|
| 14 |
+
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
|
| 15 |
+
|
| 16 |
+
export HCCL_RDMA_TIMEOUT=5
|
| 17 |
+
export HCCL_DETERMINISTIC=False
|
| 18 |
+
export ASCEND_GLOBAL_LOG_LEVEL=3
|
| 19 |
+
export CPU_AFFINITY_CONF=2
|
| 20 |
+
export VLLM_LOGGING_LEVEL=INFO
|
| 21 |
+
|
| 22 |
+
export HCCL_BUFFSIZE=1000
|
| 23 |
+
export HCCL_CONNECT_TIMEOUT=1800
|
| 24 |
+
export HCCL_EXEC_TIMEOUT=1800
|
| 25 |
+
export HCCL_INTRA_ROCE_ENABLE=1
|
| 26 |
+
export HCCL_INTRA_PCIE_ENABLE=0
|
| 27 |
+
export FORCE_ENABLE_CHUNK_PREFILL=1
|
| 28 |
+
|
| 29 |
+
export USE_REASONING=${USE_REASONING:=1}
|
| 30 |
+
export USE_TOOL=${USE_TOOL:=1}
|
| 31 |
+
if [ "$USE_REASONING" = "1" ]; then
|
| 32 |
+
reasoning="--reasoning-parser pangu"
|
| 33 |
+
fi
|
| 34 |
+
if [ "$USE_TOOL" = "1" ]; then
|
| 35 |
+
tools="--enable-auto-tool-choice --tool-call-parser pangu"
|
| 36 |
+
fi
|
| 37 |
+
|
| 38 |
+
export PYTHONPATH=/path/to/omniinfer/:$PYTHONPATH
|
| 39 |
+
rm -r -f .torchair_cache/
|
| 40 |
+
|
| 41 |
+
python start_api_servers.py \
|
| 42 |
+
--num-servers 1 \
|
| 43 |
+
--model-path /path/to/model/ \
|
| 44 |
+
--master-ip 0.0.0.0 \
|
| 45 |
+
--tp 4 \
|
| 46 |
+
--num-dp 1 \
|
| 47 |
+
--master-port 3512 \
|
| 48 |
+
--served-model-name openpangu_r_72b_2512 \
|
| 49 |
+
--log-dir apiserverlog_pangu72B_hybrid_chunk \
|
| 50 |
+
--extra-args "--max-num-batched-tokens 2048 --enforce-eager --no-enable-prefix-caching --enable-expert-parallel --max-num-seqs 32 --long-prefill-token-threshold 1024" \
|
| 51 |
+
--base-api-port 8000 \
|
| 52 |
+
--gpu-util 0.90 \
|
| 53 |
+
--no-enable-prefix-caching \
|
| 54 |
+
--max-model-len 131072 \
|
| 55 |
+
$reasoning \
|
| 56 |
+
$tools \
|
| 57 |
+
--additional-config '{"graph_model_compile_config":{"level":1, "use_ge_graph_cached":true, "decode_gear_list": [32]}, "enable_hybrid_graph_mode": false, "expert_parallel_size": 4, "expert_tensor_parallel_size": 1}' &
|
generation_config.json
ADDED
|
@@ -0,0 +1,11 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"bos_token_id": 0,
|
| 4 |
+
"eos_token_id": 45892,
|
| 5 |
+
"do_sample": true,
|
| 6 |
+
"temperature": 1.0,
|
| 7 |
+
"top_p": 0.8,
|
| 8 |
+
"top_n_sigma": 0.05,
|
| 9 |
+
"top_k": -1,
|
| 10 |
+
"transformers_version": "4.48.2"
|
| 11 |
+
}
|
model-00001-of-000051.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:faea9c15e58366768bd204ef27f631f1b91e258b1fdfbebb1eea3f481600dd06
|
| 3 |
+
size 1899599744
|
model-00002-of-000051.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4f92258b47acb3e71c7072e741c117617d49f5d4c4ffcce559d971b9638c478f
|
| 3 |
+
size 484022000
|
model-00003-of-000051.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:96a77e9c991a6f78b03edada807dff37ba8d6083c7ff4d3afd77d83dad3cd77f
|
| 3 |
+
size 484022000
|
model-00004-of-000051.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c0195156ab96017be8d878d9aec1c55c2ed522180fce949ddf30dc7f028b127b
|
| 3 |
+
size 484022000
|
model-00005-of-000051.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1f1ae6f3c3bab2c5913505284600cddafba09a887a61496e1bb7c24b56f77b66
|
| 3 |
+
size 3103607896
|
model-00006-of-000051.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fb7d09f5bd129ec8eac37d63644c3fdd52b89c6de128712865078b3f6da3a5e1
|
| 3 |
+
size 3103607896
|
model-00007-of-000051.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9a3346f29473783c660dd4c4af0e7d955cbef3c474e64a751a04bed38665e6dc
|
| 3 |
+
size 3103607896
|
model-00008-of-000051.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:52fd330d2faa49e6bca4eea2d8c78489ce6e298e55b9b9fbaff9e84eb1b7c2ce
|
| 3 |
+
size 3103607896
|
model-00009-of-000051.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9c59e8651cd10ebe8debe65fc620e1ab68efab24018a8cb7e2e0d2520b929a98
|
| 3 |
+
size 3103607896
|
model-00010-of-000051.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:348a826c4153084ed0195a2989083772ba28009cb8d2479a501bd9e3e5e26686
|
| 3 |
+
size 3103607896
|
model-00011-of-000051.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f3f60733b0617fdc21159ec975182a44a727ab292dd9d58d9ac7b400075b87a9
|
| 3 |
+
size 3103608152
|
model-00012-of-000051.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5536de41c6b14feb5df47b3be85e70a89337cf2480933eb947e1a1dc3d447fc7
|
| 3 |
+
size 3103608152
|
model-00013-of-000051.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f02a6449935c211e5a4347c73b22763f127e78f74aa736fa5b1abbeb0aaa1c87
|
| 3 |
+
size 3103608152
|
model-00014-of-000051.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6f32dd5f9f4414ea611ee0bf5ed343b7604785194a8c50e4c8f57cbec41f8d68
|
| 3 |
+
size 3103608152
|
model-00015-of-000051.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e4b7666e445966de147f4860a397f93f4b5c7635cbd469ac3c442af7ba939962
|
| 3 |
+
size 3103608152
|
model-00016-of-000051.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0ecb4ad21311887f4c44eeb94b0cfd502821db1dc1d12611b5b652a8a84d4fbe
|
| 3 |
+
size 3103608152
|
model-00017-of-000051.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:db6908a54fdb8640a406a78c6fb68c035f9cd1982bf4a9853e102941a4f50a2b
|
| 3 |
+
size 3103608152
|
model-00018-of-000051.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:10188bcb0c9ce4bd8aa8b4084ccd2bb4bb830f95507f5032fc6ac22168d676be
|
| 3 |
+
size 3103608152
|
model-00019-of-000051.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b30d5babd023187eef88826a3b505a06e586fc458b7a2c5da481473f6210e889
|
| 3 |
+
size 3103608152
|
model-00020-of-000051.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:37bf95087b8495dd8065020ccc550d73bfbc99b12419c54dc902aeba8db9d1e6
|
| 3 |
+
size 3103608152
|
model-00021-of-000051.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8741ca6de1c6f1c5c2788023d9a8db0c82b503ec9fe49ba076f07d90c5de92ba
|
| 3 |
+
size 3103608152
|
model-00022-of-000051.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:12a2c41a55ce55e98be091bc4c2333e4a7e69dee5d79fc222fd27ab30e3ce5cb
|
| 3 |
+
size 3103608152
|
model-00023-of-000051.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2b6b91c53ce720c03347dad507730b16ecb9eb9d2bd6c468a91150cd9c23cd30
|
| 3 |
+
size 3103608152
|
model-00024-of-000051.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4d441eaad36da0df2dc9825c6909b7b29d907cfd00089af1a007c45632e5e115
|
| 3 |
+
size 3103608152
|
model-00025-of-000051.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6e8f54dbb2414b22a883c110920b7bd84771a2e2dbb78a85b759c7ece195bc6c
|
| 3 |
+
size 3103608152
|
model-00026-of-000051.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1cf8a05cb3853131aaff5b1a2357cdd4462cde2d9e57ad35a5779c0f90df361e
|
| 3 |
+
size 3103608152
|
model-00027-of-000051.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ad286b9f5200c6de8a97c0eecccafafbbea7a710716abd0bad30dfc768b0650c
|
| 3 |
+
size 3103608152
|
model-00028-of-000051.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:291d66cb2467c36e2642250c40445fdca3cadbfbbf784b66fbbd6356aa3a1a18
|
| 3 |
+
size 3103608152
|
model-00029-of-000051.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f01876779fe5f926a1f4ad070e5cfbdb061dcd424272c35d3e5048254355fa13
|
| 3 |
+
size 3103608152
|
model-00030-of-000051.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:150c6e3f25cc165d23a8af45ee8859324267fb85825093472421849a2c994bcb
|
| 3 |
+
size 3103608152
|
model-00031-of-000051.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f83613272553ca7d84303dc88527ee273ba90244e6a8bda34df0b4ba2be337ed
|
| 3 |
+
size 3103608152
|
model-00032-of-000051.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0bb600e04ab0220ad2cc50e7bb8025db75c4be9fc9478f356e393d11231ac5cb
|
| 3 |
+
size 3103608152
|
model-00033-of-000051.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9b15a5bb9a59506a6ebb7224135be2c2f4816d0cbd16c096404d44dfed953712
|
| 3 |
+
size 3103608152
|
model-00034-of-000051.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9c87358d675516d50057fd7baf809508b14a69751805468ef7edd21f9b3b270b
|
| 3 |
+
size 3103608152
|
model-00035-of-000051.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:751356ba1cb106cb53f3416dcf1fad91cf799af400b9a4afa30c7a5b227b8839
|
| 3 |
+
size 3103608152
|
model-00036-of-000051.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:46b580d5e2bc27651b57b1b310bbd8c053f4da6c044f756fb5db78368b1fcfa1
|
| 3 |
+
size 3103608152
|
model-00037-of-000051.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e1e133139c8e3a80a675625abb4aebed19621c3aa7e5db072850441f2bb9a1c9
|
| 3 |
+
size 3103608152
|
model-00038-of-000051.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:206234a10d3ea746595667909a6e6e0dc728d1de719f5bd126b1e54785603bdc
|
| 3 |
+
size 3103608152
|
model-00039-of-000051.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:add0f4a5dd9bc63f900f604587c8b00f1f089dc5a8a6b39d751abf98ba11741d
|
| 3 |
+
size 3103608152
|