add special_tokens_map.json
#6
by
Baicai003
- opened
- LICENSE +1 -1
- NOTICE +1 -229
- README.md +224 -108
- assets/logo.jpg +0 -0
- assets/qwen_tokenizer.png +0 -0
- assets/tokenizer.png +0 -0
- assets/wechat.png +0 -0
- cache_autogptq_cuda_256.cpp +0 -198
- cache_autogptq_cuda_kernel_256.cu +0 -1708
- config.json +23 -14
- configuration_qwen.py +36 -29
- cpp_kernels.py +0 -55
- generation_config.json +7 -2
- model-00002-of-00008.safetensors +0 -3
- model-00003-of-00008.safetensors +0 -3
- model-00004-of-00008.safetensors +0 -3
- model-00005-of-00008.safetensors +0 -3
- model-00006-of-00008.safetensors +0 -3
- model-00007-of-00008.safetensors +0 -3
- model-00008-of-00008.safetensors +0 -3
- model.safetensors.index.json +0 -266
- modeling_qwen.py +334 -557
- model-00001-of-00008.safetensors → pytorch_model.bin +2 -2
- qwen_generation_utils.py +3 -8
- special_tokens_map.json +30 -0
- tokenization_qwen.py +16 -66
- tokenizer_config.json +1 -1
LICENSE
CHANGED
|
@@ -9,7 +9,7 @@ By clicking to agree or by using or distributing any portion or element of the T
|
|
| 9 |
b. "We"(or "Us") shall mean Alibaba Cloud.
|
| 10 |
c. "You" (or "Your") shall mean a natural person or legal entity exercising the rights granted by this Agreement and/or using the Materials for any purpose and in any field of use.
|
| 11 |
d. "Third Parties" shall mean individuals or legal entities that are not under common control with Us or You.
|
| 12 |
-
e. "Tongyi Qianwen" shall mean the large language models (including Qwen model and Qwen-Chat model), and software and algorithms, consisting of trained model weights, parameters (including optimizer states), machine-learning model code, inference-enabling code, training-enabling code, fine-tuning enabling code and other elements of the foregoing distributed by Us.
|
| 13 |
f. "Materials" shall mean, collectively, Alibaba Cloud's proprietary Tongyi Qianwen and Documentation (and any portion thereof) made available under this Agreement.
|
| 14 |
g. "Source" form shall mean the preferred form for making modifications, including but not limited to model source code, documentation source, and configuration files.
|
| 15 |
h. "Object" form shall mean any form resulting from mechanical transformation or translation of a Source form, including but not limited to compiled object code, generated documentation,
|
|
|
|
| 9 |
b. "We"(or "Us") shall mean Alibaba Cloud.
|
| 10 |
c. "You" (or "Your") shall mean a natural person or legal entity exercising the rights granted by this Agreement and/or using the Materials for any purpose and in any field of use.
|
| 11 |
d. "Third Parties" shall mean individuals or legal entities that are not under common control with Us or You.
|
| 12 |
+
e. "Tongyi Qianwen" shall mean the large language models (including Qwen-7B model and Qwen-7B-Chat model), and software and algorithms, consisting of trained model weights, parameters (including optimizer states), machine-learning model code, inference-enabling code, training-enabling code, fine-tuning enabling code and other elements of the foregoing distributed by Us.
|
| 13 |
f. "Materials" shall mean, collectively, Alibaba Cloud's proprietary Tongyi Qianwen and Documentation (and any portion thereof) made available under this Agreement.
|
| 14 |
g. "Source" form shall mean the preferred form for making modifications, including but not limited to model source code, documentation source, and configuration files.
|
| 15 |
h. "Object" form shall mean any form resulting from mechanical transformation or translation of a Source form, including but not limited to compiled object code, generated documentation,
|
NOTICE
CHANGED
|
@@ -49,232 +49,4 @@ FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
|
|
| 49 |
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
| 50 |
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
| 51 |
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
|
| 52 |
-
SOFTWARE.
|
| 53 |
-
|
| 54 |
-
------------- LICENSE FOR stanford_alpaca code --------------
|
| 55 |
-
|
| 56 |
-
Apache License
|
| 57 |
-
Version 2.0, January 2004
|
| 58 |
-
http://www.apache.org/licenses/
|
| 59 |
-
|
| 60 |
-
TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
|
| 61 |
-
|
| 62 |
-
1. Definitions.
|
| 63 |
-
|
| 64 |
-
"License" shall mean the terms and conditions for use, reproduction,
|
| 65 |
-
and distribution as defined by Sections 1 through 9 of this document.
|
| 66 |
-
|
| 67 |
-
"Licensor" shall mean the copyright owner or entity authorized by
|
| 68 |
-
the copyright owner that is granting the License.
|
| 69 |
-
|
| 70 |
-
"Legal Entity" shall mean the union of the acting entity and all
|
| 71 |
-
other entities that control, are controlled by, or are under common
|
| 72 |
-
control with that entity. For the purposes of this definition,
|
| 73 |
-
"control" means (i) the power, direct or indirect, to cause the
|
| 74 |
-
direction or management of such entity, whether by contract or
|
| 75 |
-
otherwise, or (ii) ownership of fifty percent (50%) or more of the
|
| 76 |
-
outstanding shares, or (iii) beneficial ownership of such entity.
|
| 77 |
-
|
| 78 |
-
"You" (or "Your") shall mean an individual or Legal Entity
|
| 79 |
-
exercising permissions granted by this License.
|
| 80 |
-
|
| 81 |
-
"Source" form shall mean the preferred form for making modifications,
|
| 82 |
-
including but not limited to software source code, documentation
|
| 83 |
-
source, and configuration files.
|
| 84 |
-
|
| 85 |
-
"Object" form shall mean any form resulting from mechanical
|
| 86 |
-
transformation or translation of a Source form, including but
|
| 87 |
-
not limited to compiled object code, generated documentation,
|
| 88 |
-
and conversions to other media types.
|
| 89 |
-
|
| 90 |
-
"Work" shall mean the work of authorship, whether in Source or
|
| 91 |
-
Object form, made available under the License, as indicated by a
|
| 92 |
-
copyright notice that is included in or attached to the work
|
| 93 |
-
(an example is provided in the Appendix below).
|
| 94 |
-
|
| 95 |
-
"Derivative Works" shall mean any work, whether in Source or Object
|
| 96 |
-
form, that is based on (or derived from) the Work and for which the
|
| 97 |
-
editorial revisions, annotations, elaborations, or other modifications
|
| 98 |
-
represent, as a whole, an original work of authorship. For the purposes
|
| 99 |
-
of this License, Derivative Works shall not include works that remain
|
| 100 |
-
separable from, or merely link (or bind by name) to the interfaces of,
|
| 101 |
-
the Work and Derivative Works thereof.
|
| 102 |
-
|
| 103 |
-
"Contribution" shall mean any work of authorship, including
|
| 104 |
-
the original version of the Work and any modifications or additions
|
| 105 |
-
to that Work or Derivative Works thereof, that is intentionally
|
| 106 |
-
submitted to Licensor for inclusion in the Work by the copyright owner
|
| 107 |
-
or by an individual or Legal Entity authorized to submit on behalf of
|
| 108 |
-
the copyright owner. For the purposes of this definition, "submitted"
|
| 109 |
-
means any form of electronic, verbal, or written communication sent
|
| 110 |
-
to the Licensor or its representatives, including but not limited to
|
| 111 |
-
communication on electronic mailing lists, source code control systems,
|
| 112 |
-
and issue tracking systems that are managed by, or on behalf of, the
|
| 113 |
-
Licensor for the purpose of discussing and improving the Work, but
|
| 114 |
-
excluding communication that is conspicuously marked or otherwise
|
| 115 |
-
designated in writing by the copyright owner as "Not a Contribution."
|
| 116 |
-
|
| 117 |
-
"Contributor" shall mean Licensor and any individual or Legal Entity
|
| 118 |
-
on behalf of whom a Contribution has been received by Licensor and
|
| 119 |
-
subsequently incorporated within the Work.
|
| 120 |
-
|
| 121 |
-
2. Grant of Copyright License. Subject to the terms and conditions of
|
| 122 |
-
this License, each Contributor hereby grants to You a perpetual,
|
| 123 |
-
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
|
| 124 |
-
copyright license to reproduce, prepare Derivative Works of,
|
| 125 |
-
publicly display, publicly perform, sublicense, and distribute the
|
| 126 |
-
Work and such Derivative Works in Source or Object form.
|
| 127 |
-
|
| 128 |
-
3. Grant of Patent License. Subject to the terms and conditions of
|
| 129 |
-
this License, each Contributor hereby grants to You a perpetual,
|
| 130 |
-
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
|
| 131 |
-
(except as stated in this section) patent license to make, have made,
|
| 132 |
-
use, offer to sell, sell, import, and otherwise transfer the Work,
|
| 133 |
-
where such license applies only to those patent claims licensable
|
| 134 |
-
by such Contributor that are necessarily infringed by their
|
| 135 |
-
Contribution(s) alone or by combination of their Contribution(s)
|
| 136 |
-
with the Work to which such Contribution(s) was submitted. If You
|
| 137 |
-
institute patent litigation against any entity (including a
|
| 138 |
-
cross-claim or counterclaim in a lawsuit) alleging that the Work
|
| 139 |
-
or a Contribution incorporated within the Work constitutes direct
|
| 140 |
-
or contributory patent infringement, then any patent licenses
|
| 141 |
-
granted to You under this License for that Work shall terminate
|
| 142 |
-
as of the date such litigation is filed.
|
| 143 |
-
|
| 144 |
-
4. Redistribution. You may reproduce and distribute copies of the
|
| 145 |
-
Work or Derivative Works thereof in any medium, with or without
|
| 146 |
-
modifications, and in Source or Object form, provided that You
|
| 147 |
-
meet the following conditions:
|
| 148 |
-
|
| 149 |
-
(a) You must give any other recipients of the Work or
|
| 150 |
-
Derivative Works a copy of this License; and
|
| 151 |
-
|
| 152 |
-
(b) You must cause any modified files to carry prominent notices
|
| 153 |
-
stating that You changed the files; and
|
| 154 |
-
|
| 155 |
-
(c) You must retain, in the Source form of any Derivative Works
|
| 156 |
-
that You distribute, all copyright, patent, trademark, and
|
| 157 |
-
attribution notices from the Source form of the Work,
|
| 158 |
-
excluding those notices that do not pertain to any part of
|
| 159 |
-
the Derivative Works; and
|
| 160 |
-
|
| 161 |
-
(d) If the Work includes a "NOTICE" text file as part of its
|
| 162 |
-
distribution, then any Derivative Works that You distribute must
|
| 163 |
-
include a readable copy of the attribution notices contained
|
| 164 |
-
within such NOTICE file, excluding those notices that do not
|
| 165 |
-
pertain to any part of the Derivative Works, in at least one
|
| 166 |
-
of the following places: within a NOTICE text file distributed
|
| 167 |
-
as part of the Derivative Works; within the Source form or
|
| 168 |
-
documentation, if provided along with the Derivative Works; or,
|
| 169 |
-
within a display generated by the Derivative Works, if and
|
| 170 |
-
wherever such third-party notices normally appear. The contents
|
| 171 |
-
of the NOTICE file are for informational purposes only and
|
| 172 |
-
do not modify the License. You may add Your own attribution
|
| 173 |
-
notices within Derivative Works that You distribute, alongside
|
| 174 |
-
or as an addendum to the NOTICE text from the Work, provided
|
| 175 |
-
that such additional attribution notices cannot be construed
|
| 176 |
-
as modifying the License.
|
| 177 |
-
|
| 178 |
-
You may add Your own copyright statement to Your modifications and
|
| 179 |
-
may provide additional or different license terms and conditions
|
| 180 |
-
for use, reproduction, or distribution of Your modifications, or
|
| 181 |
-
for any such Derivative Works as a whole, provided Your use,
|
| 182 |
-
reproduction, and distribution of the Work otherwise complies with
|
| 183 |
-
the conditions stated in this License.
|
| 184 |
-
|
| 185 |
-
5. Submission of Contributions. Unless You explicitly state otherwise,
|
| 186 |
-
any Contribution intentionally submitted for inclusion in the Work
|
| 187 |
-
by You to the Licensor shall be under the terms and conditions of
|
| 188 |
-
this License, without any additional terms or conditions.
|
| 189 |
-
Notwithstanding the above, nothing herein shall supersede or modify
|
| 190 |
-
the terms of any separate license agreement you may have executed
|
| 191 |
-
with Licensor regarding such Contributions.
|
| 192 |
-
|
| 193 |
-
6. Trademarks. This License does not grant permission to use the trade
|
| 194 |
-
names, trademarks, service marks, or product names of the Licensor,
|
| 195 |
-
except as required for reasonable and customary use in describing the
|
| 196 |
-
origin of the Work and reproducing the content of the NOTICE file.
|
| 197 |
-
|
| 198 |
-
7. Disclaimer of Warranty. Unless required by applicable law or
|
| 199 |
-
agreed to in writing, Licensor provides the Work (and each
|
| 200 |
-
Contributor provides its Contributions) on an "AS IS" BASIS,
|
| 201 |
-
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
|
| 202 |
-
implied, including, without limitation, any warranties or conditions
|
| 203 |
-
of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
|
| 204 |
-
PARTICULAR PURPOSE. You are solely responsible for determining the
|
| 205 |
-
appropriateness of using or redistributing the Work and assume any
|
| 206 |
-
risks associated with Your exercise of permissions under this License.
|
| 207 |
-
|
| 208 |
-
8. Limitation of Liability. In no event and under no legal theory,
|
| 209 |
-
whether in tort (including negligence), contract, or otherwise,
|
| 210 |
-
unless required by applicable law (such as deliberate and grossly
|
| 211 |
-
negligent acts) or agreed to in writing, shall any Contributor be
|
| 212 |
-
liable to You for damages, including any direct, indirect, special,
|
| 213 |
-
incidental, or consequential damages of any character arising as a
|
| 214 |
-
result of this License or out of the use or inability to use the
|
| 215 |
-
Work (including but not limited to damages for loss of goodwill,
|
| 216 |
-
work stoppage, computer failure or malfunction, or any and all
|
| 217 |
-
other commercial damages or losses), even if such Contributor
|
| 218 |
-
has been advised of the possibility of such damages.
|
| 219 |
-
|
| 220 |
-
9. Accepting Warranty or Additional Liability. While redistributing
|
| 221 |
-
the Work or Derivative Works thereof, You may choose to offer,
|
| 222 |
-
and charge a fee for, acceptance of support, warranty, indemnity,
|
| 223 |
-
or other liability obligations and/or rights consistent with this
|
| 224 |
-
License. However, in accepting such obligations, You may act only
|
| 225 |
-
on Your own behalf and on Your sole responsibility, not on behalf
|
| 226 |
-
of any other Contributor, and only if You agree to indemnify,
|
| 227 |
-
defend, and hold each Contributor harmless for any liability
|
| 228 |
-
incurred by, or claims asserted against, such Contributor by reason
|
| 229 |
-
of your accepting any such warranty or additional liability.
|
| 230 |
-
|
| 231 |
-
END OF TERMS AND CONDITIONS
|
| 232 |
-
|
| 233 |
-
APPENDIX: How to apply the Apache License to your work.
|
| 234 |
-
|
| 235 |
-
To apply the Apache License to your work, attach the following
|
| 236 |
-
boilerplate notice, with the fields enclosed by brackets "[]"
|
| 237 |
-
replaced with your own identifying information. (Don't include
|
| 238 |
-
the brackets!) The text should be enclosed in the appropriate
|
| 239 |
-
comment syntax for the file format. We also recommend that a
|
| 240 |
-
file or class name and description of purpose be included on the
|
| 241 |
-
same "printed page" as the copyright notice for easier
|
| 242 |
-
identification within third-party archives.
|
| 243 |
-
|
| 244 |
-
Copyright 2023 Rohan Taori, Ishaan Gulrajani, Tianyi Zhang, Yann Dubois, Xuechen Li
|
| 245 |
-
|
| 246 |
-
Licensed under the Apache License, Version 2.0 (the "License");
|
| 247 |
-
you may not use this file except in compliance with the License.
|
| 248 |
-
You may obtain a copy of the License at
|
| 249 |
-
|
| 250 |
-
http://www.apache.org/licenses/LICENSE-2.0
|
| 251 |
-
|
| 252 |
-
Unless required by applicable law or agreed to in writing, software
|
| 253 |
-
distributed under the License is distributed on an "AS IS" BASIS,
|
| 254 |
-
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
| 255 |
-
See the License for the specific language governing permissions and
|
| 256 |
-
limitations under the License.
|
| 257 |
-
|
| 258 |
-
------------- LICENSE FOR PanQiWei AutoGPTQ code --------------
|
| 259 |
-
|
| 260 |
-
MIT License
|
| 261 |
-
|
| 262 |
-
Copyright (c) 2023 潘其威(William)
|
| 263 |
-
|
| 264 |
-
Permission is hereby granted, free of charge, to any person obtaining a copy
|
| 265 |
-
of this software and associated documentation files (the "Software"), to deal
|
| 266 |
-
in the Software without restriction, including without limitation the rights
|
| 267 |
-
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
|
| 268 |
-
copies of the Software, and to permit persons to whom the Software is
|
| 269 |
-
furnished to do so, subject to the following conditions:
|
| 270 |
-
|
| 271 |
-
The above copyright notice and this permission notice shall be included in all
|
| 272 |
-
copies or substantial portions of the Software.
|
| 273 |
-
|
| 274 |
-
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
| 275 |
-
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
| 276 |
-
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
|
| 277 |
-
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
| 278 |
-
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
| 279 |
-
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
|
| 280 |
-
SOFTWARE.
|
|
|
|
| 49 |
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
| 50 |
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
| 51 |
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
|
| 52 |
+
SOFTWARE.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
README.md
CHANGED
|
@@ -6,57 +6,52 @@ tags:
|
|
| 6 |
- qwen
|
| 7 |
pipeline_tag: text-generation
|
| 8 |
inference: false
|
| 9 |
-
license: other
|
| 10 |
-
license_name: tongyi-qianwen-license-agreement
|
| 11 |
-
license_link: https://github.com/QwenLM/Qwen/blob/main/Tongyi%20Qianwen%20LICENSE%20AGREEMENT
|
| 12 |
---
|
| 13 |
|
| 14 |
# Qwen-7B
|
| 15 |
|
| 16 |
<p align="center">
|
| 17 |
-
<img src="https://qianwen-res.oss-cn-beijing.aliyuncs.com/
|
| 18 |
<p>
|
| 19 |
<br>
|
| 20 |
|
| 21 |
<p align="center">
|
| 22 |
-
|
| 23 |
-
<br>
|
| 24 |
-
<a href="https://github.com/QwenLM/Qwen/blob/main/assets/wechat.png">WeChat (微信)</a>   |   <a href="https://discord.gg/z3GAxXZ9Ce">Discord</a>   ｜   <a href="https://dashscope.aliyun.com">API</a>
|
| 25 |
</p>
|
| 26 |
<br>
|
| 27 |
|
| 28 |
## 介绍 (Introduction)
|
| 29 |
|
| 30 |
-
**通义千问-7B（Qwen-7B）**是阿里云研发的通义千问大模型系列的70亿参数规模的模型。Qwen-7B是基于Transformer的大语言模型, 在超大规模的预训练数据上进行训练得到。预训练数据类型多样，覆盖广泛，包括大量网络文本、专业书籍、代码等。同时，在Qwen-7B的基础上，我们使用对齐机制打造了基于大语言模型的AI助手Qwen-7B-Chat
|
| 31 |
|
| 32 |
通义千问-7B（Qwen-7B）主要有以下特点：
|
| 33 |
|
| 34 |
-
1. **大规模高质量训练语料**：使用超过2.
|
| 35 |
2. **强大的性能**：Qwen-7B在多个中英文下游评测任务上（涵盖常识推理、代码、数学、翻译等），效果显著超越现有的相近规模开源模型，甚至在部分指标上相比更大尺寸模型也有较强竞争力。具体评测结果请详见下文。
|
| 36 |
3. **覆盖更全面的词表**：相比目前以中英词表为主的开源模型，Qwen-7B使用了约15万大小的词表。该词表对多语言更加友好，方便用户在不扩展词表的情况下对部分语种进行能力增强和扩展。
|
| 37 |
|
| 38 |
-
如果您想了解更多关于通义千问7B开源模型的细节，我们建议您参阅[
|
| 39 |
|
| 40 |
-
**Qwen-7B** is the 7B-parameter version of the large language model series, Qwen (abbr. Tongyi Qianwen), proposed by
|
| 41 |
|
| 42 |
The features of Qwen-7B include:
|
| 43 |
|
| 44 |
-
1. **Large-scale high-quality training corpora**: It is pretrained on over 2.
|
| 45 |
2. **Competitive performance**: It significantly surpasses existing open-source models of similar scale on multiple Chinese and English downstream evaluation tasks (including commonsense, reasoning, code, mathematics, etc.), and even surpasses some larger-scale models in several benchmarks. See below for specific evaluation results.
|
| 46 |
3. **More comprehensive vocabulary coverage**: Compared with other open-source models based on Chinese and English vocabularies, Qwen-7B uses a vocabulary of over 150K tokens. This vocabulary is more friendly to multiple languages, enabling users to directly further enhance the capability for certain languages without expanding the vocabulary.
|
| 47 |
|
| 48 |
-
For more details about Qwen, please refer to the [
|
| 49 |
-
<br>
|
| 50 |
|
| 51 |
## 要求（Requirements）
|
| 52 |
|
| 53 |
* python 3.8及以上版本
|
| 54 |
* pytorch 1.12及以上版本，推荐2.0及以上版本
|
| 55 |
* 建议使用CUDA 11.4及以上（GPU用户、flash-attention用户等需考虑此选项）
|
|
|
|
|
|
|
| 56 |
* python 3.8 and above
|
| 57 |
* pytorch 1.12 and above, 2.0 and above are recommended
|
| 58 |
* CUDA 11.4 and above are recommended (this is for GPU users, flash-attention users, etc.)
|
| 59 |
-
<br>
|
| 60 |
|
| 61 |
## 依赖项 (Dependency)
|
| 62 |
|
|
@@ -65,21 +60,19 @@ For more details about Qwen, please refer to the [GitHub](https://github.com/Qwe
|
|
| 65 |
To run Qwen-7B, please make sure you meet the above requirements, and then execute the following pip commands to install the dependent libraries.
|
| 66 |
|
| 67 |
```bash
|
| 68 |
-
pip install transformers==4.
|
| 69 |
```
|
| 70 |
|
| 71 |
-
另外，推荐安装`flash-attention
|
| 72 |
|
| 73 |
-
In addition, it is recommended to install the `flash-attention` library
|
| 74 |
|
| 75 |
```bash
|
| 76 |
-
git clone https://github.com/Dao-AILab/flash-attention
|
| 77 |
cd flash-attention && pip install .
|
| 78 |
-
|
| 79 |
-
|
| 80 |
-
# pip install csrc/rotary
|
| 81 |
```
|
| 82 |
-
<br>
|
| 83 |
|
| 84 |
## 快速使用（Quickstart）
|
| 85 |
|
|
@@ -92,7 +85,7 @@ from transformers import AutoModelForCausalLM, AutoTokenizer
|
|
| 92 |
from transformers.generation import GenerationConfig
|
| 93 |
|
| 94 |
# Note: The default behavior now has injection attack prevention off.
|
| 95 |
-
tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen-7B", trust_remote_code=True)
|
| 96 |
|
| 97 |
# use bf16
|
| 98 |
# model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen-7B", device_map="auto", trust_remote_code=True, bf16=True).eval()
|
|
@@ -103,35 +96,25 @@ tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen-7B", trust_remote_code=True
|
|
| 103 |
# use auto mode, automatically select precision based on the device.
|
| 104 |
model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen-7B", device_map="auto", trust_remote_code=True).eval()
|
| 105 |
|
| 106 |
-
# Specify hyperparameters for generation
|
| 107 |
-
|
| 108 |
|
| 109 |
inputs = tokenizer('蒙古国的首都是乌兰巴托（Ulaanbaatar）\n冰岛的首都是雷克雅未克（Reykjavik）\n埃塞俄比亚的首都是', return_tensors='pt')
|
| 110 |
-
inputs = inputs.to(
|
| 111 |
pred = model.generate(**inputs)
|
| 112 |
print(tokenizer.decode(pred.cpu()[0], skip_special_tokens=True))
|
| 113 |
# 蒙古国的首都是乌兰巴托（Ulaanbaatar）\n冰岛的首都是雷克雅未克（Reykjavik）\n埃塞俄比亚的首都是亚的斯亚贝巴（Addis Ababa）...
|
| 114 |
```
|
| 115 |
|
| 116 |
-
|
| 117 |
-
|
| 118 |
-
For more information, please refer to our [GitHub repo](https://github.com/QwenLM/Qwen) for more information.
|
| 119 |
-
<br>
|
| 120 |
-
|
| 121 |
-
## Tokenizer
|
| 122 |
|
| 123 |
-
|
| 124 |
-
|
| 125 |
-
基于tiktoken的分词器有别于其他分词器，比如sentencepiece分词器。尤其在微调阶段，需要特别注意特殊token的使用。关于tokenizer的更多信息，以及微调时涉及的相关使用，请参阅[文档](https://github.com/QwenLM/Qwen/blob/main/tokenization_note_zh.md)。
|
| 126 |
-
|
| 127 |
-
Our tokenizer based on tiktoken is different from other tokenizers, e.g., sentencepiece tokenizer. You need to pay attention to special tokens, especially in finetuning. For more detailed information on the tokenizer and related use in fine-tuning, please refer to the [documentation](https://github.com/QwenLM/Qwen/blob/main/tokenization_note.md).
|
| 128 |
-
<br>
|
| 129 |
|
| 130 |
## 模型细节 (Model)
|
| 131 |
|
| 132 |
-
Qwen-7B
|
| 133 |
|
| 134 |
-
The details of the model architecture of Qwen-7B are listed as follows
|
| 135 |
|
| 136 |
| Hyperparameter | Value |
|
| 137 |
|:----------------|:-------|
|
|
@@ -139,7 +122,7 @@ The details of the model architecture of Qwen-7B are listed as follows.
|
|
| 139 |
| n_heads | 32 |
|
| 140 |
| d_model | 4096 |
|
| 141 |
| vocab size | 151851 |
|
| 142 |
-
| sequence length |
|
| 143 |
|
| 144 |
在位置编码、FFN激活函数和normalization的实现方式上，我们也采用了目前最流行的做法，
|
| 145 |
即RoPE相对位置编码、SwiGLU激活函数、RMSNorm（可选安装flash-attention加速）。
|
|
@@ -151,7 +134,9 @@ The details of the model architecture of Qwen-7B are listed as follows.
|
|
| 151 |
|
| 152 |
可以看到Qwen-7B在保持中英代码高效解码的前提下，对部分使用人群较多的语种（泰语th、希伯来语he、阿拉伯语ar、韩语ko、越南语vi、日语ja、土耳其语tr、印尼语id、波兰语pl、俄语ru、荷兰语nl、葡萄牙语pt、意大利语it、德语de、西班牙语es、法语fr等）上也实现了较高的压缩率，使得模型在这些语种上也具备较强的可扩展性和较高的训练和推理效率。
|
| 153 |
|
| 154 |
-
|
|
|
|
|
|
|
| 155 |
|
| 156 |
<p align="center">
|
| 157 |
<img src="assets/tokenizer.png" style="width: 1200px"/>
|
|
@@ -165,112 +150,243 @@ We randomly selected 1 million document corpus of each language to test and comp
|
|
| 165 |
|
| 166 |
As can be seen, while ensuring the efficient decoding of Chinese, English, and code, Qwen-7B also achieves a high compression rate for many other languages (such as th, he, ar, ko, vi, ja, tr, id, pl, ru, nl, pt, it, de, es, fr etc.), equipping the model with strong scalability as well as high training and inference efficiency in these languages.
|
| 167 |
|
| 168 |
-
The scale of
|
| 169 |
-
<br>
|
| 170 |
|
| 171 |
## 评测效果（Evaluation）
|
| 172 |
-
|
| 173 |
-
|
| 174 |
-
|
| 175 |
-
|
| 176 |
-
|
| 177 |
-
|
| 178 |
-
|
| 179 |
-
|
| 180 |
-
|
| 181 |
-
|
| 182 |
-
|
| 183 |
-
|
| 184 |
-
|
| 185 |
-
|
| 186 |
-
|
|
| 187 |
-
|
| 188 |
-
|
|
| 189 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 190 |
|
| 191 |
### 长序列评测（Long-Context Evaluation）
|
| 192 |
|
| 193 |
-
我们引入NTK插值，LogN
|
| 194 |
|
| 195 |
-
**(若要启用NTK和LogN注意力缩放，请将config.json里的`
|
| 196 |
|
| 197 |
We introduce NTK-aware interpolation, LogN attention scaling, Window attention, etc. to extend the context length to over 8K tokens. We conduct language modeling experiments on the arXiv dataset with the PPL evaluation. Results are demonstrated below:
|
| 198 |
|
| 199 |
**(To use NTK interpolation and LogN scaling, please set `use_dynamic_ntk` and `use_long_attn` to true in config.json.)**
|
|
|
|
| 200 |
<table>
|
| 201 |
<tr>
|
| 202 |
-
<th rowspan="2">Model</th><th colspan="
|
| 203 |
</tr>
|
| 204 |
<tr>
|
| 205 |
-
<th align="center">1024</th><th align="center">2048</th><th align="center">4096</th><th align="center">8192</th><th align="center">16384</th
|
| 206 |
-
</tr>
|
| 207 |
-
<tr>
|
| 208 |
-
<td>Qwen-7B (original)</td><td align="center">4.23</td><td align="center">3.78</td><td align="center">39.35</td><td align="center">469.81</td><td align="center">2645.09</td><td align="center">-</td>
|
| 209 |
</tr>
|
| 210 |
<tr>
|
| 211 |
-
<td
|
| 212 |
</tr>
|
| 213 |
<tr>
|
| 214 |
-
<td>+ dynamic_ntk
|
| 215 |
</tr>
|
| 216 |
<tr>
|
| 217 |
-
<td>+ dynamic_ntk + logn
|
| 218 |
</tr>
|
| 219 |
<tr>
|
| 220 |
-
|
| 221 |
-
<td>Qwen-7B</td><td align="center"><b>4.23</b></td><td align="center"><b>3.81</b></td><td align="center"><b>3.52</b></td><td align="center"><b>3.31</b></td><td align="center">7.27</td><td align="center">181.49</td>
|
| 222 |
-
</tr>
|
| 223 |
-
<tr>
|
| 224 |
-
<td>+ dynamic_ntk + logn + window_attn</td><td align="center"><b>4.23</b></td><td align="center"><b>3.81</b></td><td align="center"><b>3.52</b></td><td align="center"><b>3.33</b></td><td align="center"><b>3.22</b></td><td align="center"><b>3.17</b></td>
|
| 225 |
-
</tr>
|
| 226 |
-
<tr>
|
| 227 |
-
<td>Qwen-14B</td><td align="center"><b>-</b></td><td align="center"><b>3.46</b></td><td align="center">22.79</td><td align="center">334.65</td><td align="center">3168.35</td><td align="center">-</td>
|
| 228 |
-
</tr>
|
| 229 |
-
<tr>
|
| 230 |
-
<td>+ dynamic_ntk + logn + window_attn</td><td align="center"><b>-</b></td><td align="center"><b>3.46</b></td><td align="center"><b>3.29</b></td><td align="center"><b>3.18</b></td><td align="center">3.42</td><td align="center">-</td>
|
| 231 |
</tr>
|
| 232 |
</table>
|
| 233 |
|
| 234 |
-
##
|
| 235 |
|
| 236 |
-
|
| 237 |
|
| 238 |
-
We
|
| 239 |
-
<br>
|
| 240 |
|
| 241 |
-
|
|
|
|
|
|
|
| 242 |
|
| 243 |
-
|
| 244 |
|
| 245 |
-
|
| 246 |
-
<br>
|
| 247 |
|
| 248 |
-
|
|
|
|
|
|
|
| 249 |
|
| 250 |
-
|
| 251 |
|
| 252 |
-
|
| 253 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 254 |
```
|
| 255 |
-
|
| 256 |
-
|
| 257 |
-
|
| 258 |
-
|
| 259 |
-
|
| 260 |
-
|
| 261 |
-
|
| 262 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 263 |
|
| 264 |
## 使用协议（License Agreement）
|
| 265 |
|
| 266 |
-
我们的代码和模型权重对学术研究完全开放，并支持商用。请查看[LICENSE](https://github.com/QwenLM/Qwen/blob/main/
|
| 267 |
|
| 268 |
-
Our code and checkpoints are open to research purpose, and they are allowed for commercial purposes. Check [LICENSE](https://github.com/QwenLM/Qwen/blob/main/
|
| 269 |
-
<br>
|
| 270 |
|
| 271 |
## 联系我们（Contact Us）
|
| 272 |
|
| 273 |
-
|
| 274 |
|
| 275 |
-
If you are interested to leave a message to either our research team or product team,
|
| 276 |
|
|
|
|
| 6 |
- qwen
|
| 7 |
pipeline_tag: text-generation
|
| 8 |
inference: false
|
|
|
|
|
|
|
|
|
|
| 9 |
---
|
| 10 |
|
| 11 |
# Qwen-7B
|
| 12 |
|
| 13 |
<p align="center">
|
| 14 |
+
<img src="https://qianwen-res.oss-cn-beijing.aliyuncs.com/logo.jpg" width="400"/>
|
| 15 |
<p>
|
| 16 |
<br>
|
| 17 |
|
| 18 |
<p align="center">
|
| 19 |
+
Qwen-7B <a href="https://modelscope.cn/models/qwen/Qwen-7B/summary">🤖 </a> | <a href="https://huggingface.co/Qwen/Qwen-7B">🤗</a>  ｜ Qwen-7B-Chat <a href="https://modelscope.cn/models/qwen/Qwen-7B-Chat/summary">🤖 </a>| <a href="https://huggingface.co/Qwen/Qwen-7B-Chat">🤗</a>  ｜  <a href="https://modelscope.cn/studios/qwen/Qwen-7B-Chat-Demo/summary">Demo</a>  ｜  <a href="https://github.com/QwenLM/Qwen-7B/blob/main/tech_memo.md">Report</a>
|
|
|
|
|
|
|
| 20 |
</p>
|
| 21 |
<br>
|
| 22 |
|
| 23 |
## 介绍 (Introduction)
|
| 24 |
|
| 25 |
+
**通义千问-7B（Qwen-7B）**是阿里云研发的通义千问大模型系列的70亿参数规模的模型。Qwen-7B是基于Transformer的大语言模型, 在超大规模的预训练数据上进行训练得到。预训练数据类型多样，覆盖广泛，包括大量网络文本、专业书籍、代码等。同时，在Qwen-7B的基础上，我们使用对齐机制打造了基于大语言模型的AI助手Qwen-7B-Chat。本仓库为Qwen-7B的仓库。
|
| 26 |
|
| 27 |
通义千问-7B（Qwen-7B）主要有以下特点：
|
| 28 |
|
| 29 |
+
1. **大规模高质量训练语料**：使用超过2.2万亿tokens的数据进行预训练，包含高质量中、英、多语言、代码、数学等数据，涵盖通用及专业领域的训练语料。通过大量对比实验对预训练语料分布进行了优化。
|
| 30 |
2. **强大的性能**：Qwen-7B在多个中英文下游评测任务上（涵盖常识推理、代码、数学、翻译等），效果显著超越现有的相近规模开源模型，甚至在部分指标上相比更大尺寸模型也有较强竞争力。具体评测结果请详见下文。
|
| 31 |
3. **覆盖更全面的词表**：相比目前以中英词表为主的开源模型，Qwen-7B使用了约15万大小的词表。该词表对多语言更加友好，方便用户在不扩展词表的情况下对部分语种进行能力增强和扩展。
|
| 32 |
|
| 33 |
+
如果您想了解更多关于通义千问7B开源模型的细节，我们建议您参阅[Github代码库](https://github.com/QwenLM/Qwen-7B)。
|
| 34 |
|
| 35 |
+
**Qwen-7B** is the 7B-parameter version of the large language model series, Qwen (abbr. Tongyi Qianwen), proposed by Aibaba Cloud. Qwen-7B is a Transformer-based large language model, which is pretrained on a large volume of data, including web texts, books, codes, etc. Additionally, based on the pretrained Qwen-7B, we release Qwen-7B-Chat, a large-model-based AI assistant, which is trained with alignment techniques. This repository is the one for Qwen-7B.
|
| 36 |
|
| 37 |
The features of Qwen-7B include:
|
| 38 |
|
| 39 |
+
1. **Large-scale high-quality training corpora**: It is pretrained on over 2.2 trillion tokens, including Chinese, English, multilingual texts, code, and mathematics, covering general and professional fields. The distribution of the pre-training corpus has been optimized through a large number of ablation experiments.
|
| 40 |
2. **Competitive performance**: It significantly surpasses existing open-source models of similar scale on multiple Chinese and English downstream evaluation tasks (including commonsense, reasoning, code, mathematics, etc.), and even surpasses some larger-scale models in several benchmarks. See below for specific evaluation results.
|
| 41 |
3. **More comprehensive vocabulary coverage**: Compared with other open-source models based on Chinese and English vocabularies, Qwen-7B uses a vocabulary of over 150K tokens. This vocabulary is more friendly to multiple languages, enabling users to directly further enhance the capability for certain languages without expanding the vocabulary.
|
| 42 |
|
| 43 |
+
For more details about the open-source model of Qwen-7B, please refer to the [Github](https://github.com/QwenLM/Qwen-7B) code repository.
|
|
|
|
| 44 |
|
| 45 |
## 要求（Requirements）
|
| 46 |
|
| 47 |
* python 3.8及以上版本
|
| 48 |
* pytorch 1.12及以上版本，推荐2.0及以上版本
|
| 49 |
* 建议使用CUDA 11.4及以上（GPU用户、flash-attention用户等需考虑此选项）
|
| 50 |
+
|
| 51 |
+
|
| 52 |
* python 3.8 and above
|
| 53 |
* pytorch 1.12 and above, 2.0 and above are recommended
|
| 54 |
* CUDA 11.4 and above are recommended (this is for GPU users, flash-attention users, etc.)
|
|
|
|
| 55 |
|
| 56 |
## 依赖项 (Dependency)
|
| 57 |
|
|
|
|
| 60 |
To run Qwen-7B, please make sure you meet the above requirements, and then execute the following pip commands to install the dependent libraries.
|
| 61 |
|
| 62 |
```bash
|
| 63 |
+
pip install transformers==4.31.0 accelerate tiktoken einops
|
| 64 |
```
|
| 65 |
|
| 66 |
+
另外，推荐安装`flash-attention`库，以实现更高的效率和更低的显存占用。
|
| 67 |
|
| 68 |
+
In addition, it is recommended to install the `flash-attention` library for higher efficiency and lower memory usage.
|
| 69 |
|
| 70 |
```bash
|
| 71 |
+
git clone -b v1.0.8 https://github.com/Dao-AILab/flash-attention
|
| 72 |
cd flash-attention && pip install .
|
| 73 |
+
pip install csrc/layer_norm
|
| 74 |
+
pip install csrc/rotary
|
|
|
|
| 75 |
```
|
|
|
|
| 76 |
|
| 77 |
## 快速使用（Quickstart）
|
| 78 |
|
|
|
|
| 85 |
from transformers.generation import GenerationConfig
|
| 86 |
|
| 87 |
# Note: The default behavior now has injection attack prevention off.
|
| 88 |
+
tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen-7B-Chat", trust_remote_code=True)
|
| 89 |
|
| 90 |
# use bf16
|
| 91 |
# model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen-7B", device_map="auto", trust_remote_code=True, bf16=True).eval()
|
|
|
|
| 96 |
# use auto mode, automatically select precision based on the device.
|
| 97 |
model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen-7B", device_map="auto", trust_remote_code=True).eval()
|
| 98 |
|
| 99 |
+
# Specify hyperparameters for generation
|
| 100 |
+
model.generation_config = GenerationConfig.from_pretrained("Qwen/Qwen-7B", trust_remote_code=True)
|
| 101 |
|
| 102 |
inputs = tokenizer('蒙古国的首都是乌兰巴托（Ulaanbaatar）\n冰岛的首都是雷克雅未克（Reykjavik）\n埃塞俄比亚的首都是', return_tensors='pt')
|
| 103 |
+
inputs = inputs.to('cuda:0')
|
| 104 |
pred = model.generate(**inputs)
|
| 105 |
print(tokenizer.decode(pred.cpu()[0], skip_special_tokens=True))
|
| 106 |
# 蒙古国的首都是乌兰巴托（Ulaanbaatar）\n冰岛的首都是雷克雅未克（Reykjavik）\n埃塞俄比亚的首都是亚的斯亚贝巴（Addis Ababa）...
|
| 107 |
```
|
| 108 |
|
| 109 |
+
关于更多的使用说明，请参考我们的[Github repo](https://github.com/QwenLM/Qwen-7B)获取更多信息。
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 110 |
|
| 111 |
+
For more information, please refer to our [Github repo](https://github.com/QwenLM/Qwen-7B) for more information.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 112 |
|
| 113 |
## 模型细节 (Model)
|
| 114 |
|
| 115 |
+
Qwen-7B模型规模基本情况如下所示：
|
| 116 |
|
| 117 |
+
The details of the model architecture of Qwen-7B are listed as follows:
|
| 118 |
|
| 119 |
| Hyperparameter | Value |
|
| 120 |
|:----------------|:-------|
|
|
|
|
| 122 |
| n_heads | 32 |
|
| 123 |
| d_model | 4096 |
|
| 124 |
| vocab size | 151851 |
|
| 125 |
+
| sequence length | 2048 |
|
| 126 |
|
| 127 |
在位置编码、FFN激活函数和normalization的实现方式上，我们也采用了目前最流行的做法，
|
| 128 |
即RoPE相对位置编码、SwiGLU激活函数、RMSNorm（可选安装flash-attention加速）。
|
|
|
|
| 134 |
|
| 135 |
可以看到Qwen-7B在保持中英代码高效解码的前提下，对部分使用人群较多的语种（泰语th、希伯来语he、阿拉伯语ar、韩语ko、越南语vi、日语ja、土耳其语tr、印尼语id、波兰语pl、俄语ru、荷兰语nl、葡萄牙语pt、意大利语it、德语de、西班牙语es、法语fr等）上也实现了较高的压缩率，使得模型在这些语种上也具备较强的可扩展性和较高的训练和推理效率。
|
| 136 |
|
| 137 |
+
在预训练数据方面，Qwen-7B模型一方面利用了部分开源通用语料，
|
| 138 |
+
另一方面也积累了海量全网语料以及高质量文本内容，去重及过滤后的语料超过2.2T tokens。
|
| 139 |
+
囊括全网文本、百科、书籍、代码、数学及各个领域垂类。
|
| 140 |
|
| 141 |
<p align="center">
|
| 142 |
<img src="assets/tokenizer.png" style="width: 1200px"/>
|
|
|
|
| 150 |
|
| 151 |
As can be seen, while ensuring the efficient decoding of Chinese, English, and code, Qwen-7B also achieves a high compression rate for many other languages (such as th, he, ar, ko, vi, ja, tr, id, pl, ru, nl, pt, it, de, es, fr etc.), equipping the model with strong scalability as well as high training and inference efficiency in these languages.
|
| 152 |
|
| 153 |
+
For pre-training data, on the one hand, Qwen-7B uses part of the open-source generic corpus. On the other hand, it uses a massive amount of accumulated web corpus and high-quality text content. The scale of corpus reaches over 2.2T tokens after deduplication and filtration, encompassing web text, encyclopedias, books, code, mathematics, and various domain.
|
|
|
|
| 154 |
|
| 155 |
## 评测效果（Evaluation）
|
| 156 |
+
|
| 157 |
+
### 中文评测（Chinese Evaluation）
|
| 158 |
+
|
| 159 |
+
#### C-Eval
|
| 160 |
+
|
| 161 |
+
[C-Eval](https://arxiv.org/abs/2305.08322)是评测预训练模型中文常识能力的常用测评框架，覆盖人文、社科、理工、其他专业四个大方向共52个学科。
|
| 162 |
+
我们按照标准做法，以开发集样本作为few-shot来源，评价Qwen-7B预训练模型的5-shot验证集与测试集准确率。
|
| 163 |
+
|
| 164 |
+
[C-Eval](https://arxiv.org/abs/2305.08322) is a common evaluation benchmark for testing the common sense capability of pre-trained models in Chinese. It covers 52 subjects in four major directions: humanities, social sciences, STEM, and other specialties. According to the standard practice, we use the development set samples as the source of few-shot, to evaluate the 5-shot validation set and test set accuracy of the Qwen-7B pre-trained model.
|
| 165 |
+
|
| 166 |
+
在C-Eval验证集上，Qwen-7B模型和其他模型的准确率对比如下：
|
| 167 |
+
|
| 168 |
+
The accuracy comparison of Qwen-7B and the other models on the C-Eval validation set is shown as follows:
|
| 169 |
+
|
| 170 |
+
| Model | Avg. |
|
| 171 |
+
|:----------------|:--------:|
|
| 172 |
+
| Alpaca-7B | 28.9 |
|
| 173 |
+
| Vicuna-7B | 31.2 |
|
| 174 |
+
| ChatGLM-6B | 37.1 |
|
| 175 |
+
| Baichuan-7B | 42.7 |
|
| 176 |
+
| ChatGLM2-6B | 50.9 |
|
| 177 |
+
| InternLM-7B | 53.4 |
|
| 178 |
+
| ChatGPT | 53.5 |
|
| 179 |
+
| Claude-v1.3 | 55.5 |
|
| 180 |
+
| **Qwen-7B** | **60.8** |
|
| 181 |
+
|
| 182 |
+
在C-Eval测试集上，Qwen-7B预训练模型与其他模型的效果对比如下表所示：
|
| 183 |
+
|
| 184 |
+
The performance comparison of Qwen-7B and other models on the C-Eval test set is shown in the following table:
|
| 185 |
+
|
| 186 |
+
| Model | Avg. | Avg. (Hard) | STEM | Social Sciences | Humanities | Others |
|
| 187 |
+
|:--------------|:------:|:------:|:------:|:------:|:------:|:------:|
|
| 188 |
+
| ChatGLM-6B | 38.9 | 29.2 | 33.3 | 48.3 | 41.3 | 38.0 |
|
| 189 |
+
| Chinese-Alpaca-Plus-13B | 41.5 | 30.5 | 36.6 | 49.7 | 43.1 | 41.2 |
|
| 190 |
+
| Baichuan-7B | 42.8 | 31.5 | 38.2 | 52.0 | 46.2 | 39.3 |
|
| 191 |
+
| WestlakeLM-19B | 44.6 | 34.9 | 41.6 | 51.0 | 44.3 | 44.5 |
|
| 192 |
+
| AndesLM-13B | 46.0 | 29.7 | 38.1 | 61.0 | 51.0 | 41.9 |
|
| 193 |
+
| BatGPT-15B-sirius | 47.0 | 31.9 | 42.7 | 57.5 | 48.6 | 43.6 |
|
| 194 |
+
| ChatGLM2-6B | 51.7 | 37.1 | 48.6 | 60.5 | 51.3 | 49.8 |
|
| 195 |
+
| InternLM-7B | 52.8 | 37.1 | 48.0 | 67.4 | 55.4 | 45.8 |
|
| 196 |
+
| Baichuan-13B | 53.6 | 36.7 | 47.0 | 66.8 | 57.3 | 49.8 |
|
| 197 |
+
| Claude-v1.3 | 54.2 | 39.0 | 51.9 | 61.7 | 52.1 | 53.7 |
|
| 198 |
+
| ChatGPT | 54.4 | 41.4 | 52.9 | 61.8 | 50.9 | 53.6 |
|
| 199 |
+
| **Qwen-7B** | **59.6** | 41.0 | 52.8 | 74.1 | 63.1 | 55.2 |
|
| 200 |
+
|
| 201 |
+
可以看到，Qwen-7B在同等规模现有模型中取得了最高的分数，甚至相比更大规模模型也具有较强竞争力。
|
| 202 |
+
|
| 203 |
+
As can be seen, Qwen-7B achieves the best performance out of all existing models with similar scale and even surpasses larger-scale models.
|
| 204 |
+
|
| 205 |
+
### 英文评测（English Evaluation）
|
| 206 |
+
|
| 207 |
+
#### MMLU
|
| 208 |
+
|
| 209 |
+
[MMLU](https://arxiv.org/abs/2009.03300)是目前评测英文综合能力最权威的基准评测之一，同样覆盖了不同学科领域、不同难度层级的57个子任务。
|
| 210 |
+
|
| 211 |
+
Qwen-7B在MMLU 5-shot准确率表现如下表：
|
| 212 |
+
|
| 213 |
+
[MMLU](https://arxiv.org/abs/2009.03300) is currently one of the most recognized benchmarks for evaluating English comprehension abilities, covering 57 subtasks across different academic fields and difficulty levels. The MMLU 5-shot accuracy performance of Qwen-7B is shown in the following table:
|
| 214 |
+
|
| 215 |
+
| Model | Avg. | STEM | Social Sciences | Humanities | Others |
|
| 216 |
+
|:--------------|:------:|:------:|:------:|:------:|:------:|
|
| 217 |
+
| LLaMA-7B | 35.1 | 30.5 | 38.3 | 34.0 | 38.1 |
|
| 218 |
+
| Baichuan-7B | 42.3 | 35.6 | 48.9 | 38.4 | 48.1 |
|
| 219 |
+
| LLaMA2-7B | 45.3 | 36.4 | 51.2 | 42.9 | 52.2 |
|
| 220 |
+
| LLaMA-13B | 46.9 | 35.8 | 53.8 | 45.0 | 53.3 |
|
| 221 |
+
| ChatGLM2-6B | 47.9 | 41.2 | 54.4 | 43.7 | 54.5 |
|
| 222 |
+
| InternLM-7B | 51.0 | - | - | - | - |
|
| 223 |
+
| Baichuan-13B | 51.6 | 41.6 | 60.9 | 47.4 | 58.5 |
|
| 224 |
+
| LLaMA2-13B | 54.8 | 44.1 | 62.6 | 52.8 | 61.1 |
|
| 225 |
+
| ChatGLM2-12B | 56.2 | 48.2 | 65.1 | 52.6 | 60.9 |
|
| 226 |
+
| **Qwen-7B** | **56.7** | 47.6 | 65.9 | 51.5 | 64.7 |
|
| 227 |
+
|
| 228 |
+
在英文方面，Qwen-7B的效果同样超过了目前国内外其他同类开源预训练模型，同样对比更大规模版本的模型也具有较强竞争力。
|
| 229 |
+
|
| 230 |
+
In terms of English, Qwen-7B also surpasses other similar open-source pre-trained models, and is competitive when compared to larger versions of other models.
|
| 231 |
+
|
| 232 |
+
### 代码评测（Coding Evaluation）
|
| 233 |
+
|
| 234 |
+
我们在[HumanEval](https://github.com/openai/human-eval)（0-shot）上对比预训练模型的代码能力，结果如下：
|
| 235 |
+
|
| 236 |
+
We compared the code capabilities of pre-trained models on [HumanEval](https://github.com/openai/human-eval), and the results are as follows:
|
| 237 |
+
|
| 238 |
+
| Model | Pass@1 |
|
| 239 |
+
|:--------------|:------:|
|
| 240 |
+
| Baichuan-7B | 9.2 |
|
| 241 |
+
| ChatGLM2-6B | 9.2 |
|
| 242 |
+
| InternLM-7B | 10.4 |
|
| 243 |
+
| LLaMA-7B | 10.5 |
|
| 244 |
+
| LLaMA2-7B | 12.8 |
|
| 245 |
+
| Baichuan-13B | 12.8 |
|
| 246 |
+
| LLaMA-13B | 15.8 |
|
| 247 |
+
| MPT-7B | 18.3 |
|
| 248 |
+
| LLaMA2-13B | 18.3 |
|
| 249 |
+
| **Qwen-7B** | **24.4** |
|
| 250 |
+
|
| 251 |
+
### 数学评测（Mathematics Evaluation）
|
| 252 |
+
|
| 253 |
+
数学能力使用常用的[GSM8K](https://github.com/openai/grade-school-math)数据集（8-shot）评价：
|
| 254 |
+
|
| 255 |
+
We compared the math capabilities of pre-trained models on [GSM8K](https://github.com/openai/grade-school-math) (8-shot), and the results are as follows:
|
| 256 |
+
|
| 257 |
+
| Model | Acc. |
|
| 258 |
+
|:--------------|:------:|
|
| 259 |
+
| MPT-7B | 6.8 |
|
| 260 |
+
| Falcon-7B | 6.8 |
|
| 261 |
+
| Baichuan-7B | 9.7 |
|
| 262 |
+
| LLaMA-7B | 11.0 |
|
| 263 |
+
| LLaMA2-7B | 14.6 |
|
| 264 |
+
| LLaMA-13B | 17.8 |
|
| 265 |
+
| Baichuan-13B | 26.6 |
|
| 266 |
+
| LLaMA2-13B | 28.7 |
|
| 267 |
+
| InternLM-7B | 31.2 |
|
| 268 |
+
| ChatGLM2-6B | 32.4 |
|
| 269 |
+
| ChatGLM2-12B | 40.9 |
|
| 270 |
+
| **Qwen-7B** | **51.6** |
|
| 271 |
+
|
| 272 |
+
### 翻译评测
|
| 273 |
+
|
| 274 |
+
我们使用[WMT22](https://www.statmt.org/wmt22/translation-task.html)中-英（zh-en）和英-中（en-zh）数据集（5-shot BLEU）评测：
|
| 275 |
+
|
| 276 |
+
We compared the translation capabilities of pre-trained models on [WMT22](https://www.statmt.org/wmt22/translation-task.html) zh-en and en-zh (5-shot BLEU), and the results are as follows:
|
| 277 |
+
|
| 278 |
+
| Model | Avg. | zh-en | en-zh |
|
| 279 |
+
|:------------|:--------:|:--------:|:--------:|
|
| 280 |
+
| InternLM-7B | 11.8 | 9.0 | 14.5 |
|
| 281 |
+
| LLaMA-7B | 12.7 | 16.7 | 8.7 |
|
| 282 |
+
| LLaMA-13B | 15.8 | 19.5 | 12.0 |
|
| 283 |
+
| LLaMA2-7B | 19.9 | 21.9 | 17.9 |
|
| 284 |
+
| Bloom-7B | 20.3 | 19.1 | 21.4 |
|
| 285 |
+
| LLaMA2-13B | 23.3 | 22.4 | 24.2 |
|
| 286 |
+
| PolyLM-13B | 23.6 | 20.2 | 27.0 |
|
| 287 |
+
| Baichuan-7B | 24.6 | 22.6 | 26.6 |
|
| 288 |
+
| **Qwen-7B** | **27.5** | **24.3** | **30.6** |
|
| 289 |
|
| 290 |
### 长序列评测（Long-Context Evaluation）
|
| 291 |
|
| 292 |
+
我们引入NTK插值，LogN注意力缩放，窗口注意力等技巧，将模型的上下文长度扩展到8K以上。在arXiv数据上使用PPL指标测试Qwen-7B在不同长度下的表现，结果如下：
|
| 293 |
|
| 294 |
+
**(若要启用NTK和LogN注意力缩放，请将config.json里的`use_dynamc_ntk`和`use_logn_attn`设置为true)**
|
| 295 |
|
| 296 |
We introduce NTK-aware interpolation, LogN attention scaling, Window attention, etc. to extend the context length to over 8K tokens. We conduct language modeling experiments on the arXiv dataset with the PPL evaluation. Results are demonstrated below:
|
| 297 |
|
| 298 |
**(To use NTK interpolation and LogN scaling, please set `use_dynamic_ntk` and `use_long_attn` to true in config.json.)**
|
| 299 |
+
|
| 300 |
<table>
|
| 301 |
<tr>
|
| 302 |
+
<th rowspan="2">Model</th><th colspan="5" align="center">序列长度 Sequence Length</th>
|
| 303 |
</tr>
|
| 304 |
<tr>
|
| 305 |
+
<th align="center">1024</th><th align="center">2048</th><th align="center">4096</th><th align="center">8192</th><th align="center">16384</th>
|
|
|
|
|
|
|
|
|
|
| 306 |
</tr>
|
| 307 |
<tr>
|
| 308 |
+
<td>Qwen-7B</td><td align="center"><b>4.23</b></td><td align="center"><b>3.78</b></td><td align="center">39.35</td><td align="center">469.81</td><td align="center">2645.09</td>
|
| 309 |
</tr>
|
| 310 |
<tr>
|
| 311 |
+
<td>+ dynamic_ntk</td><td align="center"><b>4.23</b></td><td align="center"><b>3.78</b></td><td align="center">3.59</td><td align="center">3.66</td><td align="center">5.71</td>
|
| 312 |
</tr>
|
| 313 |
<tr>
|
| 314 |
+
<td>+ dynamic_ntk + logn</td><td align="center"><b>4.23</b></td><td align="center"><b>3.78</b></td><td align="center"><b>3.58</b></td><td align="center">3.56</td><td align="center">4.62</td>
|
| 315 |
</tr>
|
| 316 |
<tr>
|
| 317 |
+
<td>+ dynamic_ntk + logn + window_attn</td><td align="center"><b>4.23</b></td><td align="center"><b>3.78</b></td><td align="center"><b>3.58</b></td><td align="center"><b>3.49</b></td><td align="center"><b>4.32</b></td>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 318 |
</tr>
|
| 319 |
</table>
|
| 320 |
|
| 321 |
+
## 量化（Quantization）
|
| 322 |
|
| 323 |
+
如希望使用更低精度的量化模型，如4比特和8比特的模型，我们提供了简单的示例来说明如何快速使用量化模型。在开始前，确保你已经安装了`bitsandbytes`。请注意，`bitsandbytes`的安装要求是：
|
| 324 |
|
| 325 |
+
We provide examples to show how to load models in `NF4` and `Int8`. For starters, make sure you have implemented `bitsandbytes`. Note that the requirements for `bitsandbytes` are:
|
|
|
|
| 326 |
|
| 327 |
+
```
|
| 328 |
+
**Requirements** Python >=3.8. Linux distribution (Ubuntu, MacOS, etc.) + CUDA > 10.0.
|
| 329 |
+
```
|
| 330 |
|
| 331 |
+
Windows用户需安装特定版本的`bitsandbytes`，可选项包括[bitsandbytes-windows-webui](https://github.com/jllllll/bitsandbytes-windows-webui/releases/tag/wheels)。
|
| 332 |
|
| 333 |
+
Windows users should find another option, which might be [bitsandbytes-windows-webui](https://github.com/jllllll/bitsandbytes-windows-webui/releases/tag/wheels).
|
|
|
|
| 334 |
|
| 335 |
+
```bash
|
| 336 |
+
pip install bitsandbytes
|
| 337 |
+
```
|
| 338 |
|
| 339 |
+
你只需要在`AutoModelForCausalLM.from_pretrained`中添加你的量化配置，即可使用量化模型。如下所示：
|
| 340 |
|
| 341 |
+
Then you only need to add your quantization configuration to `AutoModelForCausalLM.from_pretrained`. See the example below:
|
| 342 |
|
| 343 |
+
```python
|
| 344 |
+
from transformers import AutoModelForCausalLM, BitsAndBytesConfig
|
| 345 |
+
|
| 346 |
+
# quantization configuration for NF4 (4 bits)
|
| 347 |
+
quantization_config = BitsAndBytesConfig(
|
| 348 |
+
load_in_4bit=True,
|
| 349 |
+
bnb_4bit_quant_type='nf4',
|
| 350 |
+
bnb_4bit_compute_dtype=torch.bfloat16
|
| 351 |
+
)
|
| 352 |
+
|
| 353 |
+
# quantization configuration for Int8 (8 bits)
|
| 354 |
+
quantization_config = BitsAndBytesConfig(load_in_8bit=True)
|
| 355 |
+
|
| 356 |
+
model = AutoModelForCausalLM.from_pretrained(
|
| 357 |
+
"Qwen/Qwen-7B",
|
| 358 |
+
device_map="cuda:0",
|
| 359 |
+
quantization_config=quantization_config,
|
| 360 |
+
max_memory=max_memory,
|
| 361 |
+
trust_remote_code=True,
|
| 362 |
+
).eval()
|
| 363 |
```
|
| 364 |
+
|
| 365 |
+
上述方法可以让我们将模型量化成`NF4`和`Int8`精度的模型进行读取，帮助我们节省显存开销。我们也提供了相关性能数据。我们发现尽管模型在效果上存在损失，但模型的显存开销大幅降低。
|
| 366 |
+
|
| 367 |
+
With this method, it is available to load Qwen-7B in `NF4` and `Int8`, which saves you memory usage. We provide related statistics of model performance below. We find that the quantization downgrades the effectiveness slightly but significantly increases inference efficiency and reduces memory costs.
|
| 368 |
+
|
| 369 |
+
| Precision | MMLU | Memory |
|
| 370 |
+
| :--------- | :-------: | :-----: |
|
| 371 |
+
| BF16 | 56.7 | 16.2G |
|
| 372 |
+
| Int8 | 52.8 | 10.1G |
|
| 373 |
+
| NF4 | 48.9 | 7.4G |
|
| 374 |
+
|
| 375 |
+
## 评测复现（Reproduction）
|
| 376 |
+
|
| 377 |
+
我们提供了评测脚本，方便大家复现模型效果，详见[链接](https://github.com/QwenLM/Qwen-7B/tree/main/eval)。提示：由于硬件和框架造成的舍入误差，复现结果如有小幅波动属于正常现象。
|
| 378 |
+
|
| 379 |
+
We have provided evaluation scripts to reproduce the performance of our model, details as [link](https://github.com/QwenLM/Qwen-7B/tree/main/eval).
|
| 380 |
|
| 381 |
## 使用协议（License Agreement）
|
| 382 |
|
| 383 |
+
我们的代码和模型权重对学术研究完全开放，并支持商用。请查看[LICENSE](https://github.com/QwenLM/Qwen-7B/blob/main/LICENSE)了解具体的开源协议细节。
|
| 384 |
|
| 385 |
+
Our code and checkpoints are open to research purpose, and they are allowed for commercial purposes. Check [LICENSE](https://github.com/QwenLM/Qwen-7B/blob/main/LICENSE) for more details about the license.
|
|
|
|
| 386 |
|
| 387 |
## 联系我们（Contact Us）
|
| 388 |
|
| 389 |
+
如果你想给我们的研发团队和产品团队留言，请通过邮件（qianwen_opensource@alibabacloud.com）联系我们。
|
| 390 |
|
| 391 |
+
If you are interested to leave a message to either our research team or product team, feel free to send an email to qianwen_opensource@alibabacloud.com.
|
| 392 |
|
assets/logo.jpg
CHANGED
|
|
assets/qwen_tokenizer.png
CHANGED
|
|
assets/tokenizer.png
DELETED
|
Binary file (80.9 kB)
|
|
|
assets/wechat.png
DELETED
|
Binary file (68.4 kB)
|
|
|
cache_autogptq_cuda_256.cpp
DELETED
|
@@ -1,198 +0,0 @@
|
|
| 1 |
-
#include <torch/all.h>
|
| 2 |
-
#include <torch/python.h>
|
| 3 |
-
#include <c10/cuda/CUDAGuard.h>
|
| 4 |
-
|
| 5 |
-
// adapted from https://github.com/PanQiWei/AutoGPTQ/blob/main/autogptq_extension/cuda_256/autogptq_cuda_256.cpp
|
| 6 |
-
void vecquant8matmul_cuda(
|
| 7 |
-
torch::Tensor vec, torch::Tensor mat, torch::Tensor mul,
|
| 8 |
-
torch::Tensor scales, torch::Tensor zeros,
|
| 9 |
-
torch::Tensor g_idx
|
| 10 |
-
);
|
| 11 |
-
|
| 12 |
-
void vecquant8matmul(
|
| 13 |
-
torch::Tensor vec, torch::Tensor mat, torch::Tensor mul,
|
| 14 |
-
torch::Tensor scales, torch::Tensor zeros,
|
| 15 |
-
torch::Tensor g_idx
|
| 16 |
-
) {
|
| 17 |
-
const at::cuda::OptionalCUDAGuard device_guard(device_of(vec));
|
| 18 |
-
vecquant8matmul_cuda(vec, mat, mul, scales, zeros, g_idx);
|
| 19 |
-
}
|
| 20 |
-
|
| 21 |
-
void vecquant8matmul_batched_cuda(
|
| 22 |
-
torch::Tensor vec, torch::Tensor mat, torch::Tensor mul,
|
| 23 |
-
torch::Tensor scales, torch::Tensor zeros
|
| 24 |
-
);
|
| 25 |
-
|
| 26 |
-
void vecquant8matmul_batched(
|
| 27 |
-
torch::Tensor vec, torch::Tensor mat, torch::Tensor mul,
|
| 28 |
-
torch::Tensor scales, torch::Tensor zeros
|
| 29 |
-
) {
|
| 30 |
-
const at::cuda::OptionalCUDAGuard device_guard(device_of(vec));
|
| 31 |
-
vecquant8matmul_batched_cuda(vec, mat, mul, scales, zeros);
|
| 32 |
-
}
|
| 33 |
-
|
| 34 |
-
void vecquant8matmul_batched_column_compression_cuda(
|
| 35 |
-
torch::Tensor vec, torch::Tensor mat, torch::Tensor mul,
|
| 36 |
-
torch::Tensor scales, torch::Tensor zeros
|
| 37 |
-
);
|
| 38 |
-
|
| 39 |
-
void vecquant8matmul_batched_column_compression(
|
| 40 |
-
torch::Tensor vec, torch::Tensor mat, torch::Tensor mul,
|
| 41 |
-
torch::Tensor scales, torch::Tensor zeros
|
| 42 |
-
) {
|
| 43 |
-
const at::cuda::OptionalCUDAGuard device_guard(device_of(vec));
|
| 44 |
-
vecquant8matmul_batched_column_compression_cuda(vec, mat, mul, scales, zeros);
|
| 45 |
-
}
|
| 46 |
-
|
| 47 |
-
void vecquant4matmul_batched_cuda(
|
| 48 |
-
torch::Tensor vec, torch::Tensor mat, torch::Tensor mul,
|
| 49 |
-
torch::Tensor scales, torch::Tensor zeros
|
| 50 |
-
);
|
| 51 |
-
|
| 52 |
-
void vecquant4matmul_batched(
|
| 53 |
-
torch::Tensor vec, torch::Tensor mat, torch::Tensor mul,
|
| 54 |
-
torch::Tensor scales, torch::Tensor zeros
|
| 55 |
-
) {
|
| 56 |
-
const at::cuda::OptionalCUDAGuard device_guard(device_of(vec));
|
| 57 |
-
vecquant4matmul_batched_cuda(vec, mat, mul, scales, zeros);
|
| 58 |
-
}
|
| 59 |
-
|
| 60 |
-
void vecquant4matmul_batched_column_compression_cuda(
|
| 61 |
-
torch::Tensor vec, torch::Tensor mat, torch::Tensor mul,
|
| 62 |
-
torch::Tensor scales, torch::Tensor zeros
|
| 63 |
-
);
|
| 64 |
-
|
| 65 |
-
void vecquant4matmul_batched_column_compression(
|
| 66 |
-
torch::Tensor vec, torch::Tensor mat, torch::Tensor mul,
|
| 67 |
-
torch::Tensor scales, torch::Tensor zeros
|
| 68 |
-
) {
|
| 69 |
-
const at::cuda::OptionalCUDAGuard device_guard(device_of(vec));
|
| 70 |
-
vecquant4matmul_batched_column_compression_cuda(vec, mat, mul, scales, zeros);
|
| 71 |
-
}
|
| 72 |
-
|
| 73 |
-
void vecquant8matmul_batched_old_cuda(
|
| 74 |
-
torch::Tensor vec, torch::Tensor mat, torch::Tensor mul,
|
| 75 |
-
torch::Tensor scales, torch::Tensor zeros
|
| 76 |
-
);
|
| 77 |
-
|
| 78 |
-
void vecquant8matmul_batched_old(
|
| 79 |
-
torch::Tensor vec, torch::Tensor mat, torch::Tensor mul,
|
| 80 |
-
torch::Tensor scales, torch::Tensor zeros
|
| 81 |
-
) {
|
| 82 |
-
const at::cuda::OptionalCUDAGuard device_guard(device_of(vec));
|
| 83 |
-
vecquant8matmul_batched_old_cuda(vec, mat, mul, scales, zeros);
|
| 84 |
-
}
|
| 85 |
-
|
| 86 |
-
|
| 87 |
-
void vecquant4matmul_batched_old_cuda(
|
| 88 |
-
torch::Tensor vec, torch::Tensor mat, torch::Tensor mul,
|
| 89 |
-
torch::Tensor scales, torch::Tensor zeros
|
| 90 |
-
);
|
| 91 |
-
|
| 92 |
-
void vecquant4matmul_batched_old(
|
| 93 |
-
torch::Tensor vec, torch::Tensor mat, torch::Tensor mul,
|
| 94 |
-
torch::Tensor scales, torch::Tensor zeros
|
| 95 |
-
) {
|
| 96 |
-
const at::cuda::OptionalCUDAGuard device_guard(device_of(vec));
|
| 97 |
-
vecquant4matmul_batched_old_cuda(vec, mat, mul, scales, zeros);
|
| 98 |
-
}
|
| 99 |
-
|
| 100 |
-
void vecquant8matmul_batched_column_compression_old_cuda(
|
| 101 |
-
torch::Tensor vec, torch::Tensor mat, torch::Tensor mul,
|
| 102 |
-
torch::Tensor scales, torch::Tensor zeros
|
| 103 |
-
);
|
| 104 |
-
|
| 105 |
-
void vecquant8matmul_batched_column_compression_old(
|
| 106 |
-
torch::Tensor vec, torch::Tensor mat, torch::Tensor mul,
|
| 107 |
-
torch::Tensor scales, torch::Tensor zeros
|
| 108 |
-
) {
|
| 109 |
-
const at::cuda::OptionalCUDAGuard device_guard(device_of(vec));
|
| 110 |
-
vecquant8matmul_batched_column_compression_old_cuda(vec, mat, mul, scales, zeros);
|
| 111 |
-
}
|
| 112 |
-
|
| 113 |
-
void vecquant4matmul_batched_column_compression_old_cuda(
|
| 114 |
-
torch::Tensor vec, torch::Tensor mat, torch::Tensor mul,
|
| 115 |
-
torch::Tensor scales, torch::Tensor zeros
|
| 116 |
-
);
|
| 117 |
-
|
| 118 |
-
void vecquant4matmul_batched_column_compression_old(
|
| 119 |
-
torch::Tensor vec, torch::Tensor mat, torch::Tensor mul,
|
| 120 |
-
torch::Tensor scales, torch::Tensor zeros
|
| 121 |
-
) {
|
| 122 |
-
const at::cuda::OptionalCUDAGuard device_guard(device_of(vec));
|
| 123 |
-
vecquant4matmul_batched_column_compression_old_cuda(vec, mat, mul, scales, zeros);
|
| 124 |
-
}
|
| 125 |
-
|
| 126 |
-
|
| 127 |
-
|
| 128 |
-
void vecquant8matmul_batched_faster_cuda(
|
| 129 |
-
torch::Tensor vec, torch::Tensor mat, torch::Tensor mul,
|
| 130 |
-
torch::Tensor scales, torch::Tensor zeros
|
| 131 |
-
);
|
| 132 |
-
|
| 133 |
-
void vecquant8matmul_batched_faster(
|
| 134 |
-
torch::Tensor vec, torch::Tensor mat, torch::Tensor mul,
|
| 135 |
-
torch::Tensor scales, torch::Tensor zeros
|
| 136 |
-
) {
|
| 137 |
-
const at::cuda::OptionalCUDAGuard device_guard(device_of(vec));
|
| 138 |
-
vecquant8matmul_batched_faster_cuda(vec, mat, mul, scales, zeros);
|
| 139 |
-
}
|
| 140 |
-
|
| 141 |
-
|
| 142 |
-
void vecquant8matmul_batched_faster_old_cuda(
|
| 143 |
-
torch::Tensor vec, torch::Tensor mat, torch::Tensor mul,
|
| 144 |
-
torch::Tensor scales, torch::Tensor zeros
|
| 145 |
-
);
|
| 146 |
-
|
| 147 |
-
void vecquant8matmul_batched_faster_old(
|
| 148 |
-
torch::Tensor vec, torch::Tensor mat, torch::Tensor mul,
|
| 149 |
-
torch::Tensor scales, torch::Tensor zeros
|
| 150 |
-
) {
|
| 151 |
-
const at::cuda::OptionalCUDAGuard device_guard(device_of(vec));
|
| 152 |
-
vecquant8matmul_batched_faster_old_cuda(vec, mat, mul, scales, zeros);
|
| 153 |
-
}
|
| 154 |
-
|
| 155 |
-
void vecquant8matmul_batched_column_compression_faster_cuda(
|
| 156 |
-
torch::Tensor vec, torch::Tensor mat, torch::Tensor mul,
|
| 157 |
-
torch::Tensor scales, torch::Tensor zeros
|
| 158 |
-
);
|
| 159 |
-
|
| 160 |
-
void vecquant8matmul_batched_column_compression_faster(
|
| 161 |
-
torch::Tensor vec, torch::Tensor mat, torch::Tensor mul,
|
| 162 |
-
torch::Tensor scales, torch::Tensor zeros
|
| 163 |
-
) {
|
| 164 |
-
const at::cuda::OptionalCUDAGuard device_guard(device_of(vec));
|
| 165 |
-
vecquant8matmul_batched_column_compression_faster_cuda(vec, mat, mul, scales, zeros);
|
| 166 |
-
}
|
| 167 |
-
|
| 168 |
-
|
| 169 |
-
void vecquant8matmul_batched_column_compression_faster_old_cuda(
|
| 170 |
-
torch::Tensor vec, torch::Tensor mat, torch::Tensor mul,
|
| 171 |
-
torch::Tensor scales, torch::Tensor zeros
|
| 172 |
-
);
|
| 173 |
-
|
| 174 |
-
void vecquant8matmul_batched_column_compression_faster_old(
|
| 175 |
-
torch::Tensor vec, torch::Tensor mat, torch::Tensor mul,
|
| 176 |
-
torch::Tensor scales, torch::Tensor zeros
|
| 177 |
-
) {
|
| 178 |
-
const at::cuda::OptionalCUDAGuard device_guard(device_of(vec));
|
| 179 |
-
vecquant8matmul_batched_column_compression_faster_old_cuda(vec, mat, mul, scales, zeros);
|
| 180 |
-
}
|
| 181 |
-
|
| 182 |
-
|
| 183 |
-
|
| 184 |
-
PYBIND11_MODULE(TORCH_EXTENSION_NAME, m) {
|
| 185 |
-
m.def("vecquant8matmul", &vecquant8matmul, "Vector 8-bit Quantized Matrix Multiplication (CUDA) (desc_act)");
|
| 186 |
-
m.def("vecquant8matmul_batched", &vecquant8matmul_batched, "Vector 8-bit Batched Quantized Matrix Multiplication (CUDA) (desc_act)");
|
| 187 |
-
m.def("vecquant8matmul_batched_old", &vecquant8matmul_batched_old, "Vector 8-bit old Batched Quantized Matrix Multiplication (CUDA) (desc_act)");
|
| 188 |
-
m.def("vecquant8matmul_batched_faster", &vecquant8matmul_batched_faster, "Vector 8-bit old Batched Quantized Matrix Multiplication (CUDA) (desc_act)");
|
| 189 |
-
m.def("vecquant8matmul_batched_faster_old", &vecquant8matmul_batched_faster_old, "Vector 8-bit old Batched Quantized Matrix Multiplication (CUDA) (desc_act)");
|
| 190 |
-
m.def("vecquant4matmul_batched_old", &vecquant4matmul_batched_old, "Vector 4-bit old Batched Quantized Matrix Multiplication (CUDA) (desc_act)");
|
| 191 |
-
m.def("vecquant8matmul_batched_column_compression", &vecquant8matmul_batched_column_compression, "Vector 8-bit Batched Quantized Matrix Multiplication (CUDA) with weight's column compressed (desc_act)");
|
| 192 |
-
m.def("vecquant8matmul_batched_column_compression_old", &vecquant8matmul_batched_column_compression_old, "Vector old 8-bit Batched Quantized Matrix Multiplication (CUDA) with weight's column compressed (desc_act)");
|
| 193 |
-
m.def("vecquant8matmul_batched_column_compression_faster", &vecquant8matmul_batched_column_compression_faster, "Vector old 8-bit Batched Quantized Matrix Multiplication (CUDA) with weight's column compressed (desc_act)");
|
| 194 |
-
m.def("vecquant8matmul_batched_column_compression_faster_old", &vecquant8matmul_batched_column_compression_faster_old, "Vector old 8-bit Batched Quantized Matrix Multiplication (CUDA) with weight's column compressed (desc_act)");
|
| 195 |
-
m.def("vecquant4matmul_batched_column_compression_old", &vecquant4matmul_batched_column_compression_old, "Vector old 4-bit Batched Quantized Matrix Multiplication (CUDA) with weight's column compressed (desc_act)");
|
| 196 |
-
m.def("vecquant4matmul_batched", &vecquant4matmul_batched, "Vector 4-bit Batched Quantized Matrix Multiplication (CUDA) (desc_act)");
|
| 197 |
-
m.def("vecquant4matmul_batched_column_compression", &vecquant4matmul_batched_column_compression, "Vector 4-bit Batched Quantized Matrix Multiplication (CUDA) with weight's column compressed (desc_act)");
|
| 198 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
cache_autogptq_cuda_kernel_256.cu
DELETED
|
@@ -1,1708 +0,0 @@
|
|
| 1 |
-
#define _CRT_SECURE_NO_WARNINGS
|
| 2 |
-
#include <torch/all.h>
|
| 3 |
-
#include <torch/python.h>
|
| 4 |
-
#include <cuda.h>
|
| 5 |
-
#include <cuda_runtime.h>
|
| 6 |
-
#include <cuda_fp16.h>
|
| 7 |
-
#include <stdint.h>
|
| 8 |
-
|
| 9 |
-
#if (defined(__CUDA_ARCH__) && __CUDA_ARCH__ < 700) || defined(USE_ROCM)
|
| 10 |
-
// adapted from https://github.com/PanQiWei/AutoGPTQ/blob/main/autogptq_extension/cuda_256/autogptq_cuda_kernel_256.cu
|
| 11 |
-
__device__ __forceinline__ void atomicAdd(c10::Half* address, c10::Half val) {
|
| 12 |
-
unsigned int *address_as_ui = reinterpret_cast<unsigned int *>(reinterpret_cast<char *>(address) - (reinterpret_cast<size_t>(address) & 2));
|
| 13 |
-
unsigned int old = *address_as_ui;
|
| 14 |
-
unsigned int assumed;
|
| 15 |
-
|
| 16 |
-
do {
|
| 17 |
-
assumed = old;
|
| 18 |
-
unsigned short hsum = reinterpret_cast<size_t>(address) & 2 ? (old >> 16) : (old & 0xffff);
|
| 19 |
-
hsum += val;
|
| 20 |
-
old = reinterpret_cast<size_t>(address) & 2
|
| 21 |
-
? (old & 0xffff) | (hsum << 16)
|
| 22 |
-
: (old & 0xffff0000) | hsum;
|
| 23 |
-
old = atomicCAS(address_as_ui, assumed, old);
|
| 24 |
-
|
| 25 |
-
// Note: uses integer comparison to avoid hang in case of NaN (since NaN != NaN)
|
| 26 |
-
} while (assumed != old);
|
| 27 |
-
}
|
| 28 |
-
__device__ __forceinline__ void atomicAdd(__half* address, c10::Half val) {
|
| 29 |
-
unsigned int * address_as_ui = (unsigned int *) ((char *)address - ((size_t)address & 2));
|
| 30 |
-
unsigned int old = *address_as_ui;
|
| 31 |
-
unsigned int assumed;
|
| 32 |
-
|
| 33 |
-
do {
|
| 34 |
-
assumed = old;
|
| 35 |
-
__half_raw hsum;
|
| 36 |
-
hsum.x = (size_t)address & 2 ? (old >> 16) : (old & 0xffff);
|
| 37 |
-
half tmpres = __hadd(hsum, val);
|
| 38 |
-
hsum = __half_raw(tmpres);
|
| 39 |
-
old = (size_t)address & 2 ? (old & 0xffff) | (hsum.x << 16) : (old & 0xffff0000) | hsum.x;
|
| 40 |
-
old = atomicCAS(address_as_ui, assumed, old);
|
| 41 |
-
} while (assumed != old);
|
| 42 |
-
}
|
| 43 |
-
#endif
|
| 44 |
-
|
| 45 |
-
template <typename scalar_t>
|
| 46 |
-
__global__ void VecQuant8MatMulKernel(
|
| 47 |
-
const scalar_t* __restrict__ vec,
|
| 48 |
-
const int* __restrict__ mat,
|
| 49 |
-
scalar_t* __restrict__ mul,
|
| 50 |
-
const scalar_t* __restrict__ scales,
|
| 51 |
-
const int* __restrict__ zeros,
|
| 52 |
-
const int* __restrict__ g_idx,
|
| 53 |
-
int batch,
|
| 54 |
-
int vec_height,
|
| 55 |
-
int height,
|
| 56 |
-
int width,
|
| 57 |
-
int zero_width
|
| 58 |
-
);
|
| 59 |
-
|
| 60 |
-
template <typename scalar_t>
|
| 61 |
-
__global__ void VecQuant8BatchMatMulColumnCompressionKernel(
|
| 62 |
-
const scalar_t* __restrict__ vec,
|
| 63 |
-
const int* __restrict__ mat,
|
| 64 |
-
scalar_t* __restrict__ mul,
|
| 65 |
-
const scalar_t* __restrict__ scales,
|
| 66 |
-
const int* __restrict__ zeros,
|
| 67 |
-
int batch,
|
| 68 |
-
int heads,
|
| 69 |
-
int vec_row,
|
| 70 |
-
int height,
|
| 71 |
-
int width
|
| 72 |
-
);
|
| 73 |
-
|
| 74 |
-
template <typename scalar_t>
|
| 75 |
-
__global__ void VecQuant4BatchMatMulColumnCompressionKernel(
|
| 76 |
-
const scalar_t* __restrict__ vec,
|
| 77 |
-
const int* __restrict__ mat,
|
| 78 |
-
scalar_t* __restrict__ mul,
|
| 79 |
-
const scalar_t* __restrict__ scales,
|
| 80 |
-
const int* __restrict__ zeros,
|
| 81 |
-
int batch,
|
| 82 |
-
int heads,
|
| 83 |
-
int vec_row,
|
| 84 |
-
int height,
|
| 85 |
-
int width
|
| 86 |
-
);
|
| 87 |
-
|
| 88 |
-
template <typename scalar_t>
|
| 89 |
-
__global__ void VecQuant8BatchMatMulKernel(
|
| 90 |
-
const scalar_t* __restrict__ vec,
|
| 91 |
-
const int* __restrict__ mat,
|
| 92 |
-
scalar_t* __restrict__ mul,
|
| 93 |
-
const scalar_t* __restrict__ scales,
|
| 94 |
-
const int* __restrict__ zeros,
|
| 95 |
-
int batch,
|
| 96 |
-
int heads,
|
| 97 |
-
int vec_row,
|
| 98 |
-
int vec_height,
|
| 99 |
-
int height,
|
| 100 |
-
int width,
|
| 101 |
-
int zero_width
|
| 102 |
-
);
|
| 103 |
-
|
| 104 |
-
template <typename scalar_t>
|
| 105 |
-
__global__ void VecQuant4BatchMatMulKernel(
|
| 106 |
-
const scalar_t* __restrict__ vec,
|
| 107 |
-
const int* __restrict__ mat,
|
| 108 |
-
scalar_t* __restrict__ mul,
|
| 109 |
-
const scalar_t* __restrict__ scales,
|
| 110 |
-
const int* __restrict__ zeros,
|
| 111 |
-
int batch,
|
| 112 |
-
int heads,
|
| 113 |
-
int vec_row,
|
| 114 |
-
int vec_height,
|
| 115 |
-
int height,
|
| 116 |
-
int width,
|
| 117 |
-
int zero_width
|
| 118 |
-
);
|
| 119 |
-
|
| 120 |
-
|
| 121 |
-
|
| 122 |
-
template <typename scalar_t>
|
| 123 |
-
__global__ void VecQuant8BatchMatMulKernel_old(
|
| 124 |
-
const scalar_t* __restrict__ vec,
|
| 125 |
-
const uint8_t* __restrict__ mat,
|
| 126 |
-
scalar_t* __restrict__ mul,
|
| 127 |
-
const scalar_t* __restrict__ scales,
|
| 128 |
-
const scalar_t* __restrict__ zeros,
|
| 129 |
-
int batch,
|
| 130 |
-
int heads,
|
| 131 |
-
int vec_row,
|
| 132 |
-
int vec_height,
|
| 133 |
-
int height,
|
| 134 |
-
int width,
|
| 135 |
-
int zero_width
|
| 136 |
-
);
|
| 137 |
-
|
| 138 |
-
__global__ void VecQuant8BatchMatMulKernel_faster(
|
| 139 |
-
const half* __restrict__ vec,
|
| 140 |
-
const uint8_t* __restrict__ mat,
|
| 141 |
-
half* __restrict__ mul,
|
| 142 |
-
const half* __restrict__ scales,
|
| 143 |
-
const half* __restrict__ zeros,
|
| 144 |
-
int batch,
|
| 145 |
-
int heads,
|
| 146 |
-
int vec_row,
|
| 147 |
-
int vec_height,
|
| 148 |
-
int height,
|
| 149 |
-
int width,
|
| 150 |
-
int zero_width
|
| 151 |
-
);
|
| 152 |
-
|
| 153 |
-
|
| 154 |
-
|
| 155 |
-
__global__ void VecQuant8BatchMatMulKernel_faster_old(
|
| 156 |
-
const half* __restrict__ vec,
|
| 157 |
-
const uint8_t* __restrict__ mat,
|
| 158 |
-
half* __restrict__ mul,
|
| 159 |
-
const half* __restrict__ scales,
|
| 160 |
-
const half* __restrict__ zeros,
|
| 161 |
-
int batch,
|
| 162 |
-
int heads,
|
| 163 |
-
int vec_row,
|
| 164 |
-
int vec_height,
|
| 165 |
-
int height,
|
| 166 |
-
int width
|
| 167 |
-
);
|
| 168 |
-
|
| 169 |
-
|
| 170 |
-
template <typename scalar_t>
|
| 171 |
-
__global__ void VecQuant4BatchMatMulKernel_old(
|
| 172 |
-
const scalar_t* __restrict__ vec,
|
| 173 |
-
const uint8_t* __restrict__ mat,
|
| 174 |
-
scalar_t* __restrict__ mul,
|
| 175 |
-
const scalar_t* __restrict__ scales,
|
| 176 |
-
const scalar_t* __restrict__ zeros,
|
| 177 |
-
int batch,
|
| 178 |
-
int heads,
|
| 179 |
-
int vec_row,
|
| 180 |
-
int vec_height,
|
| 181 |
-
int height,
|
| 182 |
-
int width,
|
| 183 |
-
int zero_width
|
| 184 |
-
);
|
| 185 |
-
|
| 186 |
-
|
| 187 |
-
template <typename scalar_t>
|
| 188 |
-
__global__ void VecQuant8BatchMatMulColumnCompressionKernel_old(
|
| 189 |
-
const scalar_t* __restrict__ vec,
|
| 190 |
-
const uint8_t* __restrict__ mat,
|
| 191 |
-
scalar_t* __restrict__ mul,
|
| 192 |
-
const scalar_t* __restrict__ scales,
|
| 193 |
-
const scalar_t* __restrict__ zeros,
|
| 194 |
-
int batch,
|
| 195 |
-
int heads,
|
| 196 |
-
int vec_row,
|
| 197 |
-
int height,
|
| 198 |
-
int width
|
| 199 |
-
);
|
| 200 |
-
|
| 201 |
-
__global__ void VecQuant8BatchMatMulColumnCompressionKernel_faster(
|
| 202 |
-
const half* __restrict__ vec,
|
| 203 |
-
const uint8_t* __restrict__ mat,
|
| 204 |
-
half* __restrict__ mul,
|
| 205 |
-
const half* __restrict__ scales,
|
| 206 |
-
const half* __restrict__ zeros,
|
| 207 |
-
int batch,
|
| 208 |
-
int heads,
|
| 209 |
-
int vec_row,
|
| 210 |
-
int height,
|
| 211 |
-
int width
|
| 212 |
-
);
|
| 213 |
-
|
| 214 |
-
__global__ void VecQuant8BatchMatMulColumnCompressionKernel_faster_old(
|
| 215 |
-
const half* __restrict__ vec,
|
| 216 |
-
const uint8_t* __restrict__ mat,
|
| 217 |
-
half* __restrict__ mul,
|
| 218 |
-
const half* __restrict__ scales,
|
| 219 |
-
const half* __restrict__ zeros,
|
| 220 |
-
int batch,
|
| 221 |
-
int heads,
|
| 222 |
-
int vec_row,
|
| 223 |
-
int height,
|
| 224 |
-
int width
|
| 225 |
-
);
|
| 226 |
-
|
| 227 |
-
|
| 228 |
-
template <typename scalar_t>
|
| 229 |
-
__global__ void VecQuant4BatchMatMulColumnCompressionKernel_old(
|
| 230 |
-
const scalar_t* __restrict__ vec,
|
| 231 |
-
const uint8_t* __restrict__ mat,
|
| 232 |
-
scalar_t* __restrict__ mul,
|
| 233 |
-
const scalar_t* __restrict__ scales,
|
| 234 |
-
const scalar_t* __restrict__ zeros,
|
| 235 |
-
int batch,
|
| 236 |
-
int heads,
|
| 237 |
-
int vec_row,
|
| 238 |
-
int height,
|
| 239 |
-
int width
|
| 240 |
-
);
|
| 241 |
-
|
| 242 |
-
|
| 243 |
-
__global__ void VecQuant8BatchMatMulKernel_faster(
|
| 244 |
-
const half* __restrict__ vec,
|
| 245 |
-
const uint8_t* __restrict__ mat,
|
| 246 |
-
half* __restrict__ mul,
|
| 247 |
-
const half* __restrict__ scales,
|
| 248 |
-
const half* __restrict__ zeros,
|
| 249 |
-
int batch,
|
| 250 |
-
int heads,
|
| 251 |
-
int vec_row,
|
| 252 |
-
int vec_height,
|
| 253 |
-
int height,
|
| 254 |
-
int width
|
| 255 |
-
);
|
| 256 |
-
|
| 257 |
-
|
| 258 |
-
__global__ void VecQuant8BatchMatMulColumnCompressionKernel_faster(
|
| 259 |
-
const half* __restrict__ vec,
|
| 260 |
-
const uint8_t* __restrict__ mat,
|
| 261 |
-
half* __restrict__ mul,
|
| 262 |
-
const half* __restrict__ scales,
|
| 263 |
-
const half* __restrict__ zeros,
|
| 264 |
-
int batch,
|
| 265 |
-
int heads,
|
| 266 |
-
int vec_row,
|
| 267 |
-
int height,
|
| 268 |
-
int width
|
| 269 |
-
);
|
| 270 |
-
|
| 271 |
-
const int BLOCKWIDTH = 128;
|
| 272 |
-
const int BLOCKHEIGHT8 = 32;
|
| 273 |
-
const int BLOCKHEIGHT4 = 16;
|
| 274 |
-
const int BLOCKHEIGHT_OLD4 = 128;
|
| 275 |
-
//const int BLOCKHEIGHT_OLD8 = 128;
|
| 276 |
-
|
| 277 |
-
__device__ inline unsigned int as_unsigned(int i) {
|
| 278 |
-
return *reinterpret_cast<unsigned int*>(&i);
|
| 279 |
-
}
|
| 280 |
-
|
| 281 |
-
__device__ inline int as_int(int i) {
|
| 282 |
-
return *reinterpret_cast<int*>(&i);
|
| 283 |
-
}
|
| 284 |
-
|
| 285 |
-
void vecquant8matmul_batched_column_compression_cuda(
|
| 286 |
-
torch::Tensor vec,
|
| 287 |
-
torch::Tensor mat,
|
| 288 |
-
torch::Tensor mul,
|
| 289 |
-
torch::Tensor scales,
|
| 290 |
-
torch::Tensor zeros
|
| 291 |
-
) {
|
| 292 |
-
int batch = vec.size(0);
|
| 293 |
-
int heads = vec.size(1);
|
| 294 |
-
int vec_row = vec.size(2);
|
| 295 |
-
int height = vec.size(3);
|
| 296 |
-
int width = mat.size(3) * 4;
|
| 297 |
-
|
| 298 |
-
dim3 blocks(
|
| 299 |
-
(height + BLOCKWIDTH - 1) / BLOCKWIDTH,
|
| 300 |
-
(width + BLOCKWIDTH - 1) / BLOCKWIDTH
|
| 301 |
-
);
|
| 302 |
-
dim3 threads(BLOCKWIDTH);
|
| 303 |
-
|
| 304 |
-
AT_DISPATCH_FLOATING_TYPES(
|
| 305 |
-
vec.type(), "vecquant8matmul_batched_cuda", ([&] {
|
| 306 |
-
VecQuant8BatchMatMulColumnCompressionKernel<<<blocks, threads>>>(
|
| 307 |
-
vec.data<scalar_t>(), mat.data<int>(), mul.data<scalar_t>(),
|
| 308 |
-
scales.data<scalar_t>(), zeros.data<int>(),
|
| 309 |
-
batch, heads, vec_row, height, width
|
| 310 |
-
);
|
| 311 |
-
})
|
| 312 |
-
);
|
| 313 |
-
|
| 314 |
-
}
|
| 315 |
-
|
| 316 |
-
template <typename scalar_t>
|
| 317 |
-
__global__ void VecQuant8BatchMatMulColumnCompressionKernel(
|
| 318 |
-
const scalar_t* __restrict__ vec,
|
| 319 |
-
const int* __restrict__ mat,
|
| 320 |
-
scalar_t* __restrict__ mul,
|
| 321 |
-
const scalar_t* __restrict__ scales,
|
| 322 |
-
const int* __restrict__ zeros,
|
| 323 |
-
int batch,
|
| 324 |
-
int heads,
|
| 325 |
-
int vec_row,
|
| 326 |
-
int height,
|
| 327 |
-
int width
|
| 328 |
-
) {
|
| 329 |
-
int weight_total = batch * heads * height * width / 4;
|
| 330 |
-
int input_total = batch * heads * vec_row * height;
|
| 331 |
-
int out_total = batch * heads * vec_row * width;
|
| 332 |
-
int tid = threadIdx.x;
|
| 333 |
-
// h is index of height with step being BLOCKWIDTH
|
| 334 |
-
int h = BLOCKWIDTH * blockIdx.x;
|
| 335 |
-
// w is index of width with step being 1
|
| 336 |
-
int w = BLOCKWIDTH * blockIdx.y + tid;
|
| 337 |
-
if (w >= width && tid >= height) {
|
| 338 |
-
return;
|
| 339 |
-
}
|
| 340 |
-
|
| 341 |
-
__shared__ scalar_t blockvec[BLOCKWIDTH];
|
| 342 |
-
int k;
|
| 343 |
-
scalar_t w_tmp;
|
| 344 |
-
|
| 345 |
-
float weight[BLOCKWIDTH];
|
| 346 |
-
|
| 347 |
-
for (int b = 0; b < batch; ++b){
|
| 348 |
-
for (int head = 0; head < heads; ++head){
|
| 349 |
-
int batch_shift = b * heads + head;
|
| 350 |
-
for (k = 0; k < BLOCKWIDTH && h + k < height; ++k){
|
| 351 |
-
int i_w = (w / 4);
|
| 352 |
-
int w_bit = (w % 4) * 8;
|
| 353 |
-
|
| 354 |
-
int w_index = (batch_shift * height + h + k) * width / 4 + i_w;
|
| 355 |
-
if (w_index >= weight_total || w >= width) {
|
| 356 |
-
weight[k] = 0;
|
| 357 |
-
} else {
|
| 358 |
-
scalar_t scale = scales[batch_shift * height + h + k];
|
| 359 |
-
scalar_t zero = zeros[batch_shift * height + h + k];
|
| 360 |
-
w_tmp = ((as_unsigned(mat[w_index]) >> w_bit) & 0xFF);
|
| 361 |
-
weight[k] = scale * (w_tmp - zero);
|
| 362 |
-
}
|
| 363 |
-
}
|
| 364 |
-
|
| 365 |
-
scalar_t res;
|
| 366 |
-
for (int vr = 0; vr < vec_row; ++vr){
|
| 367 |
-
res = 0;
|
| 368 |
-
int vec_index = (batch_shift * vec_row + vr) * height + blockIdx.x * BLOCKWIDTH + tid;
|
| 369 |
-
if (vec_index < input_total) {
|
| 370 |
-
blockvec[tid] = vec[vec_index];
|
| 371 |
-
} else {
|
| 372 |
-
blockvec[tid] = 0;
|
| 373 |
-
}
|
| 374 |
-
|
| 375 |
-
__syncthreads();
|
| 376 |
-
for (k = 0; k < BLOCKWIDTH && h + k < height; ++k){
|
| 377 |
-
// res is the dot product of BLOCKWIDTH elements (part of width)
|
| 378 |
-
res += weight[k] * blockvec[k];
|
| 379 |
-
}
|
| 380 |
-
// add res to the final result, final matrix shape: (batch, vec_row, width)
|
| 381 |
-
int out_index = (batch_shift * vec_row + vr) * width + w;
|
| 382 |
-
if (out_index < out_total) {
|
| 383 |
-
atomicAdd(&mul[out_index], res);
|
| 384 |
-
}
|
| 385 |
-
__syncthreads();
|
| 386 |
-
}
|
| 387 |
-
}
|
| 388 |
-
}
|
| 389 |
-
}
|
| 390 |
-
|
| 391 |
-
void vecquant8matmul_batched_cuda(
|
| 392 |
-
torch::Tensor vec,
|
| 393 |
-
torch::Tensor mat,
|
| 394 |
-
torch::Tensor mul,
|
| 395 |
-
torch::Tensor scales,
|
| 396 |
-
torch::Tensor zeros
|
| 397 |
-
) {
|
| 398 |
-
int batch = vec.size(0);
|
| 399 |
-
int heads = vec.size(1);
|
| 400 |
-
int vec_row = vec.size(2);
|
| 401 |
-
int vec_height = vec.size(3);
|
| 402 |
-
int height = mat.size(2);
|
| 403 |
-
int width = mat.size(3);
|
| 404 |
-
int zero_width = zeros.size(2);
|
| 405 |
-
|
| 406 |
-
dim3 blocks(
|
| 407 |
-
(height + BLOCKHEIGHT8 - 1) / BLOCKHEIGHT8,
|
| 408 |
-
(width + BLOCKWIDTH - 1) / BLOCKWIDTH
|
| 409 |
-
);
|
| 410 |
-
dim3 threads(BLOCKWIDTH);
|
| 411 |
-
|
| 412 |
-
AT_DISPATCH_FLOATING_TYPES(
|
| 413 |
-
vec.type(), "vecquant8matmul_batched_cuda", ([&] {
|
| 414 |
-
VecQuant8BatchMatMulKernel<<<blocks, threads>>>(
|
| 415 |
-
vec.data<scalar_t>(), mat.data<int>(), mul.data<scalar_t>(),
|
| 416 |
-
scales.data<scalar_t>(), zeros.data<int>(),
|
| 417 |
-
batch, heads, vec_row, vec_height, height, width, zero_width
|
| 418 |
-
);
|
| 419 |
-
})
|
| 420 |
-
);
|
| 421 |
-
|
| 422 |
-
}
|
| 423 |
-
|
| 424 |
-
template <typename scalar_t>
|
| 425 |
-
__global__ void VecQuant8BatchMatMulKernel(
|
| 426 |
-
const scalar_t* __restrict__ vec,
|
| 427 |
-
const int* __restrict__ mat,
|
| 428 |
-
scalar_t* __restrict__ mul,
|
| 429 |
-
const scalar_t* __restrict__ scales,
|
| 430 |
-
const int* __restrict__ zeros,
|
| 431 |
-
int batch,
|
| 432 |
-
int heads,
|
| 433 |
-
int vec_row,
|
| 434 |
-
int vec_height,
|
| 435 |
-
int height,
|
| 436 |
-
int width,
|
| 437 |
-
int zero_width
|
| 438 |
-
) {
|
| 439 |
-
int weight_total = batch * heads * height * width;
|
| 440 |
-
int input_total = batch * heads * vec_row * vec_height;
|
| 441 |
-
int out_total = batch * heads * vec_row * width;
|
| 442 |
-
int tid = threadIdx.x;
|
| 443 |
-
// h is index of height with step being BLOCKHEIGHT8
|
| 444 |
-
int h = BLOCKHEIGHT8 * blockIdx.x;
|
| 445 |
-
// w is index of width with step being 1
|
| 446 |
-
int w = BLOCKWIDTH * blockIdx.y + tid;
|
| 447 |
-
if (w >= width && tid >= vec_height) {
|
| 448 |
-
return;
|
| 449 |
-
}
|
| 450 |
-
|
| 451 |
-
__shared__ scalar_t blockvec[BLOCKWIDTH];
|
| 452 |
-
// i is index of mat of block first row
|
| 453 |
-
int i = width * h + w;
|
| 454 |
-
// if (i >= width * height) {
|
| 455 |
-
// return;
|
| 456 |
-
// }
|
| 457 |
-
int k;
|
| 458 |
-
scalar_t w_tmp;
|
| 459 |
-
|
| 460 |
-
int z_w = w / 4;
|
| 461 |
-
int z_mod = (w % 4) * 8;
|
| 462 |
-
|
| 463 |
-
float weight[BLOCKWIDTH];
|
| 464 |
-
|
| 465 |
-
for (int b = 0; b < batch; ++b){
|
| 466 |
-
for (int head = 0; head < heads; ++head){
|
| 467 |
-
int batch_shift = b * heads + head;
|
| 468 |
-
for (k = 0; k < BLOCKWIDTH && h * 4 + k < vec_height; ++k){
|
| 469 |
-
int k_w = (k / 4);
|
| 470 |
-
int k_bit = (k % 4) * 8;
|
| 471 |
-
|
| 472 |
-
int w_index = batch_shift * height * width + i + (k_w * width);
|
| 473 |
-
if (w_index >= weight_total || w >= width) {
|
| 474 |
-
weight[k] = 0;
|
| 475 |
-
} else {
|
| 476 |
-
scalar_t scale = scales[batch_shift * width + w];
|
| 477 |
-
scalar_t zero;
|
| 478 |
-
if (zero_width == width) {
|
| 479 |
-
zero = zeros[batch_shift * width + w];
|
| 480 |
-
} else {
|
| 481 |
-
zero = scalar_t(((as_unsigned(zeros[batch_shift * zero_width + z_w]) >> z_mod) & 0xFF) + 1);
|
| 482 |
-
}
|
| 483 |
-
w_tmp = ((as_unsigned(mat[w_index]) >> k_bit) & 0xFF);
|
| 484 |
-
weight[k] = scale * (w_tmp - zero);
|
| 485 |
-
}
|
| 486 |
-
}
|
| 487 |
-
|
| 488 |
-
scalar_t res;
|
| 489 |
-
for (int vr = 0; vr < vec_row; ++vr){
|
| 490 |
-
res = 0;
|
| 491 |
-
int vec_index = (batch_shift * vec_row + vr) * vec_height + blockIdx.x * BLOCKWIDTH + tid;
|
| 492 |
-
if (vec_index < input_total) {
|
| 493 |
-
blockvec[tid] = vec[vec_index];
|
| 494 |
-
} else {
|
| 495 |
-
blockvec[tid] = 0;
|
| 496 |
-
}
|
| 497 |
-
|
| 498 |
-
__syncthreads();
|
| 499 |
-
for (k = 0; k < BLOCKWIDTH && h * 4 + k < vec_height; ++k){
|
| 500 |
-
// res is the dot product of BLOCKWIDTH elements (part of width)
|
| 501 |
-
res += weight[k] * blockvec[k];
|
| 502 |
-
}
|
| 503 |
-
// add res to the final result, final matrix shape: (batch, vec_row, width)
|
| 504 |
-
int out_index = (batch_shift * vec_row + vr) * width + w;
|
| 505 |
-
if (out_index < out_total) {
|
| 506 |
-
atomicAdd(&mul[out_index], res);
|
| 507 |
-
}
|
| 508 |
-
__syncthreads();
|
| 509 |
-
}
|
| 510 |
-
}
|
| 511 |
-
}
|
| 512 |
-
}
|
| 513 |
-
|
| 514 |
-
|
| 515 |
-
void vecquant8matmul_cuda(
|
| 516 |
-
torch::Tensor vec,
|
| 517 |
-
torch::Tensor mat,
|
| 518 |
-
torch::Tensor mul,
|
| 519 |
-
torch::Tensor scales,
|
| 520 |
-
torch::Tensor zeros,
|
| 521 |
-
torch::Tensor g_idx
|
| 522 |
-
) {
|
| 523 |
-
int batch = vec.size(0);
|
| 524 |
-
int vec_height = vec.size(1);
|
| 525 |
-
int height = mat.size(0);
|
| 526 |
-
int width = mat.size(1);
|
| 527 |
-
int zero_width = zeros.size(1);
|
| 528 |
-
|
| 529 |
-
dim3 blocks(
|
| 530 |
-
(height + BLOCKHEIGHT8 - 1) / BLOCKHEIGHT8,
|
| 531 |
-
(width + BLOCKWIDTH - 1) / BLOCKWIDTH
|
| 532 |
-
);
|
| 533 |
-
dim3 threads(BLOCKWIDTH);
|
| 534 |
-
|
| 535 |
-
AT_DISPATCH_FLOATING_TYPES(
|
| 536 |
-
vec.type(), "vecquant8matmul_cuda", ([&] {
|
| 537 |
-
VecQuant8MatMulKernel<<<blocks, threads>>>(
|
| 538 |
-
vec.data<scalar_t>(), mat.data<int>(), mul.data<scalar_t>(),
|
| 539 |
-
scales.data<scalar_t>(), zeros.data<int>(), g_idx.data<int>(),
|
| 540 |
-
batch, vec_height, height, width, zero_width
|
| 541 |
-
);
|
| 542 |
-
})
|
| 543 |
-
);
|
| 544 |
-
}
|
| 545 |
-
|
| 546 |
-
template <typename scalar_t>
|
| 547 |
-
__global__ void VecQuant8MatMulKernel(
|
| 548 |
-
const scalar_t* __restrict__ vec,
|
| 549 |
-
const int* __restrict__ mat,
|
| 550 |
-
scalar_t* __restrict__ mul,
|
| 551 |
-
const scalar_t* __restrict__ scales,
|
| 552 |
-
const int* __restrict__ zeros,
|
| 553 |
-
const int* __restrict__ g_idx,
|
| 554 |
-
int batch,
|
| 555 |
-
int vec_height,
|
| 556 |
-
int height,
|
| 557 |
-
int width,
|
| 558 |
-
int zero_width
|
| 559 |
-
) {
|
| 560 |
-
int h = BLOCKHEIGHT8 * blockIdx.x;
|
| 561 |
-
int w = BLOCKWIDTH * blockIdx.y + threadIdx.x;
|
| 562 |
-
|
| 563 |
-
__shared__ scalar_t blockvec[BLOCKWIDTH];
|
| 564 |
-
int i = width * h + w;
|
| 565 |
-
int g_h = h * 4;
|
| 566 |
-
int k;
|
| 567 |
-
unsigned int g;
|
| 568 |
-
scalar_t w_tmp;
|
| 569 |
-
|
| 570 |
-
int z_w = w / 4;
|
| 571 |
-
int z_mod = (w % 4) * 8;
|
| 572 |
-
|
| 573 |
-
float weight[BLOCKWIDTH];
|
| 574 |
-
|
| 575 |
-
for (k = 0; k < BLOCKWIDTH; ++k){
|
| 576 |
-
int k_w = (k / 4);
|
| 577 |
-
int k_bit = (k % 4) * 8;
|
| 578 |
-
|
| 579 |
-
g = as_int(g_idx[g_h + k]);
|
| 580 |
-
scalar_t scale = scales[g * width + w];
|
| 581 |
-
scalar_t zero = scalar_t(((as_unsigned(zeros[g * zero_width + z_w]) >> z_mod) & 0xFF) + 1);
|
| 582 |
-
|
| 583 |
-
w_tmp = ((as_unsigned(mat[i + (k_w * width)]) >> k_bit) & 0xFF);
|
| 584 |
-
|
| 585 |
-
weight[k] = scale * (w_tmp - zero);
|
| 586 |
-
}
|
| 587 |
-
|
| 588 |
-
|
| 589 |
-
scalar_t res;
|
| 590 |
-
for (int b = 0; b < batch; ++b){
|
| 591 |
-
res = 0;
|
| 592 |
-
blockvec[threadIdx.x] = vec[b * vec_height + blockIdx.x * BLOCKWIDTH + threadIdx.x];
|
| 593 |
-
__syncthreads();
|
| 594 |
-
for (k = 0; k < BLOCKWIDTH; ++k){
|
| 595 |
-
res += weight[k] * blockvec[k];
|
| 596 |
-
}
|
| 597 |
-
atomicAdd(&mul[b * width + w], res);
|
| 598 |
-
__syncthreads();
|
| 599 |
-
}
|
| 600 |
-
}
|
| 601 |
-
|
| 602 |
-
|
| 603 |
-
|
| 604 |
-
void vecquant4matmul_batched_cuda(
|
| 605 |
-
torch::Tensor vec,
|
| 606 |
-
torch::Tensor mat,
|
| 607 |
-
torch::Tensor mul,
|
| 608 |
-
torch::Tensor scales,
|
| 609 |
-
torch::Tensor zeros
|
| 610 |
-
) {
|
| 611 |
-
int batch = vec.size(0);
|
| 612 |
-
int heads = vec.size(1);
|
| 613 |
-
int vec_row = vec.size(2);
|
| 614 |
-
int vec_height = vec.size(3);
|
| 615 |
-
int height = mat.size(2);
|
| 616 |
-
int width = mat.size(3);
|
| 617 |
-
int zero_width = zeros.size(2);
|
| 618 |
-
|
| 619 |
-
dim3 blocks(
|
| 620 |
-
(height + BLOCKHEIGHT4 - 1) / BLOCKHEIGHT4,
|
| 621 |
-
(width + BLOCKWIDTH - 1) / BLOCKWIDTH
|
| 622 |
-
);
|
| 623 |
-
dim3 threads(BLOCKWIDTH);
|
| 624 |
-
|
| 625 |
-
AT_DISPATCH_FLOATING_TYPES(
|
| 626 |
-
vec.type(), "vecquant4matmul_batched_cuda", ([&] {
|
| 627 |
-
VecQuant4BatchMatMulKernel<<<blocks, threads>>>(
|
| 628 |
-
vec.data<scalar_t>(), mat.data<int>(), mul.data<scalar_t>(),
|
| 629 |
-
scales.data<scalar_t>(), zeros.data<int>(),
|
| 630 |
-
batch, heads, vec_row, vec_height, height, width, zero_width
|
| 631 |
-
);
|
| 632 |
-
})
|
| 633 |
-
);
|
| 634 |
-
|
| 635 |
-
}
|
| 636 |
-
|
| 637 |
-
template <typename scalar_t>
|
| 638 |
-
__global__ void VecQuant4BatchMatMulKernel(
|
| 639 |
-
const scalar_t* __restrict__ vec,
|
| 640 |
-
const int* __restrict__ mat,
|
| 641 |
-
scalar_t* __restrict__ mul,
|
| 642 |
-
const scalar_t* __restrict__ scales,
|
| 643 |
-
const int* __restrict__ zeros,
|
| 644 |
-
int batch,
|
| 645 |
-
int heads,
|
| 646 |
-
int vec_row,
|
| 647 |
-
int vec_height,
|
| 648 |
-
int height,
|
| 649 |
-
int width,
|
| 650 |
-
int zero_width
|
| 651 |
-
) {
|
| 652 |
-
int weight_total = batch * heads * height * width;
|
| 653 |
-
int input_total = batch * heads * vec_row * vec_height;
|
| 654 |
-
int out_total = batch * heads * vec_row * width;
|
| 655 |
-
int tid = threadIdx.x;
|
| 656 |
-
// h is index of height with step being BLOCKHEIGHT4
|
| 657 |
-
int h = BLOCKHEIGHT4 * blockIdx.x;
|
| 658 |
-
// w is index of width with step being 1
|
| 659 |
-
int w = BLOCKWIDTH * blockIdx.y + tid;
|
| 660 |
-
if (w >= width && tid >= vec_height) {
|
| 661 |
-
return;
|
| 662 |
-
}
|
| 663 |
-
|
| 664 |
-
__shared__ scalar_t blockvec[BLOCKWIDTH];
|
| 665 |
-
// i is index of mat of block first row
|
| 666 |
-
int i = width * h + w;
|
| 667 |
-
int k;
|
| 668 |
-
scalar_t w_tmp;
|
| 669 |
-
|
| 670 |
-
int z_w = w / 8;
|
| 671 |
-
int z_mod = (w % 8) * 4;
|
| 672 |
-
|
| 673 |
-
float weight[BLOCKWIDTH];
|
| 674 |
-
|
| 675 |
-
for (int b = 0; b < batch; ++b){
|
| 676 |
-
for (int head = 0; head < heads; ++head){
|
| 677 |
-
int batch_shift = b * heads + head;
|
| 678 |
-
for (k = 0; k < BLOCKWIDTH && h * 8 + k < vec_height; ++k){
|
| 679 |
-
int k_w = (k / 8);
|
| 680 |
-
int k_bit = (k % 8) * 4;
|
| 681 |
-
|
| 682 |
-
int w_index = batch_shift * height * width + i + (k_w * width);
|
| 683 |
-
if (w_index >= weight_total || w >= width) {
|
| 684 |
-
weight[k] = 0;
|
| 685 |
-
} else {
|
| 686 |
-
scalar_t scale = scales[batch_shift * width + w];
|
| 687 |
-
scalar_t zero;
|
| 688 |
-
if (zero_width == width) {
|
| 689 |
-
zero = zeros[batch_shift * width + w];
|
| 690 |
-
} else {
|
| 691 |
-
zero = scalar_t(((as_unsigned(zeros[batch_shift * zero_width + z_w]) >> z_mod) & 0xF));
|
| 692 |
-
}
|
| 693 |
-
w_tmp = ((as_unsigned(mat[w_index]) >> k_bit) & 0xF);
|
| 694 |
-
weight[k] = scale * (w_tmp - zero);
|
| 695 |
-
}
|
| 696 |
-
}
|
| 697 |
-
|
| 698 |
-
scalar_t res;
|
| 699 |
-
for (int vr = 0; vr < vec_row; ++vr){
|
| 700 |
-
res = 0;
|
| 701 |
-
int vec_index = (batch_shift * vec_row + vr) * vec_height + blockIdx.x * BLOCKWIDTH + tid;
|
| 702 |
-
if (vec_index < input_total) {
|
| 703 |
-
blockvec[tid] = vec[vec_index];
|
| 704 |
-
} else {
|
| 705 |
-
blockvec[tid] = 0;
|
| 706 |
-
}
|
| 707 |
-
|
| 708 |
-
__syncthreads();
|
| 709 |
-
for (k = 0; k < BLOCKWIDTH && h * 8 + k < vec_height; ++k){
|
| 710 |
-
// res is the dot product of BLOCKWIDTH elements (part of width)
|
| 711 |
-
res += weight[k] * blockvec[k];
|
| 712 |
-
}
|
| 713 |
-
// add res to the final result, final matrix shape: (batch, vec_row, width)
|
| 714 |
-
int out_index = (batch_shift * vec_row + vr) * width + w;
|
| 715 |
-
if (out_index < out_total) {
|
| 716 |
-
atomicAdd(&mul[out_index], res);
|
| 717 |
-
}
|
| 718 |
-
__syncthreads();
|
| 719 |
-
}
|
| 720 |
-
}
|
| 721 |
-
}
|
| 722 |
-
}
|
| 723 |
-
|
| 724 |
-
|
| 725 |
-
|
| 726 |
-
void vecquant4matmul_batched_column_compression_cuda(
|
| 727 |
-
torch::Tensor vec,
|
| 728 |
-
torch::Tensor mat,
|
| 729 |
-
torch::Tensor mul,
|
| 730 |
-
torch::Tensor scales,
|
| 731 |
-
torch::Tensor zeros
|
| 732 |
-
) {
|
| 733 |
-
int batch = vec.size(0);
|
| 734 |
-
int heads = vec.size(1);
|
| 735 |
-
int vec_row = vec.size(2);
|
| 736 |
-
int height = vec.size(3);
|
| 737 |
-
int width = mat.size(3) * 8;
|
| 738 |
-
|
| 739 |
-
dim3 blocks(
|
| 740 |
-
(height + BLOCKWIDTH - 1) / BLOCKWIDTH,
|
| 741 |
-
(width + BLOCKWIDTH - 1) / BLOCKWIDTH
|
| 742 |
-
);
|
| 743 |
-
dim3 threads(BLOCKWIDTH);
|
| 744 |
-
|
| 745 |
-
AT_DISPATCH_FLOATING_TYPES(
|
| 746 |
-
vec.type(), "vecquant4matmul_batched_cuda", ([&] {
|
| 747 |
-
VecQuant4BatchMatMulColumnCompressionKernel<<<blocks, threads>>>(
|
| 748 |
-
vec.data<scalar_t>(), mat.data<int>(), mul.data<scalar_t>(),
|
| 749 |
-
scales.data<scalar_t>(), zeros.data<int>(),
|
| 750 |
-
batch, heads, vec_row, height, width
|
| 751 |
-
);
|
| 752 |
-
})
|
| 753 |
-
);
|
| 754 |
-
|
| 755 |
-
}
|
| 756 |
-
|
| 757 |
-
template <typename scalar_t>
|
| 758 |
-
__global__ void VecQuant4BatchMatMulColumnCompressionKernel(
|
| 759 |
-
const scalar_t* __restrict__ vec,
|
| 760 |
-
const int* __restrict__ mat,
|
| 761 |
-
scalar_t* __restrict__ mul,
|
| 762 |
-
const scalar_t* __restrict__ scales,
|
| 763 |
-
const int* __restrict__ zeros,
|
| 764 |
-
int batch,
|
| 765 |
-
int heads,
|
| 766 |
-
int vec_row,
|
| 767 |
-
int height,
|
| 768 |
-
int width
|
| 769 |
-
) {
|
| 770 |
-
int weight_total = batch * heads * height * width / 8;
|
| 771 |
-
int input_total = batch * heads * vec_row * height;
|
| 772 |
-
int out_total = batch * heads * vec_row * width;
|
| 773 |
-
int tid = threadIdx.x;
|
| 774 |
-
// h is index of height with step being BLOCKWIDTH
|
| 775 |
-
int h = BLOCKWIDTH * blockIdx.x;
|
| 776 |
-
// w is index of width with step being 1
|
| 777 |
-
int w = BLOCKWIDTH * blockIdx.y + tid;
|
| 778 |
-
if (w >= width && tid >= height) {
|
| 779 |
-
return;
|
| 780 |
-
}
|
| 781 |
-
|
| 782 |
-
__shared__ scalar_t blockvec[BLOCKWIDTH];
|
| 783 |
-
int k;
|
| 784 |
-
scalar_t w_tmp;
|
| 785 |
-
|
| 786 |
-
float weight[BLOCKWIDTH];
|
| 787 |
-
|
| 788 |
-
for (int b = 0; b < batch; ++b){
|
| 789 |
-
for (int head = 0; head < heads; ++head){
|
| 790 |
-
int batch_shift = b * heads + head;
|
| 791 |
-
for (k = 0; k < BLOCKWIDTH && h + k < height; ++k){
|
| 792 |
-
int i_w = (w / 8);
|
| 793 |
-
int w_bit = (w % 8) * 4;
|
| 794 |
-
|
| 795 |
-
int w_index = (batch_shift * height + h + k) * width / 8 + i_w;
|
| 796 |
-
if (w_index >= weight_total || w >= width) {
|
| 797 |
-
weight[k] = 0;
|
| 798 |
-
} else {
|
| 799 |
-
scalar_t scale = scales[batch_shift * height + h + k];
|
| 800 |
-
scalar_t zero = zeros[batch_shift * height + h + k];
|
| 801 |
-
w_tmp = ((as_unsigned(mat[w_index]) >> w_bit) & 0xF);
|
| 802 |
-
weight[k] = scale * (w_tmp - zero);
|
| 803 |
-
}
|
| 804 |
-
}
|
| 805 |
-
|
| 806 |
-
scalar_t res;
|
| 807 |
-
for (int vr = 0; vr < vec_row; ++vr){
|
| 808 |
-
res = 0;
|
| 809 |
-
int vec_index = (batch_shift * vec_row + vr) * height + blockIdx.x * BLOCKWIDTH + tid;
|
| 810 |
-
if (vec_index < input_total) {
|
| 811 |
-
blockvec[tid] = vec[vec_index];
|
| 812 |
-
} else {
|
| 813 |
-
blockvec[tid] = 0;
|
| 814 |
-
}
|
| 815 |
-
|
| 816 |
-
__syncthreads();
|
| 817 |
-
for (k = 0; k < BLOCKWIDTH && h + k < height; ++k){
|
| 818 |
-
// res is the dot product of BLOCKWIDTH elements (part of width)
|
| 819 |
-
res += weight[k] * blockvec[k];
|
| 820 |
-
}
|
| 821 |
-
// add res to the final result, final matrix shape: (batch, vec_row, width)
|
| 822 |
-
int out_index = (batch_shift * vec_row + vr) * width + w;
|
| 823 |
-
if (out_index < out_total) {
|
| 824 |
-
atomicAdd(&mul[out_index], res);
|
| 825 |
-
}
|
| 826 |
-
__syncthreads();
|
| 827 |
-
}
|
| 828 |
-
}
|
| 829 |
-
}
|
| 830 |
-
}
|
| 831 |
-
|
| 832 |
-
|
| 833 |
-
void vecquant8matmul_batched_old_cuda(
|
| 834 |
-
torch::Tensor vec,
|
| 835 |
-
torch::Tensor mat,
|
| 836 |
-
torch::Tensor mul,
|
| 837 |
-
torch::Tensor scales,
|
| 838 |
-
torch::Tensor zeros
|
| 839 |
-
) {
|
| 840 |
-
int batch = vec.size(0);
|
| 841 |
-
int heads = vec.size(1);
|
| 842 |
-
int vec_row = vec.size(2);
|
| 843 |
-
int vec_height = vec.size(3);
|
| 844 |
-
int height = mat.size(2);
|
| 845 |
-
int width = mat.size(3);
|
| 846 |
-
int zero_width = zeros.size(2);
|
| 847 |
-
|
| 848 |
-
dim3 blocks(
|
| 849 |
-
(height + BLOCKWIDTH - 1) / BLOCKWIDTH,
|
| 850 |
-
(width + BLOCKWIDTH - 1) / BLOCKWIDTH
|
| 851 |
-
);
|
| 852 |
-
dim3 threads(BLOCKWIDTH);
|
| 853 |
-
|
| 854 |
-
AT_DISPATCH_FLOATING_TYPES(
|
| 855 |
-
vec.type(), "vecquant8matmul_batched_old_cuda", ([&] {
|
| 856 |
-
VecQuant8BatchMatMulKernel_old<<<blocks, threads>>>(
|
| 857 |
-
vec.data<scalar_t>(), mat.data<uint8_t>(), mul.data<scalar_t>(),
|
| 858 |
-
scales.data<scalar_t>(), zeros.data<scalar_t>(),
|
| 859 |
-
batch, heads, vec_row, vec_height, height, width, zero_width
|
| 860 |
-
);
|
| 861 |
-
})
|
| 862 |
-
);
|
| 863 |
-
}
|
| 864 |
-
|
| 865 |
-
|
| 866 |
-
template <typename scalar_t>
|
| 867 |
-
__global__ void VecQuant8BatchMatMulKernel_old(
|
| 868 |
-
const scalar_t* __restrict__ vec,
|
| 869 |
-
const uint8_t* __restrict__ mat,
|
| 870 |
-
scalar_t* __restrict__ mul,
|
| 871 |
-
const scalar_t* __restrict__ scales,
|
| 872 |
-
const scalar_t* __restrict__ zeros,
|
| 873 |
-
int batch,
|
| 874 |
-
int heads,
|
| 875 |
-
int vec_row,
|
| 876 |
-
int vec_height,
|
| 877 |
-
int height,
|
| 878 |
-
int width,
|
| 879 |
-
int zero_width
|
| 880 |
-
) {
|
| 881 |
-
int weight_total = batch * heads * height * width;
|
| 882 |
-
int input_total = batch * heads * vec_row * vec_height;
|
| 883 |
-
int out_total = batch * heads * vec_row * width;
|
| 884 |
-
int tid = threadIdx.x;
|
| 885 |
-
// h is index of height with step being BLOCKHEIGHT8
|
| 886 |
-
int h = BLOCKWIDTH * blockIdx.x;
|
| 887 |
-
// w is index of width with step being 1
|
| 888 |
-
int w = BLOCKWIDTH * blockIdx.y + tid;
|
| 889 |
-
if (w >= width && tid >= vec_height) {
|
| 890 |
-
return;
|
| 891 |
-
}
|
| 892 |
-
|
| 893 |
-
__shared__ scalar_t blockvec[BLOCKWIDTH];
|
| 894 |
-
// i is index of mat of block first row
|
| 895 |
-
int i = width * h + w;
|
| 896 |
-
int k;
|
| 897 |
-
scalar_t w_tmp;
|
| 898 |
-
|
| 899 |
-
float weight[BLOCKWIDTH];
|
| 900 |
-
for (int b = 0; b < batch; ++b){
|
| 901 |
-
for (int head = 0; head < heads; ++head){
|
| 902 |
-
int batch_shift = b * heads + head;
|
| 903 |
-
for (k = 0; k < BLOCKWIDTH && h + k < vec_height; ++k){
|
| 904 |
-
int k_w = k;
|
| 905 |
-
int w_index = batch_shift * height * width + i + (k_w * width);
|
| 906 |
-
if (w_index >= weight_total || w >= width) {
|
| 907 |
-
weight[k] = 0;
|
| 908 |
-
} else {
|
| 909 |
-
scalar_t scale = scales[batch_shift * width + w];
|
| 910 |
-
scalar_t zero = zeros[batch_shift * width + w];
|
| 911 |
-
w_tmp = as_unsigned(mat[w_index]);
|
| 912 |
-
weight[k] = scale * (w_tmp - zero);
|
| 913 |
-
}
|
| 914 |
-
}
|
| 915 |
-
|
| 916 |
-
scalar_t res;
|
| 917 |
-
for (int vr = 0; vr < vec_row; ++vr){
|
| 918 |
-
res = 0;
|
| 919 |
-
int vec_index = (batch_shift * vec_row + vr) * vec_height + blockIdx.x * BLOCKWIDTH + tid;
|
| 920 |
-
if (vec_index < input_total) {
|
| 921 |
-
blockvec[tid] = vec[vec_index];
|
| 922 |
-
} else {
|
| 923 |
-
blockvec[tid] = 0;
|
| 924 |
-
}
|
| 925 |
-
|
| 926 |
-
__syncthreads();
|
| 927 |
-
for (k = 0; k < BLOCKWIDTH && h + k < vec_height; ++k){
|
| 928 |
-
// res is the dot product of BLOCKWIDTH elements (part of width)
|
| 929 |
-
res += weight[k] * blockvec[k];
|
| 930 |
-
}
|
| 931 |
-
// add res to the final result, final matrix shape: (batch, vec_row, width)
|
| 932 |
-
int out_index = (batch_shift * vec_row + vr) * width + w;
|
| 933 |
-
if (out_index < out_total) {
|
| 934 |
-
atomicAdd(&mul[out_index], res);
|
| 935 |
-
}
|
| 936 |
-
__syncthreads();
|
| 937 |
-
}
|
| 938 |
-
}
|
| 939 |
-
}
|
| 940 |
-
}
|
| 941 |
-
|
| 942 |
-
|
| 943 |
-
|
| 944 |
-
void vecquant8matmul_batched_faster_cuda(
|
| 945 |
-
torch::Tensor vec,
|
| 946 |
-
torch::Tensor mat,
|
| 947 |
-
torch::Tensor mul,
|
| 948 |
-
torch::Tensor scales,
|
| 949 |
-
torch::Tensor zeros
|
| 950 |
-
) {
|
| 951 |
-
int batch = vec.size(0);
|
| 952 |
-
int heads = vec.size(1);
|
| 953 |
-
int vec_row = vec.size(2);
|
| 954 |
-
int vec_height = vec.size(3);
|
| 955 |
-
int height = mat.size(2);
|
| 956 |
-
int width = mat.size(3);
|
| 957 |
-
int zero_width = zeros.size(2);
|
| 958 |
-
|
| 959 |
-
dim3 blocks(
|
| 960 |
-
(height + BLOCKWIDTH - 1) / BLOCKWIDTH,
|
| 961 |
-
(width + BLOCKWIDTH - 1) / BLOCKWIDTH
|
| 962 |
-
);
|
| 963 |
-
dim3 threads(BLOCKWIDTH);
|
| 964 |
-
|
| 965 |
-
VecQuant8BatchMatMulKernel_faster<<<blocks, threads>>>(
|
| 966 |
-
(half*) vec.data_ptr(),
|
| 967 |
-
(uint8_t*) mat.data_ptr(),
|
| 968 |
-
(half*) mul.data_ptr(),
|
| 969 |
-
(half*) scales.data_ptr(),
|
| 970 |
-
(half*) zeros.data_ptr(),
|
| 971 |
-
batch, heads, vec_row, vec_height, height, width, zero_width
|
| 972 |
-
);
|
| 973 |
-
}
|
| 974 |
-
|
| 975 |
-
|
| 976 |
-
|
| 977 |
-
__global__ void VecQuant8BatchMatMulKernel_faster(
|
| 978 |
-
const half* __restrict__ vec,
|
| 979 |
-
const uint8_t* __restrict__ mat,
|
| 980 |
-
half* __restrict__ mul,
|
| 981 |
-
const half* __restrict__ scales,
|
| 982 |
-
const half* __restrict__ zeros,
|
| 983 |
-
int batch,
|
| 984 |
-
int heads,
|
| 985 |
-
int vec_row,
|
| 986 |
-
int vec_height,
|
| 987 |
-
int height,
|
| 988 |
-
int width,
|
| 989 |
-
int zero_width
|
| 990 |
-
) {
|
| 991 |
-
//int weight_total = batch * heads * height * width;
|
| 992 |
-
int input_total = batch * heads * vec_row * vec_height;
|
| 993 |
-
int out_total = batch * heads * vec_row * width;
|
| 994 |
-
int tid = threadIdx.x;
|
| 995 |
-
int h = BLOCKWIDTH * blockIdx.x;
|
| 996 |
-
int w = BLOCKWIDTH * blockIdx.y + tid;
|
| 997 |
-
if (w >= width && tid >= height) {
|
| 998 |
-
return;
|
| 999 |
-
}
|
| 1000 |
-
|
| 1001 |
-
__shared__ float blockvec[BLOCKWIDTH];
|
| 1002 |
-
int i = width * h + w;
|
| 1003 |
-
int k;
|
| 1004 |
-
float w_tmp;
|
| 1005 |
-
|
| 1006 |
-
float weight[BLOCKWIDTH];
|
| 1007 |
-
for (int b = 0; b < batch; ++b){
|
| 1008 |
-
for (int head = 0; head < heads; ++head){
|
| 1009 |
-
int batch_shift = b * heads + head;
|
| 1010 |
-
for (k = 0; k < BLOCKWIDTH && h + k < vec_height; ++k){
|
| 1011 |
-
int k_w = k;
|
| 1012 |
-
int w_index = batch_shift * height * width + i + (k_w * width);
|
| 1013 |
-
float scale = __half2float(scales[batch_shift * width + w]);
|
| 1014 |
-
float zero = __half2float(zeros[batch_shift * width + w]);
|
| 1015 |
-
w_tmp = as_unsigned(mat[w_index]);
|
| 1016 |
-
weight[k] = scale *(w_tmp-zero);
|
| 1017 |
-
}
|
| 1018 |
-
|
| 1019 |
-
float res;
|
| 1020 |
-
for (int vr = 0; vr < vec_row; ++vr){
|
| 1021 |
-
res = 0;
|
| 1022 |
-
int vec_index = (batch_shift * vec_row + vr) * vec_height + blockIdx.x * BLOCKWIDTH + tid;
|
| 1023 |
-
if (vec_index < input_total) {
|
| 1024 |
-
blockvec[tid] = __half2float(vec[vec_index]);
|
| 1025 |
-
} else {
|
| 1026 |
-
blockvec[tid] = 0;
|
| 1027 |
-
}
|
| 1028 |
-
__syncthreads();
|
| 1029 |
-
for (k = 0; k < BLOCKWIDTH && h + k < vec_height; ++k){
|
| 1030 |
-
float temp_res = weight[k]*blockvec[k];
|
| 1031 |
-
res += temp_res;
|
| 1032 |
-
}
|
| 1033 |
-
int out_index = (batch_shift * vec_row + vr) * width + w;
|
| 1034 |
-
if (out_index < out_total) {
|
| 1035 |
-
atomicAdd(&mul[out_index], __float2half(res));
|
| 1036 |
-
}
|
| 1037 |
-
__syncthreads();
|
| 1038 |
-
}
|
| 1039 |
-
}
|
| 1040 |
-
}
|
| 1041 |
-
}
|
| 1042 |
-
|
| 1043 |
-
|
| 1044 |
-
|
| 1045 |
-
|
| 1046 |
-
void vecquant8matmul_batched_column_compression_faster_cuda(
|
| 1047 |
-
torch::Tensor vec,
|
| 1048 |
-
torch::Tensor mat,
|
| 1049 |
-
torch::Tensor mul,
|
| 1050 |
-
torch::Tensor scales,
|
| 1051 |
-
torch::Tensor zeros
|
| 1052 |
-
) {
|
| 1053 |
-
int batch = vec.size(0);
|
| 1054 |
-
int heads = vec.size(1);
|
| 1055 |
-
int vec_row = vec.size(2);
|
| 1056 |
-
int height = vec.size(3);
|
| 1057 |
-
int width = mat.size(3);
|
| 1058 |
-
|
| 1059 |
-
dim3 blocks(
|
| 1060 |
-
(height + BLOCKWIDTH - 1) / BLOCKWIDTH,
|
| 1061 |
-
(width + BLOCKWIDTH - 1) / BLOCKWIDTH
|
| 1062 |
-
);
|
| 1063 |
-
dim3 threads(BLOCKWIDTH);
|
| 1064 |
-
|
| 1065 |
-
VecQuant8BatchMatMulColumnCompressionKernel_faster<<<blocks, threads>>>(
|
| 1066 |
-
(half*) vec.data_ptr(),
|
| 1067 |
-
(uint8_t*) mat.data_ptr(),
|
| 1068 |
-
(half*) mul.data_ptr(),
|
| 1069 |
-
(half*) scales.data_ptr(),
|
| 1070 |
-
(half*) zeros.data_ptr(),
|
| 1071 |
-
batch, heads, vec_row, height, width
|
| 1072 |
-
);
|
| 1073 |
-
|
| 1074 |
-
}
|
| 1075 |
-
|
| 1076 |
-
__global__ void VecQuant8BatchMatMulColumnCompressionKernel_faster(
|
| 1077 |
-
const half* __restrict__ vec,
|
| 1078 |
-
const uint8_t* __restrict__ mat,
|
| 1079 |
-
half* __restrict__ mul,
|
| 1080 |
-
const half* __restrict__ scales,
|
| 1081 |
-
const half* __restrict__ zeros,
|
| 1082 |
-
int batch,
|
| 1083 |
-
int heads,
|
| 1084 |
-
int vec_row,
|
| 1085 |
-
int height,
|
| 1086 |
-
int width
|
| 1087 |
-
) {
|
| 1088 |
-
//int weight_total = batch * heads * height * width;
|
| 1089 |
-
int input_total = batch * heads * vec_row * height;
|
| 1090 |
-
int out_total = batch * heads * vec_row * width;
|
| 1091 |
-
int tid = threadIdx.x;
|
| 1092 |
-
int h = BLOCKWIDTH * blockIdx.x;
|
| 1093 |
-
int w = BLOCKWIDTH * blockIdx.y + tid;
|
| 1094 |
-
if (w >= width && tid >= height) {
|
| 1095 |
-
return;
|
| 1096 |
-
}
|
| 1097 |
-
|
| 1098 |
-
__shared__ float blockvec[BLOCKWIDTH];
|
| 1099 |
-
int k;
|
| 1100 |
-
float w_tmp;
|
| 1101 |
-
float weight[BLOCKWIDTH];
|
| 1102 |
-
|
| 1103 |
-
for (int b = 0; b < batch; ++b){
|
| 1104 |
-
for (int head = 0; head < heads; ++head){
|
| 1105 |
-
int batch_shift = b * heads + head;
|
| 1106 |
-
for (k = 0; k < BLOCKWIDTH; ++k){
|
| 1107 |
-
int w_index = (batch_shift * height + h + k) * width + w;
|
| 1108 |
-
float scale = __half2float(scales[batch_shift * height + h + k]);
|
| 1109 |
-
float zero = __half2float(zeros[batch_shift * height + h + k]);
|
| 1110 |
-
w_tmp = mat[w_index];
|
| 1111 |
-
weight[k] = scale * (w_tmp-zero);
|
| 1112 |
-
}
|
| 1113 |
-
|
| 1114 |
-
float res;
|
| 1115 |
-
for (int vr = 0; vr < vec_row; ++vr){
|
| 1116 |
-
res = 0;
|
| 1117 |
-
int vec_index = (batch_shift * vec_row + vr) * height + blockIdx.x * BLOCKWIDTH + tid;
|
| 1118 |
-
if (vec_index < input_total) {
|
| 1119 |
-
blockvec[tid] = __half2float(vec[vec_index]);
|
| 1120 |
-
} else {
|
| 1121 |
-
blockvec[tid] = 0;
|
| 1122 |
-
}
|
| 1123 |
-
__syncthreads();
|
| 1124 |
-
for (k = 0; k < BLOCKWIDTH; ++k){
|
| 1125 |
-
res += weight[k]*blockvec[k];
|
| 1126 |
-
}
|
| 1127 |
-
int out_index = (batch_shift * vec_row + vr) * width + w;
|
| 1128 |
-
if (out_index < out_total) {
|
| 1129 |
-
atomicAdd(&mul[out_index], __float2half(res));
|
| 1130 |
-
}
|
| 1131 |
-
__syncthreads();
|
| 1132 |
-
}
|
| 1133 |
-
}
|
| 1134 |
-
}
|
| 1135 |
-
}
|
| 1136 |
-
|
| 1137 |
-
|
| 1138 |
-
|
| 1139 |
-
void vecquant8matmul_batched_column_compression_old_cuda(
|
| 1140 |
-
torch::Tensor vec,
|
| 1141 |
-
torch::Tensor mat,
|
| 1142 |
-
torch::Tensor mul,
|
| 1143 |
-
torch::Tensor scales,
|
| 1144 |
-
torch::Tensor zeros
|
| 1145 |
-
) {
|
| 1146 |
-
int batch = vec.size(0);
|
| 1147 |
-
int heads = vec.size(1);
|
| 1148 |
-
int vec_row = vec.size(2);
|
| 1149 |
-
int height = vec.size(3);
|
| 1150 |
-
int width = mat.size(3);
|
| 1151 |
-
|
| 1152 |
-
dim3 blocks(
|
| 1153 |
-
(height + BLOCKWIDTH - 1) / BLOCKWIDTH,
|
| 1154 |
-
(width + BLOCKWIDTH - 1) / BLOCKWIDTH
|
| 1155 |
-
);
|
| 1156 |
-
dim3 threads(BLOCKWIDTH);
|
| 1157 |
-
|
| 1158 |
-
AT_DISPATCH_FLOATING_TYPES(
|
| 1159 |
-
vec.type(), "vecquant8matmul_batched_column_compression_old_cuda", ([&] {
|
| 1160 |
-
VecQuant8BatchMatMulColumnCompressionKernel_old<<<blocks, threads>>>(
|
| 1161 |
-
vec.data<scalar_t>(), mat.data<uint8_t>(), mul.data<scalar_t>(),
|
| 1162 |
-
scales.data<scalar_t>(), zeros.data<scalar_t>(),
|
| 1163 |
-
batch, heads, vec_row, height, width
|
| 1164 |
-
);
|
| 1165 |
-
})
|
| 1166 |
-
);
|
| 1167 |
-
|
| 1168 |
-
}
|
| 1169 |
-
|
| 1170 |
-
template <typename scalar_t>
|
| 1171 |
-
__global__ void VecQuant8BatchMatMulColumnCompressionKernel_old(
|
| 1172 |
-
const scalar_t* __restrict__ vec,
|
| 1173 |
-
const uint8_t* __restrict__ mat,
|
| 1174 |
-
scalar_t* __restrict__ mul,
|
| 1175 |
-
const scalar_t* __restrict__ scales,
|
| 1176 |
-
const scalar_t* __restrict__ zeros,
|
| 1177 |
-
int batch,
|
| 1178 |
-
int heads,
|
| 1179 |
-
int vec_row,
|
| 1180 |
-
int height,
|
| 1181 |
-
int width
|
| 1182 |
-
) {
|
| 1183 |
-
int weight_total = batch * heads * height * width;
|
| 1184 |
-
int input_total = batch * heads * vec_row * height;
|
| 1185 |
-
int out_total = batch * heads * vec_row * width;
|
| 1186 |
-
int tid = threadIdx.x;
|
| 1187 |
-
// h is index of height with step being BLOCKWIDTH
|
| 1188 |
-
int h = BLOCKWIDTH * blockIdx.x;
|
| 1189 |
-
// w is index of width with step being 1
|
| 1190 |
-
int w = BLOCKWIDTH * blockIdx.y + tid;
|
| 1191 |
-
if (w >= width && tid >= height) {
|
| 1192 |
-
return;
|
| 1193 |
-
}
|
| 1194 |
-
|
| 1195 |
-
__shared__ scalar_t blockvec[BLOCKWIDTH];
|
| 1196 |
-
int k;
|
| 1197 |
-
scalar_t w_tmp;
|
| 1198 |
-
|
| 1199 |
-
float weight[BLOCKWIDTH];
|
| 1200 |
-
|
| 1201 |
-
for (int b = 0; b < batch; ++b){
|
| 1202 |
-
for (int head = 0; head < heads; ++head){
|
| 1203 |
-
int batch_shift = b * heads + head;
|
| 1204 |
-
for (k = 0; k < BLOCKWIDTH && h + k < height; ++k){
|
| 1205 |
-
int w_index = (batch_shift * height + h + k) * width + w;
|
| 1206 |
-
if (w_index >= weight_total || w >= width) {
|
| 1207 |
-
weight[k] = 0;
|
| 1208 |
-
} else {
|
| 1209 |
-
scalar_t scale = scales[batch_shift * height + h + k];
|
| 1210 |
-
scalar_t zero = zeros[batch_shift * height + h + k];
|
| 1211 |
-
w_tmp = mat[w_index];
|
| 1212 |
-
weight[k] = scale * (w_tmp - zero);
|
| 1213 |
-
}
|
| 1214 |
-
}
|
| 1215 |
-
|
| 1216 |
-
scalar_t res;
|
| 1217 |
-
for (int vr = 0; vr < vec_row; ++vr){
|
| 1218 |
-
res = 0;
|
| 1219 |
-
int vec_index = (batch_shift * vec_row + vr) * height + blockIdx.x * BLOCKWIDTH + tid;
|
| 1220 |
-
if (vec_index < input_total) {
|
| 1221 |
-
blockvec[tid] = vec[vec_index];
|
| 1222 |
-
} else {
|
| 1223 |
-
blockvec[tid] = 0;
|
| 1224 |
-
}
|
| 1225 |
-
|
| 1226 |
-
__syncthreads();
|
| 1227 |
-
for (k = 0; k < BLOCKWIDTH && h + k < height; ++k){
|
| 1228 |
-
// res is the dot product of BLOCKWIDTH elements (part of width)
|
| 1229 |
-
res += weight[k] * blockvec[k];
|
| 1230 |
-
}
|
| 1231 |
-
// add res to the final result, final matrix shape: (batch, vec_row, width)
|
| 1232 |
-
int out_index = (batch_shift * vec_row + vr) * width + w;
|
| 1233 |
-
if (out_index < out_total) {
|
| 1234 |
-
atomicAdd(&mul[out_index], res);
|
| 1235 |
-
}
|
| 1236 |
-
__syncthreads();
|
| 1237 |
-
}
|
| 1238 |
-
}
|
| 1239 |
-
}
|
| 1240 |
-
}
|
| 1241 |
-
|
| 1242 |
-
|
| 1243 |
-
void vecquant4matmul_batched_old_cuda(
|
| 1244 |
-
torch::Tensor vec,
|
| 1245 |
-
torch::Tensor mat,
|
| 1246 |
-
torch::Tensor mul,
|
| 1247 |
-
torch::Tensor scales,
|
| 1248 |
-
torch::Tensor zeros
|
| 1249 |
-
) {
|
| 1250 |
-
int batch = vec.size(0);
|
| 1251 |
-
int heads = vec.size(1);
|
| 1252 |
-
int vec_row = vec.size(2);
|
| 1253 |
-
int vec_height = vec.size(3);
|
| 1254 |
-
int height = mat.size(2);
|
| 1255 |
-
int width = mat.size(3);
|
| 1256 |
-
int zero_width = zeros.size(2);
|
| 1257 |
-
|
| 1258 |
-
dim3 blocks(
|
| 1259 |
-
(height + BLOCKHEIGHT_OLD4 - 1) / BLOCKHEIGHT_OLD4,
|
| 1260 |
-
(width + BLOCKWIDTH - 1) / BLOCKWIDTH
|
| 1261 |
-
);
|
| 1262 |
-
dim3 threads(BLOCKWIDTH);
|
| 1263 |
-
|
| 1264 |
-
AT_DISPATCH_FLOATING_TYPES(
|
| 1265 |
-
vec.type(), "vecquant4matmul_batched_old_cuda", ([&] {
|
| 1266 |
-
VecQuant4BatchMatMulKernel_old<<<blocks, threads>>>(
|
| 1267 |
-
vec.data<scalar_t>(), mat.data<uint8_t>(), mul.data<scalar_t>(),
|
| 1268 |
-
scales.data<scalar_t>(), zeros.data<scalar_t>(),
|
| 1269 |
-
batch, heads, vec_row, vec_height, height, width, zero_width
|
| 1270 |
-
);
|
| 1271 |
-
})
|
| 1272 |
-
);
|
| 1273 |
-
|
| 1274 |
-
}
|
| 1275 |
-
|
| 1276 |
-
template <typename scalar_t>
|
| 1277 |
-
__global__ void VecQuant4BatchMatMulKernel_old(
|
| 1278 |
-
const scalar_t* __restrict__ vec,
|
| 1279 |
-
const uint8_t* __restrict__ mat,
|
| 1280 |
-
scalar_t* __restrict__ mul,
|
| 1281 |
-
const scalar_t* __restrict__ scales,
|
| 1282 |
-
const scalar_t* __restrict__ zeros,
|
| 1283 |
-
int batch,
|
| 1284 |
-
int heads,
|
| 1285 |
-
int vec_row,
|
| 1286 |
-
int vec_height,
|
| 1287 |
-
int height,
|
| 1288 |
-
int width,
|
| 1289 |
-
int zero_width
|
| 1290 |
-
) {
|
| 1291 |
-
int weight_total = batch * heads * height * width;
|
| 1292 |
-
int input_total = batch * heads * vec_row * vec_height;
|
| 1293 |
-
int out_total = batch * heads * vec_row * width;
|
| 1294 |
-
int tid = threadIdx.x;
|
| 1295 |
-
// h is index of height with step being BLOCKHEIGHT_OLD4
|
| 1296 |
-
int h = BLOCKHEIGHT_OLD4 * blockIdx.x;
|
| 1297 |
-
// w is index of width with step being 1
|
| 1298 |
-
int w = BLOCKWIDTH * blockIdx.y + tid;
|
| 1299 |
-
if (w >= width && tid >= vec_height) {
|
| 1300 |
-
return;
|
| 1301 |
-
}
|
| 1302 |
-
|
| 1303 |
-
__shared__ scalar_t blockvec[BLOCKWIDTH];
|
| 1304 |
-
// i is index of mat of block first row
|
| 1305 |
-
int i = width * h + w;
|
| 1306 |
-
int k;
|
| 1307 |
-
scalar_t w_tmp;
|
| 1308 |
-
|
| 1309 |
-
float weight[BLOCKWIDTH];
|
| 1310 |
-
for (int b = 0; b < batch; ++b){
|
| 1311 |
-
for (int head = 0; head < heads; ++head){
|
| 1312 |
-
int batch_shift = b * heads + head;
|
| 1313 |
-
for (k = 0; k < BLOCKWIDTH && h*2 + k < vec_height; ++k){
|
| 1314 |
-
int k_w = (k / 2);
|
| 1315 |
-
int k_bit = (k % 2) * 4;
|
| 1316 |
-
int w_index = batch_shift * height * width + i + (k_w * width);
|
| 1317 |
-
if (w_index >= weight_total || w >= width) {
|
| 1318 |
-
weight[k] = 0;
|
| 1319 |
-
} else {
|
| 1320 |
-
scalar_t scale = scales[batch_shift * width + w];
|
| 1321 |
-
scalar_t zero = zeros[batch_shift * width + w];
|
| 1322 |
-
w_tmp = ((as_unsigned(mat[w_index]) >> k_bit) & 0xF);
|
| 1323 |
-
weight[k] = scale * (w_tmp - zero);
|
| 1324 |
-
}
|
| 1325 |
-
}
|
| 1326 |
-
|
| 1327 |
-
scalar_t res;
|
| 1328 |
-
for (int vr = 0; vr < vec_row; ++vr){
|
| 1329 |
-
res = 0;
|
| 1330 |
-
int vec_index = (batch_shift * vec_row + vr) * vec_height + blockIdx.x * BLOCKWIDTH + tid;
|
| 1331 |
-
if (vec_index < input_total) {
|
| 1332 |
-
blockvec[tid] = vec[vec_index];
|
| 1333 |
-
} else {
|
| 1334 |
-
blockvec[tid] = 0;
|
| 1335 |
-
}
|
| 1336 |
-
|
| 1337 |
-
__syncthreads();
|
| 1338 |
-
for (k = 0; k < BLOCKWIDTH && h*2 + k < vec_height; ++k){
|
| 1339 |
-
// res is the dot product of BLOCKWIDTH elements (part of width)
|
| 1340 |
-
res += weight[k] * blockvec[k];
|
| 1341 |
-
}
|
| 1342 |
-
// add res to the final result, final matrix shape: (batch, vec_row, width)
|
| 1343 |
-
int out_index = (batch_shift * vec_row + vr) * width + w;
|
| 1344 |
-
if (out_index < out_total) {
|
| 1345 |
-
atomicAdd(&mul[out_index], res);
|
| 1346 |
-
}
|
| 1347 |
-
__syncthreads();
|
| 1348 |
-
}
|
| 1349 |
-
}
|
| 1350 |
-
}
|
| 1351 |
-
}
|
| 1352 |
-
|
| 1353 |
-
|
| 1354 |
-
|
| 1355 |
-
|
| 1356 |
-
|
| 1357 |
-
void vecquant4matmul_batched_column_compression_old_cuda(
|
| 1358 |
-
torch::Tensor vec,
|
| 1359 |
-
torch::Tensor mat,
|
| 1360 |
-
torch::Tensor mul,
|
| 1361 |
-
torch::Tensor scales,
|
| 1362 |
-
torch::Tensor zeros
|
| 1363 |
-
) {
|
| 1364 |
-
int batch = vec.size(0);
|
| 1365 |
-
int heads = vec.size(1);
|
| 1366 |
-
int vec_row = vec.size(2);
|
| 1367 |
-
int height = vec.size(3);
|
| 1368 |
-
int width = mat.size(3);
|
| 1369 |
-
|
| 1370 |
-
dim3 blocks(
|
| 1371 |
-
(height + BLOCKHEIGHT_OLD4 - 1) / BLOCKHEIGHT_OLD4,
|
| 1372 |
-
(width + BLOCKWIDTH - 1) / BLOCKWIDTH
|
| 1373 |
-
);
|
| 1374 |
-
dim3 threads(BLOCKWIDTH);
|
| 1375 |
-
|
| 1376 |
-
AT_DISPATCH_FLOATING_TYPES(
|
| 1377 |
-
vec.type(), "vecquant4matmul_batched_column_compression_old_cuda", ([&] {
|
| 1378 |
-
VecQuant4BatchMatMulColumnCompressionKernel_old<<<blocks, threads>>>(
|
| 1379 |
-
vec.data<scalar_t>(), mat.data<uint8_t>(), mul.data<scalar_t>(),
|
| 1380 |
-
scales.data<scalar_t>(), zeros.data<scalar_t>(),
|
| 1381 |
-
batch, heads, vec_row, height, width
|
| 1382 |
-
);
|
| 1383 |
-
})
|
| 1384 |
-
);
|
| 1385 |
-
|
| 1386 |
-
}
|
| 1387 |
-
|
| 1388 |
-
template <typename scalar_t>
|
| 1389 |
-
__global__ void VecQuant4BatchMatMulColumnCompressionKernel_old(
|
| 1390 |
-
const scalar_t* __restrict__ vec,
|
| 1391 |
-
const uint8_t* __restrict__ mat,
|
| 1392 |
-
scalar_t* __restrict__ mul,
|
| 1393 |
-
const scalar_t* __restrict__ scales,
|
| 1394 |
-
const scalar_t* __restrict__ zeros,
|
| 1395 |
-
int batch,
|
| 1396 |
-
int heads,
|
| 1397 |
-
int vec_row,
|
| 1398 |
-
int height,
|
| 1399 |
-
int width
|
| 1400 |
-
) {
|
| 1401 |
-
int weight_total = batch * heads * height * width;
|
| 1402 |
-
int input_total = batch * heads * vec_row * height;
|
| 1403 |
-
int out_total = batch * heads * vec_row * width;
|
| 1404 |
-
int tid = threadIdx.x;
|
| 1405 |
-
// h is index of height with step being BLOCKWIDTH
|
| 1406 |
-
int h = BLOCKHEIGHT_OLD4 * blockIdx.x;
|
| 1407 |
-
// w is index of width with step being 1
|
| 1408 |
-
int w = BLOCKWIDTH * blockIdx.y + tid;
|
| 1409 |
-
if (w >= width && tid >= height) {
|
| 1410 |
-
return;
|
| 1411 |
-
}
|
| 1412 |
-
|
| 1413 |
-
__shared__ scalar_t blockvec[BLOCKWIDTH];
|
| 1414 |
-
int k;
|
| 1415 |
-
scalar_t w_tmp;
|
| 1416 |
-
|
| 1417 |
-
float weight[BLOCKWIDTH];
|
| 1418 |
-
|
| 1419 |
-
for (int b = 0; b < batch; ++b){
|
| 1420 |
-
for (int head = 0; head < heads; ++head){
|
| 1421 |
-
int batch_shift = b * heads + head;
|
| 1422 |
-
for (k = 0; k < BLOCKWIDTH && h*2 + k < height; ++k){
|
| 1423 |
-
int k_w = (k / 2);
|
| 1424 |
-
int k_bit = (k % 2) * 4;
|
| 1425 |
-
int w_index = (batch_shift * height + h + k) * width + k_w;
|
| 1426 |
-
if (w_index >= weight_total || w >= width) {
|
| 1427 |
-
weight[k] = 0;
|
| 1428 |
-
} else {
|
| 1429 |
-
scalar_t scale = scales[batch_shift * height + h + k];
|
| 1430 |
-
scalar_t zero = zeros[batch_shift * height + h + k];
|
| 1431 |
-
w_tmp = ((as_unsigned(mat[w_index]) >> k_bit) & 0xF);
|
| 1432 |
-
weight[k] = scale * (w_tmp - zero);
|
| 1433 |
-
}
|
| 1434 |
-
}
|
| 1435 |
-
|
| 1436 |
-
scalar_t res;
|
| 1437 |
-
for (int vr = 0; vr < vec_row; ++vr){
|
| 1438 |
-
res = 0;
|
| 1439 |
-
int vec_index = (batch_shift * vec_row + vr) * height + blockIdx.x * BLOCKWIDTH + tid;
|
| 1440 |
-
if (vec_index < input_total) {
|
| 1441 |
-
blockvec[tid] = vec[vec_index];
|
| 1442 |
-
} else {
|
| 1443 |
-
blockvec[tid] = 0;
|
| 1444 |
-
}
|
| 1445 |
-
|
| 1446 |
-
__syncthreads();
|
| 1447 |
-
for (k = 0; k < BLOCKWIDTH && h*2 + k < height; ++k){
|
| 1448 |
-
// res is the dot product of BLOCKWIDTH elements (part of width)
|
| 1449 |
-
res += weight[k] * blockvec[k];
|
| 1450 |
-
}
|
| 1451 |
-
// add res to the final result, final matrix shape: (batch, vec_row, width)
|
| 1452 |
-
int out_index = (batch_shift * vec_row + vr) * width + w;
|
| 1453 |
-
if (out_index < out_total) {
|
| 1454 |
-
atomicAdd(&mul[out_index], res);
|
| 1455 |
-
}
|
| 1456 |
-
__syncthreads();
|
| 1457 |
-
}
|
| 1458 |
-
}
|
| 1459 |
-
}
|
| 1460 |
-
}
|
| 1461 |
-
|
| 1462 |
-
|
| 1463 |
-
|
| 1464 |
-
|
| 1465 |
-
|
| 1466 |
-
void vecquant8matmul_batched_faster_old_cuda(
|
| 1467 |
-
torch::Tensor vec,
|
| 1468 |
-
torch::Tensor mat,
|
| 1469 |
-
torch::Tensor mul,
|
| 1470 |
-
torch::Tensor scales,
|
| 1471 |
-
torch::Tensor zeros
|
| 1472 |
-
) {
|
| 1473 |
-
int batch = vec.size(0);
|
| 1474 |
-
int heads = vec.size(1);
|
| 1475 |
-
int vec_row = vec.size(2);
|
| 1476 |
-
int vec_height = vec.size(3);
|
| 1477 |
-
int height = mat.size(2);
|
| 1478 |
-
int width = mat.size(3);
|
| 1479 |
-
|
| 1480 |
-
dim3 blocks(
|
| 1481 |
-
(height + BLOCKWIDTH - 1) / BLOCKWIDTH,
|
| 1482 |
-
(width + BLOCKWIDTH - 1) / BLOCKWIDTH
|
| 1483 |
-
);
|
| 1484 |
-
dim3 threads(BLOCKWIDTH);
|
| 1485 |
-
|
| 1486 |
-
VecQuant8BatchMatMulKernel_faster_old<<<blocks, threads>>>(
|
| 1487 |
-
(half*) vec.data_ptr(),
|
| 1488 |
-
(uint8_t*) mat.data_ptr(),
|
| 1489 |
-
(half*) mul.data_ptr(),
|
| 1490 |
-
(half*) scales.data_ptr(),
|
| 1491 |
-
(half*) zeros.data_ptr(),
|
| 1492 |
-
batch, heads, vec_row, vec_height, height, width
|
| 1493 |
-
);
|
| 1494 |
-
}
|
| 1495 |
-
|
| 1496 |
-
|
| 1497 |
-
__global__ void VecQuant8BatchMatMulKernel_faster_old(
|
| 1498 |
-
const half* __restrict__ vec,
|
| 1499 |
-
const uint8_t* __restrict__ mat,
|
| 1500 |
-
half* __restrict__ mul,
|
| 1501 |
-
const half* __restrict__ scales,
|
| 1502 |
-
const half* __restrict__ zeros,
|
| 1503 |
-
int batch,
|
| 1504 |
-
int heads,
|
| 1505 |
-
int vec_row,
|
| 1506 |
-
int vec_height,
|
| 1507 |
-
int height,
|
| 1508 |
-
int width
|
| 1509 |
-
) {
|
| 1510 |
-
int weight_total = batch * heads * height * width;
|
| 1511 |
-
int input_total = batch * heads * vec_row * vec_height;
|
| 1512 |
-
int out_total = batch * heads * vec_row * width;
|
| 1513 |
-
int tid = threadIdx.x;
|
| 1514 |
-
const int BLOCKWIDTH_half = BLOCKWIDTH/2;
|
| 1515 |
-
|
| 1516 |
-
int h = BLOCKWIDTH * blockIdx.x; //head_dim, dim=-1
|
| 1517 |
-
int w = BLOCKWIDTH * blockIdx.y + tid; //seq-len, +0-256 ,dim=-2
|
| 1518 |
-
/*
|
| 1519 |
-
if (w >= width && tid >= vec_height) {
|
| 1520 |
-
return;
|
| 1521 |
-
}
|
| 1522 |
-
*/
|
| 1523 |
-
__shared__ half blockvec[BLOCKWIDTH]; //256
|
| 1524 |
-
int i = width * h + w;
|
| 1525 |
-
int k;
|
| 1526 |
-
|
| 1527 |
-
half w_tmp1 = __float2half(0);
|
| 1528 |
-
half w_tmp2 = __float2half(0);
|
| 1529 |
-
|
| 1530 |
-
half2 weight[BLOCKWIDTH_half];
|
| 1531 |
-
for (int b = 0; b < batch; ++b){
|
| 1532 |
-
for (int head = 0; head < heads; ++head){
|
| 1533 |
-
int batch_shift = b * heads + head;
|
| 1534 |
-
//int zero_index = batch_shift;
|
| 1535 |
-
for (k = 0; k < BLOCKWIDTH_half; ++k){
|
| 1536 |
-
int w_index1 = batch_shift * height * width + i + (2 * k * width); // [batch,head,h+k, w]
|
| 1537 |
-
int w_index2 = batch_shift * height * width + i + ((2 * k + 1) * width);
|
| 1538 |
-
int zero_index = batch_shift * width + w; // [batch,head, w]
|
| 1539 |
-
if (w_index1 >= weight_total || w >= width || (2 * k + h) >= height) {
|
| 1540 |
-
weight[k] = __float2half2_rn(0);
|
| 1541 |
-
} else {
|
| 1542 |
-
float zero_f=__half2float(zeros[zero_index]);
|
| 1543 |
-
float scale_f= __half2float(scales[zero_index]);
|
| 1544 |
-
if (w_index2 >= weight_total){
|
| 1545 |
-
w_tmp1 = __float2half((as_unsigned(mat[w_index1]) -zero_f)*scale_f);
|
| 1546 |
-
w_tmp2 = __float2half(0);
|
| 1547 |
-
weight[k] = __halves2half2(w_tmp1,w_tmp2);
|
| 1548 |
-
//printf("zero_index is %d w is %d height is %d width is %d w_index1 is %d w_tmp1 is %f w_tmp2 is %f zero is %f scale is %f low is %f high is %f \n ",zero_index,w,height, width,w_index1,__half2float(w_tmp1),__half2float(w_tmp2),zero_f,scale_f,__low2float(weight[k]),__high2float(weight[k]));
|
| 1549 |
-
}else{
|
| 1550 |
-
w_tmp1 = __int2half_rn(as_unsigned(mat[w_index1]));
|
| 1551 |
-
w_tmp2 = __int2half_rn(as_unsigned(mat[w_index2]));
|
| 1552 |
-
|
| 1553 |
-
//weight[k] = __hmul2(__hsub2(__halves2half2(w_tmp1,w_tmp2), __halves2half2(zero,zero)),__halves2half2(scale,scale));
|
| 1554 |
-
weight[k] = __hfma2(__halves2half2(w_tmp1,w_tmp2), __float2half2_rn(scale_f), __float2half2_rn(-(scale_f * zero_f)));
|
| 1555 |
-
//printf("zero_index1 is %d zero_index2 is %d k is %d head is %d w is %d h is %d height is %d width is %d w_index1 is %d w_index2 is %d zero is %f scale is %f low is %f high is %f \n ",zero_index1,zero_index2,k,head,w,h,height, width,w_index1,w_index2,__half2float(zero1),__half2float(scale1),__low2float(weight[k]),__high2float(weight[k]));
|
| 1556 |
-
}
|
| 1557 |
-
}
|
| 1558 |
-
}
|
| 1559 |
-
|
| 1560 |
-
|
| 1561 |
-
for (int vr = 0; vr < vec_row; ++vr){
|
| 1562 |
-
float res=0;
|
| 1563 |
-
int vec_index = (batch_shift * vec_row + vr) * height + blockIdx.x * BLOCKWIDTH + tid;
|
| 1564 |
-
int out_index = (batch_shift * vec_row + vr) * width + w;
|
| 1565 |
-
if (vec_index < input_total) {
|
| 1566 |
-
//blockvec[tid] = __half2float(vec[vec_index]);// [batch, head, vr, tid(seq_len dim+)]
|
| 1567 |
-
blockvec[tid] = vec[vec_index];
|
| 1568 |
-
//printf("width is %d height is %d h is %d w is %d vec_index is %d out_index is %d vec_row is %d vec_height is %d,vr is %d tid is %d blockvec is %f\n",width,height, h,w,vec_index,out_index,vec_row,vec_height,vr,tid,blockvec[tid]);
|
| 1569 |
-
} else {
|
| 1570 |
-
blockvec[tid] = __float2half(0);
|
| 1571 |
-
}
|
| 1572 |
-
__syncthreads();
|
| 1573 |
-
if (out_index < out_total) {
|
| 1574 |
-
for (k = 0; k < BLOCKWIDTH_half; ++k){
|
| 1575 |
-
half2 res2 = __hmul2(weight[k],__halves2half2(blockvec[2*k],blockvec[2*k+1]));
|
| 1576 |
-
res += __low2float(res2) + __high2float(res2);
|
| 1577 |
-
}
|
| 1578 |
-
atomicAdd(&mul[out_index], __float2half(res));
|
| 1579 |
-
}
|
| 1580 |
-
__syncthreads();
|
| 1581 |
-
}
|
| 1582 |
-
}
|
| 1583 |
-
}
|
| 1584 |
-
}
|
| 1585 |
-
|
| 1586 |
-
|
| 1587 |
-
void vecquant8matmul_batched_column_compression_faster_old_cuda(
|
| 1588 |
-
torch::Tensor vec, // [batch,heads, seq_q, seq_v]
|
| 1589 |
-
torch::Tensor mat, // [batch,heads, seq_v, head_dim]
|
| 1590 |
-
torch::Tensor mul, // [batch,heads, seq_q,head_dim]
|
| 1591 |
-
torch::Tensor scales, // [batch,heads, head_dim]
|
| 1592 |
-
torch::Tensor zeros
|
| 1593 |
-
) {
|
| 1594 |
-
int batch = vec.size(0);
|
| 1595 |
-
int heads = vec.size(1);
|
| 1596 |
-
int vec_row = vec.size(2); //ql
|
| 1597 |
-
int height = mat.size(2); //vl
|
| 1598 |
-
int width = mat.size(3); //head_dim
|
| 1599 |
-
|
| 1600 |
-
dim3 blocks(
|
| 1601 |
-
(height + BLOCKWIDTH - 1) / BLOCKWIDTH,
|
| 1602 |
-
(width + BLOCKWIDTH - 1) / BLOCKWIDTH
|
| 1603 |
-
);
|
| 1604 |
-
dim3 threads(BLOCKWIDTH);
|
| 1605 |
-
|
| 1606 |
-
VecQuant8BatchMatMulColumnCompressionKernel_faster_old<<<blocks, threads>>>(
|
| 1607 |
-
(half*) vec.data_ptr(),
|
| 1608 |
-
(uint8_t*) mat.data_ptr(),
|
| 1609 |
-
(half*) mul.data_ptr(),
|
| 1610 |
-
(half*) scales.data_ptr(),
|
| 1611 |
-
(half*) zeros.data_ptr(),
|
| 1612 |
-
batch, heads, vec_row, height, width
|
| 1613 |
-
);
|
| 1614 |
-
|
| 1615 |
-
}
|
| 1616 |
-
|
| 1617 |
-
|
| 1618 |
-
__global__ void VecQuant8BatchMatMulColumnCompressionKernel_faster_old(
|
| 1619 |
-
const half* __restrict__ vec, // [batch,heads, seq_q, seq_v]
|
| 1620 |
-
const uint8_t* __restrict__ mat, // [batch,heads, seq_v, head_dim]
|
| 1621 |
-
half* __restrict__ mul, // [batch,heads, seq_q,head_dim]
|
| 1622 |
-
const half* __restrict__ scales, // [batch,heads, seq_v]
|
| 1623 |
-
const half* __restrict__ zeros,
|
| 1624 |
-
int batch,
|
| 1625 |
-
int heads,
|
| 1626 |
-
int vec_row, //seq_q
|
| 1627 |
-
int height, //seq_v
|
| 1628 |
-
int width //head_dim
|
| 1629 |
-
) {
|
| 1630 |
-
int weight_total = batch * heads * height * width;
|
| 1631 |
-
int input_total = batch * heads * vec_row * height;
|
| 1632 |
-
int out_total = batch * heads * vec_row * width;
|
| 1633 |
-
int tid = threadIdx.x;
|
| 1634 |
-
int h = BLOCKWIDTH * blockIdx.x; // vl
|
| 1635 |
-
int w = BLOCKWIDTH * blockIdx.y + tid; //head_dim + block
|
| 1636 |
-
if (w >= width && tid >= height) {
|
| 1637 |
-
return;
|
| 1638 |
-
}
|
| 1639 |
-
__shared__ half blockvec[BLOCKWIDTH];
|
| 1640 |
-
int k;
|
| 1641 |
-
half w_tmp1 = __float2half(0);
|
| 1642 |
-
half w_tmp2 = __float2half(0);
|
| 1643 |
-
int i = width * h + w;
|
| 1644 |
-
const int BLOCKWIDTH_half = BLOCKWIDTH/2;
|
| 1645 |
-
half2 weight[BLOCKWIDTH_half];
|
| 1646 |
-
|
| 1647 |
-
for (int b = 0; b < batch; ++b){
|
| 1648 |
-
for (int head = 0; head < heads; ++head){
|
| 1649 |
-
int batch_shift = b * heads + head;
|
| 1650 |
-
//int zero_index = batch_shift;
|
| 1651 |
-
for (k = 0; k < BLOCKWIDTH_half; ++k){
|
| 1652 |
-
int w_index1 = batch_shift * height * width + i + (2 * k) * width; // [batch,head, h+k, w]
|
| 1653 |
-
int w_index2 = batch_shift * height * width + i + ((2 * k + 1) * width);
|
| 1654 |
-
int zero_index1 = batch_shift * height + h + 2*k; // [batch,head, w]
|
| 1655 |
-
int zero_index2 = batch_shift * height + h + 2*k+1; // [batch,head, w]
|
| 1656 |
-
|
| 1657 |
-
if (w_index1 >= weight_total || (2 * k + h)>=height) {
|
| 1658 |
-
weight[k]=__float2half2_rn(0);
|
| 1659 |
-
} else{
|
| 1660 |
-
//int zero_index = batch_shift + h; // [batch,head, w]
|
| 1661 |
-
//float scale_f1 = __half2float(scales[zero_index1]);
|
| 1662 |
-
//float zero_f1 = __half2float(zeros[zero_index1]);
|
| 1663 |
-
if (w_index2>=weight_total){
|
| 1664 |
-
w_tmp1 = __float2half((as_unsigned(mat[w_index1]) - __half2float(zeros[zero_index1]))* __half2float(scales[zero_index1]));
|
| 1665 |
-
w_tmp2 = __float2half(0);
|
| 1666 |
-
weight[k] = __halves2half2(w_tmp1,w_tmp2);
|
| 1667 |
-
//printf("zero_index is %d k is %d w is %d head is %d height is %d width is %d w_index1 is %d w_tmp1 is %f w_tmp2 is %f zero is %f scale is %f low is %f high is %f \n ",zero_index,k,w,head,height, width,w_index1,__half2float(w_tmp1),__half2float(w_tmp2),zero_f,scale_f,__low2float(weight[k]),__high2float(weight[k]));
|
| 1668 |
-
}else{
|
| 1669 |
-
w_tmp1 = __int2half_rn(as_unsigned(mat[w_index1]));
|
| 1670 |
-
w_tmp2 = __int2half_rn(as_unsigned(mat[w_index2]));
|
| 1671 |
-
half zero1=zeros[zero_index1];
|
| 1672 |
-
half zero2=zeros[zero_index2];
|
| 1673 |
-
half scale1=scales[zero_index1];
|
| 1674 |
-
half scale2=scales[zero_index2];
|
| 1675 |
-
weight[k] = __hmul2(__hsub2(__halves2half2(w_tmp1,w_tmp2), __halves2half2(zero1,zero2)),__halves2half2(scale1,scale2));
|
| 1676 |
-
//weight[k] = __hfma2(__halves2half2(w_tmp1,w_tmp2), __float2half2_rn(scale_f), __float2half2_rn(-(scale_f * zero_f)));
|
| 1677 |
-
//printf("zero_index1 is %d zero_index2 is %d k is %d head is %d w is %d h is %d height is %d width is %d w_index1 is %d w_index2 is %d zero is %f scale is %f low is %f high is %f \n ",zero_index1,zero_index2,k,head,w,h,height, width,w_index1,w_index2,__half2float(zero1),__half2float(scale1),__low2float(weight[k]),__high2float(weight[k]));
|
| 1678 |
-
}
|
| 1679 |
-
}
|
| 1680 |
-
}
|
| 1681 |
-
|
| 1682 |
-
|
| 1683 |
-
for (int vr = 0; vr < vec_row; ++vr){
|
| 1684 |
-
float res=0;
|
| 1685 |
-
int vec_index = (batch_shift * vec_row + vr) * height + blockIdx.x * BLOCKWIDTH + tid;
|
| 1686 |
-
int out_index = (batch_shift * vec_row + vr) * width + w;
|
| 1687 |
-
|
| 1688 |
-
if (vec_index < input_total) {
|
| 1689 |
-
//blockvec[tid] = __half2float(vec[vec_index]);
|
| 1690 |
-
blockvec[tid] = vec[vec_index];
|
| 1691 |
-
//printf("vec_index is %d out_index is %d vec_row is %d ,vr is %d tid is %d blockvec is %f\n",vec_index,out_index,vec_row,vr,tid,blockvec[tid]);
|
| 1692 |
-
} else {
|
| 1693 |
-
blockvec[tid] = __float2half(0);
|
| 1694 |
-
//blockvec[tid] = 0;
|
| 1695 |
-
}
|
| 1696 |
-
__syncthreads();
|
| 1697 |
-
if (out_index < out_total) {
|
| 1698 |
-
for (k = 0; k < BLOCKWIDTH_half; ++k){
|
| 1699 |
-
half2 res2 = __hmul2(weight[k],__halves2half2(blockvec[2*k],blockvec[2*k+1]));
|
| 1700 |
-
res += __low2float(res2) + __high2float(res2);
|
| 1701 |
-
}
|
| 1702 |
-
atomicAdd(&mul[out_index], __float2half(res));
|
| 1703 |
-
}
|
| 1704 |
-
__syncthreads();
|
| 1705 |
-
}
|
| 1706 |
-
}
|
| 1707 |
-
}
|
| 1708 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
config.json
CHANGED
|
@@ -1,4 +1,6 @@
|
|
| 1 |
{
|
|
|
|
|
|
|
| 2 |
"architectures": [
|
| 3 |
"QWenLMHeadModel"
|
| 4 |
],
|
|
@@ -6,32 +8,39 @@
|
|
| 6 |
"AutoConfig": "configuration_qwen.QWenConfig",
|
| 7 |
"AutoModelForCausalLM": "modeling_qwen.QWenLMHeadModel"
|
| 8 |
},
|
| 9 |
-
"
|
| 10 |
"bf16": false,
|
| 11 |
-
"emb_dropout_prob": 0.0,
|
| 12 |
"fp16": false,
|
| 13 |
"fp32": false,
|
| 14 |
-
"
|
| 15 |
-
"
|
|
|
|
|
|
|
|
|
|
| 16 |
"initializer_range": 0.02,
|
| 17 |
"kv_channels": 128,
|
| 18 |
"layer_norm_epsilon": 1e-06,
|
| 19 |
-
"max_position_embeddings": 32768,
|
| 20 |
"model_type": "qwen",
|
|
|
|
|
|
|
|
|
|
|
|
|
| 21 |
"no_bias": true,
|
| 22 |
-
"num_attention_heads": 32,
|
| 23 |
-
"num_hidden_layers": 32,
|
| 24 |
"onnx_safe": null,
|
|
|
|
|
|
|
|
|
|
|
|
|
| 25 |
"rotary_emb_base": 10000,
|
| 26 |
"rotary_pct": 1.0,
|
| 27 |
"scale_attn_weights": true,
|
| 28 |
-
"seq_length":
|
| 29 |
"tie_word_embeddings": false,
|
| 30 |
-
"
|
| 31 |
-
"transformers_version": "4.
|
| 32 |
"use_cache": true,
|
| 33 |
-
"use_dynamic_ntk": true,
|
| 34 |
"use_flash_attn": "auto",
|
| 35 |
-
"
|
| 36 |
-
"
|
| 37 |
-
|
|
|
|
|
|
| 1 |
{
|
| 2 |
+
"activation": "swiglu",
|
| 3 |
+
"apply_residual_connection_post_layernorm": false,
|
| 4 |
"architectures": [
|
| 5 |
"QWenLMHeadModel"
|
| 6 |
],
|
|
|
|
| 8 |
"AutoConfig": "configuration_qwen.QWenConfig",
|
| 9 |
"AutoModelForCausalLM": "modeling_qwen.QWenLMHeadModel"
|
| 10 |
},
|
| 11 |
+
"attn_pdrop": 0.0,
|
| 12 |
"bf16": false,
|
|
|
|
| 13 |
"fp16": false,
|
| 14 |
"fp32": false,
|
| 15 |
+
"bias_dropout_fusion": true,
|
| 16 |
+
"bos_token_id": 151643,
|
| 17 |
+
"embd_pdrop": 0.0,
|
| 18 |
+
"eos_token_id": 151643,
|
| 19 |
+
"ffn_hidden_size": 22016,
|
| 20 |
"initializer_range": 0.02,
|
| 21 |
"kv_channels": 128,
|
| 22 |
"layer_norm_epsilon": 1e-06,
|
|
|
|
| 23 |
"model_type": "qwen",
|
| 24 |
+
"n_embd": 4096,
|
| 25 |
+
"n_head": 32,
|
| 26 |
+
"n_layer": 32,
|
| 27 |
+
"n_positions": 6144,
|
| 28 |
"no_bias": true,
|
|
|
|
|
|
|
| 29 |
"onnx_safe": null,
|
| 30 |
+
"padded_vocab_size": 151936,
|
| 31 |
+
"params_dtype": "torch.bfloat16",
|
| 32 |
+
"pos_emb": "rotary",
|
| 33 |
+
"resid_pdrop": 0.1,
|
| 34 |
"rotary_emb_base": 10000,
|
| 35 |
"rotary_pct": 1.0,
|
| 36 |
"scale_attn_weights": true,
|
| 37 |
+
"seq_length": 2048,
|
| 38 |
"tie_word_embeddings": false,
|
| 39 |
+
"tokenizer_type": "QWenTokenizer",
|
| 40 |
+
"transformers_version": "4.31.0",
|
| 41 |
"use_cache": true,
|
|
|
|
| 42 |
"use_flash_attn": "auto",
|
| 43 |
+
"vocab_size": 151936,
|
| 44 |
+
"use_dynamic_ntk": true,
|
| 45 |
+
"use_logn_attn": true
|
| 46 |
+
}
|
configuration_qwen.py
CHANGED
|
@@ -9,49 +9,61 @@ from transformers import PretrainedConfig
|
|
| 9 |
class QWenConfig(PretrainedConfig):
|
| 10 |
model_type = "qwen"
|
| 11 |
keys_to_ignore_at_inference = ["past_key_values"]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 12 |
|
| 13 |
def __init__(
|
| 14 |
self,
|
| 15 |
-
vocab_size=
|
| 16 |
-
|
| 17 |
-
|
| 18 |
-
|
| 19 |
-
|
| 20 |
-
|
| 21 |
-
|
|
|
|
| 22 |
initializer_range=0.02,
|
| 23 |
-
max_position_embeddings=8192,
|
| 24 |
scale_attn_weights=True,
|
| 25 |
use_cache=True,
|
|
|
|
|
|
|
| 26 |
bf16=False,
|
| 27 |
fp16=False,
|
| 28 |
fp32=False,
|
| 29 |
kv_channels=128,
|
| 30 |
rotary_pct=1.0,
|
| 31 |
rotary_emb_base=10000,
|
| 32 |
-
use_dynamic_ntk=
|
| 33 |
-
use_logn_attn=
|
| 34 |
-
use_flash_attn=
|
| 35 |
-
|
| 36 |
no_bias=True,
|
| 37 |
tie_word_embeddings=False,
|
| 38 |
-
use_cache_quantization=False,
|
| 39 |
-
use_cache_kernel=False,
|
| 40 |
-
softmax_in_fp32=False,
|
| 41 |
**kwargs,
|
| 42 |
):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 43 |
self.vocab_size = vocab_size
|
| 44 |
-
self.
|
| 45 |
-
self.
|
| 46 |
-
self.
|
| 47 |
-
self.
|
| 48 |
-
self.
|
| 49 |
-
self.
|
| 50 |
self.layer_norm_epsilon = layer_norm_epsilon
|
| 51 |
self.initializer_range = initializer_range
|
| 52 |
self.scale_attn_weights = scale_attn_weights
|
| 53 |
self.use_cache = use_cache
|
| 54 |
-
self.
|
|
|
|
|
|
|
| 55 |
self.bf16 = bf16
|
| 56 |
self.fp16 = fp16
|
| 57 |
self.fp32 = fp32
|
|
@@ -61,11 +73,6 @@ class QWenConfig(PretrainedConfig):
|
|
| 61 |
self.use_dynamic_ntk = use_dynamic_ntk
|
| 62 |
self.use_logn_attn = use_logn_attn
|
| 63 |
self.use_flash_attn = use_flash_attn
|
|
|
|
| 64 |
self.no_bias = no_bias
|
| 65 |
-
self.
|
| 66 |
-
self.use_cache_kernel = use_cache_kernel
|
| 67 |
-
self.softmax_in_fp32 = softmax_in_fp32
|
| 68 |
-
super().__init__(
|
| 69 |
-
tie_word_embeddings=tie_word_embeddings,
|
| 70 |
-
**kwargs
|
| 71 |
-
)
|
|
|
|
| 9 |
class QWenConfig(PretrainedConfig):
|
| 10 |
model_type = "qwen"
|
| 11 |
keys_to_ignore_at_inference = ["past_key_values"]
|
| 12 |
+
attribute_map = {
|
| 13 |
+
"hidden_size": "n_embd",
|
| 14 |
+
"num_attention_heads": "n_head",
|
| 15 |
+
"max_position_embeddings": "n_positions",
|
| 16 |
+
"num_hidden_layers": "n_layer",
|
| 17 |
+
}
|
| 18 |
|
| 19 |
def __init__(
|
| 20 |
self,
|
| 21 |
+
vocab_size=151851,
|
| 22 |
+
n_embd=4096,
|
| 23 |
+
n_layer=32,
|
| 24 |
+
n_head=32,
|
| 25 |
+
n_inner=None,
|
| 26 |
+
embd_pdrop=0.0,
|
| 27 |
+
attn_pdrop=0.0,
|
| 28 |
+
layer_norm_epsilon=1e-5,
|
| 29 |
initializer_range=0.02,
|
|
|
|
| 30 |
scale_attn_weights=True,
|
| 31 |
use_cache=True,
|
| 32 |
+
eos_token_id=151643,
|
| 33 |
+
apply_residual_connection_post_layernorm=False,
|
| 34 |
bf16=False,
|
| 35 |
fp16=False,
|
| 36 |
fp32=False,
|
| 37 |
kv_channels=128,
|
| 38 |
rotary_pct=1.0,
|
| 39 |
rotary_emb_base=10000,
|
| 40 |
+
use_dynamic_ntk=False,
|
| 41 |
+
use_logn_attn=False,
|
| 42 |
+
use_flash_attn=True,
|
| 43 |
+
ffn_hidden_size=22016,
|
| 44 |
no_bias=True,
|
| 45 |
tie_word_embeddings=False,
|
|
|
|
|
|
|
|
|
|
| 46 |
**kwargs,
|
| 47 |
):
|
| 48 |
+
self.eos_token_id = eos_token_id
|
| 49 |
+
super().__init__(
|
| 50 |
+
eos_token_id=eos_token_id, tie_word_embeddings=tie_word_embeddings, **kwargs
|
| 51 |
+
)
|
| 52 |
+
|
| 53 |
self.vocab_size = vocab_size
|
| 54 |
+
self.n_embd = n_embd
|
| 55 |
+
self.n_layer = n_layer
|
| 56 |
+
self.n_head = n_head
|
| 57 |
+
self.n_inner = n_inner
|
| 58 |
+
self.embd_pdrop = embd_pdrop
|
| 59 |
+
self.attn_pdrop = attn_pdrop
|
| 60 |
self.layer_norm_epsilon = layer_norm_epsilon
|
| 61 |
self.initializer_range = initializer_range
|
| 62 |
self.scale_attn_weights = scale_attn_weights
|
| 63 |
self.use_cache = use_cache
|
| 64 |
+
self.apply_residual_connection_post_layernorm = (
|
| 65 |
+
apply_residual_connection_post_layernorm
|
| 66 |
+
)
|
| 67 |
self.bf16 = bf16
|
| 68 |
self.fp16 = fp16
|
| 69 |
self.fp32 = fp32
|
|
|
|
| 73 |
self.use_dynamic_ntk = use_dynamic_ntk
|
| 74 |
self.use_logn_attn = use_logn_attn
|
| 75 |
self.use_flash_attn = use_flash_attn
|
| 76 |
+
self.ffn_hidden_size = ffn_hidden_size
|
| 77 |
self.no_bias = no_bias
|
| 78 |
+
self.tie_word_embeddings = tie_word_embeddings
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
cpp_kernels.py
DELETED
|
@@ -1,55 +0,0 @@
|
|
| 1 |
-
from torch.utils import cpp_extension
|
| 2 |
-
import pathlib
|
| 3 |
-
import os
|
| 4 |
-
import subprocess
|
| 5 |
-
|
| 6 |
-
def _get_cuda_bare_metal_version(cuda_dir):
|
| 7 |
-
raw_output = subprocess.check_output([cuda_dir + "/bin/nvcc", "-V"],
|
| 8 |
-
universal_newlines=True)
|
| 9 |
-
output = raw_output.split()
|
| 10 |
-
release_idx = output.index("release") + 1
|
| 11 |
-
release = output[release_idx].split(".")
|
| 12 |
-
bare_metal_major = release[0]
|
| 13 |
-
bare_metal_minor = release[1][0]
|
| 14 |
-
|
| 15 |
-
return raw_output, bare_metal_major, bare_metal_minor
|
| 16 |
-
|
| 17 |
-
def _create_build_dir(buildpath):
|
| 18 |
-
try:
|
| 19 |
-
os.mkdir(buildpath)
|
| 20 |
-
except OSError:
|
| 21 |
-
if not os.path.isdir(buildpath):
|
| 22 |
-
print(f"Creation of the build directory {buildpath} failed")
|
| 23 |
-
|
| 24 |
-
# Check if cuda 11 is installed for compute capability 8.0
|
| 25 |
-
cc_flag = []
|
| 26 |
-
_, bare_metal_major, bare_metal_minor = _get_cuda_bare_metal_version(cpp_extension.CUDA_HOME)
|
| 27 |
-
if int(bare_metal_major) >= 11:
|
| 28 |
-
cc_flag.append('-gencode')
|
| 29 |
-
cc_flag.append('arch=compute_80,code=sm_80')
|
| 30 |
-
if int(bare_metal_minor) >= 7:
|
| 31 |
-
cc_flag.append('-gencode')
|
| 32 |
-
cc_flag.append('arch=compute_90,code=sm_90')
|
| 33 |
-
|
| 34 |
-
# Build path
|
| 35 |
-
srcpath = pathlib.Path(__file__).parent.absolute()
|
| 36 |
-
buildpath = srcpath / 'build'
|
| 37 |
-
_create_build_dir(buildpath)
|
| 38 |
-
|
| 39 |
-
def _cpp_extention_load_helper(name, sources, extra_cuda_flags):
|
| 40 |
-
return cpp_extension.load(
|
| 41 |
-
name=name,
|
| 42 |
-
sources=sources,
|
| 43 |
-
build_directory=buildpath,
|
| 44 |
-
extra_cflags=['-O3', ],
|
| 45 |
-
extra_cuda_cflags=['-O3',
|
| 46 |
-
'-gencode', 'arch=compute_70,code=sm_70',
|
| 47 |
-
'--use_fast_math'] + extra_cuda_flags + cc_flag,
|
| 48 |
-
verbose=1
|
| 49 |
-
)
|
| 50 |
-
|
| 51 |
-
extra_flags = []
|
| 52 |
-
|
| 53 |
-
cache_autogptq_cuda_256_sources = ["./cache_autogptq_cuda_256.cpp",
|
| 54 |
-
"./cache_autogptq_cuda_kernel_256.cu"]
|
| 55 |
-
cache_autogptq_cuda_256 = _cpp_extention_load_helper("cache_autogptq_cuda_256", cache_autogptq_cuda_256_sources, extra_flags)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
generation_config.json
CHANGED
|
@@ -1,11 +1,16 @@
|
|
| 1 |
{
|
| 2 |
"chat_format": "raw",
|
|
|
|
|
|
|
| 3 |
"eos_token_id": 151643,
|
|
|
|
|
|
|
|
|
|
|
|
|
| 4 |
"pad_token_id": 151643,
|
| 5 |
"stop_words_ids": [[151643]],
|
| 6 |
-
"max_new_tokens": 512,
|
| 7 |
"do_sample": true,
|
| 8 |
"top_k": 0,
|
| 9 |
"top_p": 0.8,
|
| 10 |
"transformers_version": "4.31.0"
|
| 11 |
-
}
|
|
|
|
| 1 |
{
|
| 2 |
"chat_format": "raw",
|
| 3 |
+
"decay_bound": 0.0,
|
| 4 |
+
"decay_factor": 1.0,
|
| 5 |
"eos_token_id": 151643,
|
| 6 |
+
"factual_nucleus_sampling": false,
|
| 7 |
+
"max_context_size": 1024,
|
| 8 |
+
"max_generate_size": 512,
|
| 9 |
+
"max_new_tokens": 512,
|
| 10 |
"pad_token_id": 151643,
|
| 11 |
"stop_words_ids": [[151643]],
|
|
|
|
| 12 |
"do_sample": true,
|
| 13 |
"top_k": 0,
|
| 14 |
"top_p": 0.8,
|
| 15 |
"transformers_version": "4.31.0"
|
| 16 |
+
}
|
model-00002-of-00008.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:3dedac66034371aa3b284a7886e9ce0fde9245ebac60f507f089b33ef82a2912
|
| 3 |
-
size 2023960808
|
|
|
|
|
|
|
|
|
|
|
|
model-00003-of-00008.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:81b25b14a58b62300d11b16c66933a8b631400bf846b27a2a5c5344629cd26e8
|
| 3 |
-
size 2023960816
|
|
|
|
|
|
|
|
|
|
|
|
model-00004-of-00008.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:59a22cb822f9e6d0a6a8415a7bab7b8448ce9672fc71f09266db264afc26ce48
|
| 3 |
-
size 2023960848
|
|
|
|
|
|
|
|
|
|
|
|
model-00005-of-00008.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:e61126f7ad8c520112c49808a73d59bee18c6da7693d9a963a4867f389c91e4a
|
| 3 |
-
size 2023960848
|
|
|
|
|
|
|
|
|
|
|
|
model-00006-of-00008.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:0ced56cc3265fe03ee73658f845dc23c713389b5d68087e918d24d0e2dea624f
|
| 3 |
-
size 2023960848
|
|
|
|
|
|
|
|
|
|
|
|
model-00007-of-00008.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:b0999d47ea087bf79a075ed10889aa3497caff2356200204b032fba208109517
|
| 3 |
-
size 2023960848
|
|
|
|
|
|
|
|
|
|
|
|
model-00008-of-00008.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:7bc05473a78ade06d526cc206e4a01722563abe99099367cdcbb9b3bf670a5de
|
| 3 |
-
size 1334845784
|
|
|
|
|
|
|
|
|
|
|
|
model.safetensors.index.json
DELETED
|
@@ -1,266 +0,0 @@
|
|
| 1 |
-
{
|
| 2 |
-
"metadata": {
|
| 3 |
-
"total_size": 15442649088
|
| 4 |
-
},
|
| 5 |
-
"weight_map": {
|
| 6 |
-
"lm_head.weight": "model-00008-of-00008.safetensors",
|
| 7 |
-
"transformer.h.0.attn.c_attn.bias": "model-00001-of-00008.safetensors",
|
| 8 |
-
"transformer.h.0.attn.c_attn.weight": "model-00001-of-00008.safetensors",
|
| 9 |
-
"transformer.h.0.attn.c_proj.weight": "model-00001-of-00008.safetensors",
|
| 10 |
-
"transformer.h.0.ln_1.weight": "model-00001-of-00008.safetensors",
|
| 11 |
-
"transformer.h.0.ln_2.weight": "model-00001-of-00008.safetensors",
|
| 12 |
-
"transformer.h.0.mlp.c_proj.weight": "model-00001-of-00008.safetensors",
|
| 13 |
-
"transformer.h.0.mlp.w1.weight": "model-00001-of-00008.safetensors",
|
| 14 |
-
"transformer.h.0.mlp.w2.weight": "model-00001-of-00008.safetensors",
|
| 15 |
-
"transformer.h.1.attn.c_attn.bias": "model-00001-of-00008.safetensors",
|
| 16 |
-
"transformer.h.1.attn.c_attn.weight": "model-00001-of-00008.safetensors",
|
| 17 |
-
"transformer.h.1.attn.c_proj.weight": "model-00001-of-00008.safetensors",
|
| 18 |
-
"transformer.h.1.ln_1.weight": "model-00001-of-00008.safetensors",
|
| 19 |
-
"transformer.h.1.ln_2.weight": "model-00001-of-00008.safetensors",
|
| 20 |
-
"transformer.h.1.mlp.c_proj.weight": "model-00002-of-00008.safetensors",
|
| 21 |
-
"transformer.h.1.mlp.w1.weight": "model-00001-of-00008.safetensors",
|
| 22 |
-
"transformer.h.1.mlp.w2.weight": "model-00001-of-00008.safetensors",
|
| 23 |
-
"transformer.h.10.attn.c_attn.bias": "model-00003-of-00008.safetensors",
|
| 24 |
-
"transformer.h.10.attn.c_attn.weight": "model-00003-of-00008.safetensors",
|
| 25 |
-
"transformer.h.10.attn.c_proj.weight": "model-00003-of-00008.safetensors",
|
| 26 |
-
"transformer.h.10.ln_1.weight": "model-00003-of-00008.safetensors",
|
| 27 |
-
"transformer.h.10.ln_2.weight": "model-00003-of-00008.safetensors",
|
| 28 |
-
"transformer.h.10.mlp.c_proj.weight": "model-00003-of-00008.safetensors",
|
| 29 |
-
"transformer.h.10.mlp.w1.weight": "model-00003-of-00008.safetensors",
|
| 30 |
-
"transformer.h.10.mlp.w2.weight": "model-00003-of-00008.safetensors",
|
| 31 |
-
"transformer.h.11.attn.c_attn.bias": "model-00003-of-00008.safetensors",
|
| 32 |
-
"transformer.h.11.attn.c_attn.weight": "model-00003-of-00008.safetensors",
|
| 33 |
-
"transformer.h.11.attn.c_proj.weight": "model-00003-of-00008.safetensors",
|
| 34 |
-
"transformer.h.11.ln_1.weight": "model-00003-of-00008.safetensors",
|
| 35 |
-
"transformer.h.11.ln_2.weight": "model-00003-of-00008.safetensors",
|
| 36 |
-
"transformer.h.11.mlp.c_proj.weight": "model-00004-of-00008.safetensors",
|
| 37 |
-
"transformer.h.11.mlp.w1.weight": "model-00003-of-00008.safetensors",
|
| 38 |
-
"transformer.h.11.mlp.w2.weight": "model-00003-of-00008.safetensors",
|
| 39 |
-
"transformer.h.12.attn.c_attn.bias": "model-00004-of-00008.safetensors",
|
| 40 |
-
"transformer.h.12.attn.c_attn.weight": "model-00004-of-00008.safetensors",
|
| 41 |
-
"transformer.h.12.attn.c_proj.weight": "model-00004-of-00008.safetensors",
|
| 42 |
-
"transformer.h.12.ln_1.weight": "model-00004-of-00008.safetensors",
|
| 43 |
-
"transformer.h.12.ln_2.weight": "model-00004-of-00008.safetensors",
|
| 44 |
-
"transformer.h.12.mlp.c_proj.weight": "model-00004-of-00008.safetensors",
|
| 45 |
-
"transformer.h.12.mlp.w1.weight": "model-00004-of-00008.safetensors",
|
| 46 |
-
"transformer.h.12.mlp.w2.weight": "model-00004-of-00008.safetensors",
|
| 47 |
-
"transformer.h.13.attn.c_attn.bias": "model-00004-of-00008.safetensors",
|
| 48 |
-
"transformer.h.13.attn.c_attn.weight": "model-00004-of-00008.safetensors",
|
| 49 |
-
"transformer.h.13.attn.c_proj.weight": "model-00004-of-00008.safetensors",
|
| 50 |
-
"transformer.h.13.ln_1.weight": "model-00004-of-00008.safetensors",
|
| 51 |
-
"transformer.h.13.ln_2.weight": "model-00004-of-00008.safetensors",
|
| 52 |
-
"transformer.h.13.mlp.c_proj.weight": "model-00004-of-00008.safetensors",
|
| 53 |
-
"transformer.h.13.mlp.w1.weight": "model-00004-of-00008.safetensors",
|
| 54 |
-
"transformer.h.13.mlp.w2.weight": "model-00004-of-00008.safetensors",
|
| 55 |
-
"transformer.h.14.attn.c_attn.bias": "model-00004-of-00008.safetensors",
|
| 56 |
-
"transformer.h.14.attn.c_attn.weight": "model-00004-of-00008.safetensors",
|
| 57 |
-
"transformer.h.14.attn.c_proj.weight": "model-00004-of-00008.safetensors",
|
| 58 |
-
"transformer.h.14.ln_1.weight": "model-00004-of-00008.safetensors",
|
| 59 |
-
"transformer.h.14.ln_2.weight": "model-00004-of-00008.safetensors",
|
| 60 |
-
"transformer.h.14.mlp.c_proj.weight": "model-00004-of-00008.safetensors",
|
| 61 |
-
"transformer.h.14.mlp.w1.weight": "model-00004-of-00008.safetensors",
|
| 62 |
-
"transformer.h.14.mlp.w2.weight": "model-00004-of-00008.safetensors",
|
| 63 |
-
"transformer.h.15.attn.c_attn.bias": "model-00004-of-00008.safetensors",
|
| 64 |
-
"transformer.h.15.attn.c_attn.weight": "model-00004-of-00008.safetensors",
|
| 65 |
-
"transformer.h.15.attn.c_proj.weight": "model-00004-of-00008.safetensors",
|
| 66 |
-
"transformer.h.15.ln_1.weight": "model-00004-of-00008.safetensors",
|
| 67 |
-
"transformer.h.15.ln_2.weight": "model-00004-of-00008.safetensors",
|
| 68 |
-
"transformer.h.15.mlp.c_proj.weight": "model-00004-of-00008.safetensors",
|
| 69 |
-
"transformer.h.15.mlp.w1.weight": "model-00004-of-00008.safetensors",
|
| 70 |
-
"transformer.h.15.mlp.w2.weight": "model-00004-of-00008.safetensors",
|
| 71 |
-
"transformer.h.16.attn.c_attn.bias": "model-00004-of-00008.safetensors",
|
| 72 |
-
"transformer.h.16.attn.c_attn.weight": "model-00004-of-00008.safetensors",
|
| 73 |
-
"transformer.h.16.attn.c_proj.weight": "model-00004-of-00008.safetensors",
|
| 74 |
-
"transformer.h.16.ln_1.weight": "model-00004-of-00008.safetensors",
|
| 75 |
-
"transformer.h.16.ln_2.weight": "model-00004-of-00008.safetensors",
|
| 76 |
-
"transformer.h.16.mlp.c_proj.weight": "model-00005-of-00008.safetensors",
|
| 77 |
-
"transformer.h.16.mlp.w1.weight": "model-00004-of-00008.safetensors",
|
| 78 |
-
"transformer.h.16.mlp.w2.weight": "model-00004-of-00008.safetensors",
|
| 79 |
-
"transformer.h.17.attn.c_attn.bias": "model-00005-of-00008.safetensors",
|
| 80 |
-
"transformer.h.17.attn.c_attn.weight": "model-00005-of-00008.safetensors",
|
| 81 |
-
"transformer.h.17.attn.c_proj.weight": "model-00005-of-00008.safetensors",
|
| 82 |
-
"transformer.h.17.ln_1.weight": "model-00005-of-00008.safetensors",
|
| 83 |
-
"transformer.h.17.ln_2.weight": "model-00005-of-00008.safetensors",
|
| 84 |
-
"transformer.h.17.mlp.c_proj.weight": "model-00005-of-00008.safetensors",
|
| 85 |
-
"transformer.h.17.mlp.w1.weight": "model-00005-of-00008.safetensors",
|
| 86 |
-
"transformer.h.17.mlp.w2.weight": "model-00005-of-00008.safetensors",
|
| 87 |
-
"transformer.h.18.attn.c_attn.bias": "model-00005-of-00008.safetensors",
|
| 88 |
-
"transformer.h.18.attn.c_attn.weight": "model-00005-of-00008.safetensors",
|
| 89 |
-
"transformer.h.18.attn.c_proj.weight": "model-00005-of-00008.safetensors",
|
| 90 |
-
"transformer.h.18.ln_1.weight": "model-00005-of-00008.safetensors",
|
| 91 |
-
"transformer.h.18.ln_2.weight": "model-00005-of-00008.safetensors",
|
| 92 |
-
"transformer.h.18.mlp.c_proj.weight": "model-00005-of-00008.safetensors",
|
| 93 |
-
"transformer.h.18.mlp.w1.weight": "model-00005-of-00008.safetensors",
|
| 94 |
-
"transformer.h.18.mlp.w2.weight": "model-00005-of-00008.safetensors",
|
| 95 |
-
"transformer.h.19.attn.c_attn.bias": "model-00005-of-00008.safetensors",
|
| 96 |
-
"transformer.h.19.attn.c_attn.weight": "model-00005-of-00008.safetensors",
|
| 97 |
-
"transformer.h.19.attn.c_proj.weight": "model-00005-of-00008.safetensors",
|
| 98 |
-
"transformer.h.19.ln_1.weight": "model-00005-of-00008.safetensors",
|
| 99 |
-
"transformer.h.19.ln_2.weight": "model-00005-of-00008.safetensors",
|
| 100 |
-
"transformer.h.19.mlp.c_proj.weight": "model-00005-of-00008.safetensors",
|
| 101 |
-
"transformer.h.19.mlp.w1.weight": "model-00005-of-00008.safetensors",
|
| 102 |
-
"transformer.h.19.mlp.w2.weight": "model-00005-of-00008.safetensors",
|
| 103 |
-
"transformer.h.2.attn.c_attn.bias": "model-00002-of-00008.safetensors",
|
| 104 |
-
"transformer.h.2.attn.c_attn.weight": "model-00002-of-00008.safetensors",
|
| 105 |
-
"transformer.h.2.attn.c_proj.weight": "model-00002-of-00008.safetensors",
|
| 106 |
-
"transformer.h.2.ln_1.weight": "model-00002-of-00008.safetensors",
|
| 107 |
-
"transformer.h.2.ln_2.weight": "model-00002-of-00008.safetensors",
|
| 108 |
-
"transformer.h.2.mlp.c_proj.weight": "model-00002-of-00008.safetensors",
|
| 109 |
-
"transformer.h.2.mlp.w1.weight": "model-00002-of-00008.safetensors",
|
| 110 |
-
"transformer.h.2.mlp.w2.weight": "model-00002-of-00008.safetensors",
|
| 111 |
-
"transformer.h.20.attn.c_attn.bias": "model-00005-of-00008.safetensors",
|
| 112 |
-
"transformer.h.20.attn.c_attn.weight": "model-00005-of-00008.safetensors",
|
| 113 |
-
"transformer.h.20.attn.c_proj.weight": "model-00005-of-00008.safetensors",
|
| 114 |
-
"transformer.h.20.ln_1.weight": "model-00005-of-00008.safetensors",
|
| 115 |
-
"transformer.h.20.ln_2.weight": "model-00005-of-00008.safetensors",
|
| 116 |
-
"transformer.h.20.mlp.c_proj.weight": "model-00005-of-00008.safetensors",
|
| 117 |
-
"transformer.h.20.mlp.w1.weight": "model-00005-of-00008.safetensors",
|
| 118 |
-
"transformer.h.20.mlp.w2.weight": "model-00005-of-00008.safetensors",
|
| 119 |
-
"transformer.h.21.attn.c_attn.bias": "model-00005-of-00008.safetensors",
|
| 120 |
-
"transformer.h.21.attn.c_attn.weight": "model-00005-of-00008.safetensors",
|
| 121 |
-
"transformer.h.21.attn.c_proj.weight": "model-00005-of-00008.safetensors",
|
| 122 |
-
"transformer.h.21.ln_1.weight": "model-00005-of-00008.safetensors",
|
| 123 |
-
"transformer.h.21.ln_2.weight": "model-00005-of-00008.safetensors",
|
| 124 |
-
"transformer.h.21.mlp.c_proj.weight": "model-00006-of-00008.safetensors",
|
| 125 |
-
"transformer.h.21.mlp.w1.weight": "model-00005-of-00008.safetensors",
|
| 126 |
-
"transformer.h.21.mlp.w2.weight": "model-00005-of-00008.safetensors",
|
| 127 |
-
"transformer.h.22.attn.c_attn.bias": "model-00006-of-00008.safetensors",
|
| 128 |
-
"transformer.h.22.attn.c_attn.weight": "model-00006-of-00008.safetensors",
|
| 129 |
-
"transformer.h.22.attn.c_proj.weight": "model-00006-of-00008.safetensors",
|
| 130 |
-
"transformer.h.22.ln_1.weight": "model-00006-of-00008.safetensors",
|
| 131 |
-
"transformer.h.22.ln_2.weight": "model-00006-of-00008.safetensors",
|
| 132 |
-
"transformer.h.22.mlp.c_proj.weight": "model-00006-of-00008.safetensors",
|
| 133 |
-
"transformer.h.22.mlp.w1.weight": "model-00006-of-00008.safetensors",
|
| 134 |
-
"transformer.h.22.mlp.w2.weight": "model-00006-of-00008.safetensors",
|
| 135 |
-
"transformer.h.23.attn.c_attn.bias": "model-00006-of-00008.safetensors",
|
| 136 |
-
"transformer.h.23.attn.c_attn.weight": "model-00006-of-00008.safetensors",
|
| 137 |
-
"transformer.h.23.attn.c_proj.weight": "model-00006-of-00008.safetensors",
|
| 138 |
-
"transformer.h.23.ln_1.weight": "model-00006-of-00008.safetensors",
|
| 139 |
-
"transformer.h.23.ln_2.weight": "model-00006-of-00008.safetensors",
|
| 140 |
-
"transformer.h.23.mlp.c_proj.weight": "model-00006-of-00008.safetensors",
|
| 141 |
-
"transformer.h.23.mlp.w1.weight": "model-00006-of-00008.safetensors",
|
| 142 |
-
"transformer.h.23.mlp.w2.weight": "model-00006-of-00008.safetensors",
|
| 143 |
-
"transformer.h.24.attn.c_attn.bias": "model-00006-of-00008.safetensors",
|
| 144 |
-
"transformer.h.24.attn.c_attn.weight": "model-00006-of-00008.safetensors",
|
| 145 |
-
"transformer.h.24.attn.c_proj.weight": "model-00006-of-00008.safetensors",
|
| 146 |
-
"transformer.h.24.ln_1.weight": "model-00006-of-00008.safetensors",
|
| 147 |
-
"transformer.h.24.ln_2.weight": "model-00006-of-00008.safetensors",
|
| 148 |
-
"transformer.h.24.mlp.c_proj.weight": "model-00006-of-00008.safetensors",
|
| 149 |
-
"transformer.h.24.mlp.w1.weight": "model-00006-of-00008.safetensors",
|
| 150 |
-
"transformer.h.24.mlp.w2.weight": "model-00006-of-00008.safetensors",
|
| 151 |
-
"transformer.h.25.attn.c_attn.bias": "model-00006-of-00008.safetensors",
|
| 152 |
-
"transformer.h.25.attn.c_attn.weight": "model-00006-of-00008.safetensors",
|
| 153 |
-
"transformer.h.25.attn.c_proj.weight": "model-00006-of-00008.safetensors",
|
| 154 |
-
"transformer.h.25.ln_1.weight": "model-00006-of-00008.safetensors",
|
| 155 |
-
"transformer.h.25.ln_2.weight": "model-00006-of-00008.safetensors",
|
| 156 |
-
"transformer.h.25.mlp.c_proj.weight": "model-00006-of-00008.safetensors",
|
| 157 |
-
"transformer.h.25.mlp.w1.weight": "model-00006-of-00008.safetensors",
|
| 158 |
-
"transformer.h.25.mlp.w2.weight": "model-00006-of-00008.safetensors",
|
| 159 |
-
"transformer.h.26.attn.c_attn.bias": "model-00006-of-00008.safetensors",
|
| 160 |
-
"transformer.h.26.attn.c_attn.weight": "model-00006-of-00008.safetensors",
|
| 161 |
-
"transformer.h.26.attn.c_proj.weight": "model-00006-of-00008.safetensors",
|
| 162 |
-
"transformer.h.26.ln_1.weight": "model-00006-of-00008.safetensors",
|
| 163 |
-
"transformer.h.26.ln_2.weight": "model-00006-of-00008.safetensors",
|
| 164 |
-
"transformer.h.26.mlp.c_proj.weight": "model-00007-of-00008.safetensors",
|
| 165 |
-
"transformer.h.26.mlp.w1.weight": "model-00006-of-00008.safetensors",
|
| 166 |
-
"transformer.h.26.mlp.w2.weight": "model-00006-of-00008.safetensors",
|
| 167 |
-
"transformer.h.27.attn.c_attn.bias": "model-00007-of-00008.safetensors",
|
| 168 |
-
"transformer.h.27.attn.c_attn.weight": "model-00007-of-00008.safetensors",
|
| 169 |
-
"transformer.h.27.attn.c_proj.weight": "model-00007-of-00008.safetensors",
|
| 170 |
-
"transformer.h.27.ln_1.weight": "model-00007-of-00008.safetensors",
|
| 171 |
-
"transformer.h.27.ln_2.weight": "model-00007-of-00008.safetensors",
|
| 172 |
-
"transformer.h.27.mlp.c_proj.weight": "model-00007-of-00008.safetensors",
|
| 173 |
-
"transformer.h.27.mlp.w1.weight": "model-00007-of-00008.safetensors",
|
| 174 |
-
"transformer.h.27.mlp.w2.weight": "model-00007-of-00008.safetensors",
|
| 175 |
-
"transformer.h.28.attn.c_attn.bias": "model-00007-of-00008.safetensors",
|
| 176 |
-
"transformer.h.28.attn.c_attn.weight": "model-00007-of-00008.safetensors",
|
| 177 |
-
"transformer.h.28.attn.c_proj.weight": "model-00007-of-00008.safetensors",
|
| 178 |
-
"transformer.h.28.ln_1.weight": "model-00007-of-00008.safetensors",
|
| 179 |
-
"transformer.h.28.ln_2.weight": "model-00007-of-00008.safetensors",
|
| 180 |
-
"transformer.h.28.mlp.c_proj.weight": "model-00007-of-00008.safetensors",
|
| 181 |
-
"transformer.h.28.mlp.w1.weight": "model-00007-of-00008.safetensors",
|
| 182 |
-
"transformer.h.28.mlp.w2.weight": "model-00007-of-00008.safetensors",
|
| 183 |
-
"transformer.h.29.attn.c_attn.bias": "model-00007-of-00008.safetensors",
|
| 184 |
-
"transformer.h.29.attn.c_attn.weight": "model-00007-of-00008.safetensors",
|
| 185 |
-
"transformer.h.29.attn.c_proj.weight": "model-00007-of-00008.safetensors",
|
| 186 |
-
"transformer.h.29.ln_1.weight": "model-00007-of-00008.safetensors",
|
| 187 |
-
"transformer.h.29.ln_2.weight": "model-00007-of-00008.safetensors",
|
| 188 |
-
"transformer.h.29.mlp.c_proj.weight": "model-00007-of-00008.safetensors",
|
| 189 |
-
"transformer.h.29.mlp.w1.weight": "model-00007-of-00008.safetensors",
|
| 190 |
-
"transformer.h.29.mlp.w2.weight": "model-00007-of-00008.safetensors",
|
| 191 |
-
"transformer.h.3.attn.c_attn.bias": "model-00002-of-00008.safetensors",
|
| 192 |
-
"transformer.h.3.attn.c_attn.weight": "model-00002-of-00008.safetensors",
|
| 193 |
-
"transformer.h.3.attn.c_proj.weight": "model-00002-of-00008.safetensors",
|
| 194 |
-
"transformer.h.3.ln_1.weight": "model-00002-of-00008.safetensors",
|
| 195 |
-
"transformer.h.3.ln_2.weight": "model-00002-of-00008.safetensors",
|
| 196 |
-
"transformer.h.3.mlp.c_proj.weight": "model-00002-of-00008.safetensors",
|
| 197 |
-
"transformer.h.3.mlp.w1.weight": "model-00002-of-00008.safetensors",
|
| 198 |
-
"transformer.h.3.mlp.w2.weight": "model-00002-of-00008.safetensors",
|
| 199 |
-
"transformer.h.30.attn.c_attn.bias": "model-00007-of-00008.safetensors",
|
| 200 |
-
"transformer.h.30.attn.c_attn.weight": "model-00007-of-00008.safetensors",
|
| 201 |
-
"transformer.h.30.attn.c_proj.weight": "model-00007-of-00008.safetensors",
|
| 202 |
-
"transformer.h.30.ln_1.weight": "model-00007-of-00008.safetensors",
|
| 203 |
-
"transformer.h.30.ln_2.weight": "model-00007-of-00008.safetensors",
|
| 204 |
-
"transformer.h.30.mlp.c_proj.weight": "model-00007-of-00008.safetensors",
|
| 205 |
-
"transformer.h.30.mlp.w1.weight": "model-00007-of-00008.safetensors",
|
| 206 |
-
"transformer.h.30.mlp.w2.weight": "model-00007-of-00008.safetensors",
|
| 207 |
-
"transformer.h.31.attn.c_attn.bias": "model-00007-of-00008.safetensors",
|
| 208 |
-
"transformer.h.31.attn.c_attn.weight": "model-00007-of-00008.safetensors",
|
| 209 |
-
"transformer.h.31.attn.c_proj.weight": "model-00007-of-00008.safetensors",
|
| 210 |
-
"transformer.h.31.ln_1.weight": "model-00007-of-00008.safetensors",
|
| 211 |
-
"transformer.h.31.ln_2.weight": "model-00007-of-00008.safetensors",
|
| 212 |
-
"transformer.h.31.mlp.c_proj.weight": "model-00008-of-00008.safetensors",
|
| 213 |
-
"transformer.h.31.mlp.w1.weight": "model-00007-of-00008.safetensors",
|
| 214 |
-
"transformer.h.31.mlp.w2.weight": "model-00007-of-00008.safetensors",
|
| 215 |
-
"transformer.h.4.attn.c_attn.bias": "model-00002-of-00008.safetensors",
|
| 216 |
-
"transformer.h.4.attn.c_attn.weight": "model-00002-of-00008.safetensors",
|
| 217 |
-
"transformer.h.4.attn.c_proj.weight": "model-00002-of-00008.safetensors",
|
| 218 |
-
"transformer.h.4.ln_1.weight": "model-00002-of-00008.safetensors",
|
| 219 |
-
"transformer.h.4.ln_2.weight": "model-00002-of-00008.safetensors",
|
| 220 |
-
"transformer.h.4.mlp.c_proj.weight": "model-00002-of-00008.safetensors",
|
| 221 |
-
"transformer.h.4.mlp.w1.weight": "model-00002-of-00008.safetensors",
|
| 222 |
-
"transformer.h.4.mlp.w2.weight": "model-00002-of-00008.safetensors",
|
| 223 |
-
"transformer.h.5.attn.c_attn.bias": "model-00002-of-00008.safetensors",
|
| 224 |
-
"transformer.h.5.attn.c_attn.weight": "model-00002-of-00008.safetensors",
|
| 225 |
-
"transformer.h.5.attn.c_proj.weight": "model-00002-of-00008.safetensors",
|
| 226 |
-
"transformer.h.5.ln_1.weight": "model-00002-of-00008.safetensors",
|
| 227 |
-
"transformer.h.5.ln_2.weight": "model-00002-of-00008.safetensors",
|
| 228 |
-
"transformer.h.5.mlp.c_proj.weight": "model-00002-of-00008.safetensors",
|
| 229 |
-
"transformer.h.5.mlp.w1.weight": "model-00002-of-00008.safetensors",
|
| 230 |
-
"transformer.h.5.mlp.w2.weight": "model-00002-of-00008.safetensors",
|
| 231 |
-
"transformer.h.6.attn.c_attn.bias": "model-00002-of-00008.safetensors",
|
| 232 |
-
"transformer.h.6.attn.c_attn.weight": "model-00002-of-00008.safetensors",
|
| 233 |
-
"transformer.h.6.attn.c_proj.weight": "model-00002-of-00008.safetensors",
|
| 234 |
-
"transformer.h.6.ln_1.weight": "model-00002-of-00008.safetensors",
|
| 235 |
-
"transformer.h.6.ln_2.weight": "model-00002-of-00008.safetensors",
|
| 236 |
-
"transformer.h.6.mlp.c_proj.weight": "model-00003-of-00008.safetensors",
|
| 237 |
-
"transformer.h.6.mlp.w1.weight": "model-00002-of-00008.safetensors",
|
| 238 |
-
"transformer.h.6.mlp.w2.weight": "model-00002-of-00008.safetensors",
|
| 239 |
-
"transformer.h.7.attn.c_attn.bias": "model-00003-of-00008.safetensors",
|
| 240 |
-
"transformer.h.7.attn.c_attn.weight": "model-00003-of-00008.safetensors",
|
| 241 |
-
"transformer.h.7.attn.c_proj.weight": "model-00003-of-00008.safetensors",
|
| 242 |
-
"transformer.h.7.ln_1.weight": "model-00003-of-00008.safetensors",
|
| 243 |
-
"transformer.h.7.ln_2.weight": "model-00003-of-00008.safetensors",
|
| 244 |
-
"transformer.h.7.mlp.c_proj.weight": "model-00003-of-00008.safetensors",
|
| 245 |
-
"transformer.h.7.mlp.w1.weight": "model-00003-of-00008.safetensors",
|
| 246 |
-
"transformer.h.7.mlp.w2.weight": "model-00003-of-00008.safetensors",
|
| 247 |
-
"transformer.h.8.attn.c_attn.bias": "model-00003-of-00008.safetensors",
|
| 248 |
-
"transformer.h.8.attn.c_attn.weight": "model-00003-of-00008.safetensors",
|
| 249 |
-
"transformer.h.8.attn.c_proj.weight": "model-00003-of-00008.safetensors",
|
| 250 |
-
"transformer.h.8.ln_1.weight": "model-00003-of-00008.safetensors",
|
| 251 |
-
"transformer.h.8.ln_2.weight": "model-00003-of-00008.safetensors",
|
| 252 |
-
"transformer.h.8.mlp.c_proj.weight": "model-00003-of-00008.safetensors",
|
| 253 |
-
"transformer.h.8.mlp.w1.weight": "model-00003-of-00008.safetensors",
|
| 254 |
-
"transformer.h.8.mlp.w2.weight": "model-00003-of-00008.safetensors",
|
| 255 |
-
"transformer.h.9.attn.c_attn.bias": "model-00003-of-00008.safetensors",
|
| 256 |
-
"transformer.h.9.attn.c_attn.weight": "model-00003-of-00008.safetensors",
|
| 257 |
-
"transformer.h.9.attn.c_proj.weight": "model-00003-of-00008.safetensors",
|
| 258 |
-
"transformer.h.9.ln_1.weight": "model-00003-of-00008.safetensors",
|
| 259 |
-
"transformer.h.9.ln_2.weight": "model-00003-of-00008.safetensors",
|
| 260 |
-
"transformer.h.9.mlp.c_proj.weight": "model-00003-of-00008.safetensors",
|
| 261 |
-
"transformer.h.9.mlp.w1.weight": "model-00003-of-00008.safetensors",
|
| 262 |
-
"transformer.h.9.mlp.w2.weight": "model-00003-of-00008.safetensors",
|
| 263 |
-
"transformer.ln_f.weight": "model-00008-of-00008.safetensors",
|
| 264 |
-
"transformer.wte.weight": "model-00001-of-00008.safetensors"
|
| 265 |
-
}
|
| 266 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
modeling_qwen.py
CHANGED
|
@@ -3,16 +3,14 @@
|
|
| 3 |
# This source code is licensed under the license found in the
|
| 4 |
# LICENSE file in the root directory of this source tree.
|
| 5 |
|
| 6 |
-
import copy
|
| 7 |
import importlib
|
| 8 |
import math
|
| 9 |
-
import
|
| 10 |
-
from typing import TYPE_CHECKING, Optional, Tuple, Union, Callable, List, Any, Generator
|
| 11 |
|
| 12 |
import torch
|
| 13 |
import torch.nn.functional as F
|
| 14 |
import torch.utils.checkpoint
|
| 15 |
-
import
|
| 16 |
|
| 17 |
from torch.nn import CrossEntropyLoss
|
| 18 |
from transformers import PreTrainedTokenizer, GenerationConfig, StoppingCriteriaList
|
|
@@ -37,8 +35,10 @@ from torch import nn
|
|
| 37 |
SUPPORT_CUDA = torch.cuda.is_available()
|
| 38 |
SUPPORT_BF16 = SUPPORT_CUDA and torch.cuda.is_bf16_supported()
|
| 39 |
SUPPORT_FP16 = SUPPORT_CUDA and torch.cuda.get_device_capability(0)[0] >= 7
|
| 40 |
-
SUPPORT_TORCH2 = hasattr(torch, '__version__') and int(torch.__version__.split(".")[0]) >= 2
|
| 41 |
|
|
|
|
|
|
|
|
|
|
| 42 |
|
| 43 |
from .configuration_qwen import QWenConfig
|
| 44 |
from .qwen_generation_utils import (
|
|
@@ -57,95 +57,6 @@ _CONFIG_FOR_DOC = "QWenConfig"
|
|
| 57 |
|
| 58 |
QWen_PRETRAINED_MODEL_ARCHIVE_LIST = ["qwen-7b"]
|
| 59 |
|
| 60 |
-
_ERROR_BAD_CHAT_FORMAT = """\
|
| 61 |
-
We detect you are probably using the pretrained model (rather than chat model) for chatting, since the chat_format in generation_config is not "chatml".
|
| 62 |
-
If you are directly using the model downloaded from Huggingface, please make sure you are using our "Qwen/Qwen-7B-Chat" Huggingface model (rather than "Qwen/Qwen-7B") when you call model.chat().
|
| 63 |
-
我们检测到您可能在使用预训练模型（而非chat模型）进行多轮chat，因为您当前在generation_config指定的chat_format，并未设置为我们在对话中所支持的"chatml"格式。
|
| 64 |
-
如果您在直接使用我们从Huggingface提供的模型，请确保您在调用model.chat()时，使用的是"Qwen/Qwen-7B-Chat"模型（而非"Qwen/Qwen-7B"预训练模型）。
|
| 65 |
-
"""
|
| 66 |
-
|
| 67 |
-
_SENTINEL = object()
|
| 68 |
-
_ERROR_STREAM_IN_CHAT = """\
|
| 69 |
-
Pass argument `stream` to model.chat() is buggy, deprecated, and marked for removal. Please use model.chat_stream(...) instead of model.chat(..., stream=True).
|
| 70 |
-
向model.chat()传入参数stream的用法可能存在Bug，该用法已被废弃，将在未来被移除。请使用model.chat_stream(...)代替model.chat(..., stream=True)。
|
| 71 |
-
"""
|
| 72 |
-
|
| 73 |
-
_ERROR_INPUT_CPU_QUERY_WITH_FLASH_ATTN_ACTIVATED = """\
|
| 74 |
-
We detect you have activated flash attention support, but running model computation on CPU. Please make sure that your input data has been placed on GPU. If you actually want to run CPU computation, please following the readme and set device_map="cpu" to disable flash attention when loading the model (calling AutoModelForCausalLM.from_pretrained).
|
| 75 |
-
检测到您的模型已激活了flash attention支持，但正在执行CPU运算任务。如使用flash attention，请您确认模型输入已经传到GPU上。如果您确认要执行CPU运算，请您在载入模型（调用AutoModelForCausalLM.from_pretrained）时，按照readme说法，指定device_map="cpu"以禁用flash attention。
|
| 76 |
-
"""
|
| 77 |
-
|
| 78 |
-
apply_rotary_emb_func = None
|
| 79 |
-
rms_norm = None
|
| 80 |
-
flash_attn_unpadded_func = None
|
| 81 |
-
flash_attn_func = None
|
| 82 |
-
|
| 83 |
-
def _import_flash_attn():
|
| 84 |
-
global apply_rotary_emb_func, rms_norm, flash_attn_unpadded_func, flash_attn_func
|
| 85 |
-
try:
|
| 86 |
-
from flash_attn.layers.rotary import apply_rotary_emb_func as __apply_rotary_emb_func
|
| 87 |
-
apply_rotary_emb_func = __apply_rotary_emb_func
|
| 88 |
-
except ImportError:
|
| 89 |
-
logger.warn(
|
| 90 |
-
"Warning: import flash_attn rotary fail, please install FlashAttention rotary to get higher efficiency "
|
| 91 |
-
"https://github.com/Dao-AILab/flash-attention/tree/main/csrc/rotary"
|
| 92 |
-
)
|
| 93 |
-
|
| 94 |
-
try:
|
| 95 |
-
from flash_attn.ops.rms_norm import rms_norm as __rms_norm
|
| 96 |
-
rms_norm = __rms_norm
|
| 97 |
-
except ImportError:
|
| 98 |
-
logger.warn(
|
| 99 |
-
"Warning: import flash_attn rms_norm fail, please install FlashAttention layer_norm to get higher efficiency "
|
| 100 |
-
"https://github.com/Dao-AILab/flash-attention/tree/main/csrc/layer_norm"
|
| 101 |
-
)
|
| 102 |
-
|
| 103 |
-
try:
|
| 104 |
-
import flash_attn
|
| 105 |
-
_flash_attn_func = None
|
| 106 |
-
if not hasattr(flash_attn, '__version__'):
|
| 107 |
-
from flash_attn.flash_attn_interface import flash_attn_unpadded_func as __flash_attn_unpadded_func
|
| 108 |
-
else:
|
| 109 |
-
if int(flash_attn.__version__.split(".")[0]) >= 2:
|
| 110 |
-
if int(flash_attn.__version__.split(".")[1]) >= 1:
|
| 111 |
-
from flash_attn.flash_attn_interface import flash_attn_func as _flash_attn_func
|
| 112 |
-
from flash_attn.flash_attn_interface import flash_attn_varlen_func as __flash_attn_unpadded_func
|
| 113 |
-
else:
|
| 114 |
-
from flash_attn.flash_attn_interface import flash_attn_unpadded_func as __flash_attn_unpadded_func
|
| 115 |
-
flash_attn_unpadded_func = __flash_attn_unpadded_func
|
| 116 |
-
flash_attn_func = _flash_attn_func
|
| 117 |
-
except ImportError:
|
| 118 |
-
logger.warn(
|
| 119 |
-
"Warning: import flash_attn fail, please install FlashAttention to get higher efficiency "
|
| 120 |
-
"https://github.com/Dao-AILab/flash-attention"
|
| 121 |
-
)
|
| 122 |
-
|
| 123 |
-
def quantize_cache_v(fdata, bits, qmax, qmin):
|
| 124 |
-
# b, s, head, h-dim->b, head, s, h-dim
|
| 125 |
-
qtype = torch.uint8
|
| 126 |
-
device = fdata.device
|
| 127 |
-
shape = fdata.shape
|
| 128 |
-
|
| 129 |
-
fdata_cal = torch.flatten(fdata, 2)
|
| 130 |
-
fmax = torch.amax(fdata_cal, dim=-1, keepdim=True)
|
| 131 |
-
fmin = torch.amin(fdata_cal, dim=-1, keepdim=True)
|
| 132 |
-
# Compute params
|
| 133 |
-
if qmax.device != fmax.device:
|
| 134 |
-
qmax = qmax.to(device)
|
| 135 |
-
qmin = qmin.to(device)
|
| 136 |
-
scale = (fmax - fmin) / (qmax - qmin)
|
| 137 |
-
zero = qmin - fmin / scale
|
| 138 |
-
scale = scale.unsqueeze(-1).repeat(1,1,shape[2],1).contiguous()
|
| 139 |
-
zero = zero.unsqueeze(-1).repeat(1,1,shape[2],1).contiguous()
|
| 140 |
-
# Quantize
|
| 141 |
-
res_data = fdata / scale + zero
|
| 142 |
-
qdata = torch.clamp(res_data, qmin, qmax).to(qtype)
|
| 143 |
-
return qdata.contiguous(), scale, zero
|
| 144 |
-
|
| 145 |
-
def dequantize_cache_torch(qdata, scale, zero):
|
| 146 |
-
data = scale * (qdata - zero)
|
| 147 |
-
return data
|
| 148 |
-
|
| 149 |
class FlashSelfAttention(torch.nn.Module):
|
| 150 |
def __init__(
|
| 151 |
self,
|
|
@@ -164,33 +75,11 @@ class FlashSelfAttention(torch.nn.Module):
|
|
| 164 |
self.softmax_scale = softmax_scale
|
| 165 |
self.dropout_p = attention_dropout
|
| 166 |
|
| 167 |
-
def
|
| 168 |
-
valid_mask = attention_mask.squeeze(1).squeeze(1).eq(0)
|
| 169 |
-
seqlens_in_batch = valid_mask.sum(dim=-1, dtype=torch.int32)
|
| 170 |
-
indices = torch.nonzero(valid_mask.flatten(), as_tuple=False).flatten()
|
| 171 |
-
max_seqlen_in_batch = seqlens_in_batch.max().item()
|
| 172 |
-
cu_seqlens = F.pad(torch.cumsum(seqlens_in_batch, dim=0, dtype=torch.torch.int32), (1, 0))
|
| 173 |
-
hidden_states = hidden_states[indices]
|
| 174 |
-
return hidden_states, indices, cu_seqlens, max_seqlen_in_batch
|
| 175 |
-
|
| 176 |
-
def pad_input(self, hidden_states, indices, batch, seqlen):
|
| 177 |
-
output = torch.zeros(batch * seqlen, *hidden_states.shape[1:], device=hidden_states.device,
|
| 178 |
-
dtype=hidden_states.dtype)
|
| 179 |
-
output[indices] = hidden_states
|
| 180 |
-
return rearrange(output, '(b s) ... -> b s ...', b=batch)
|
| 181 |
-
|
| 182 |
-
def forward(self, q, k, v, attention_mask=None):
|
| 183 |
assert all((i.dtype in [torch.float16, torch.bfloat16] for i in (q, k, v)))
|
| 184 |
assert all((i.is_cuda for i in (q, k, v)))
|
| 185 |
batch_size, seqlen_q = q.shape[0], q.shape[1]
|
| 186 |
seqlen_k = k.shape[1]
|
| 187 |
-
seqlen_out = seqlen_q
|
| 188 |
-
|
| 189 |
-
if flash_attn_func is not None and batch_size == 1:
|
| 190 |
-
dropout_p = self.dropout_p if self.training else 0
|
| 191 |
-
output = flash_attn_func(q, k, v, dropout_p, softmax_scale=self.softmax_scale, causal=self.causal)
|
| 192 |
-
return output
|
| 193 |
-
|
| 194 |
q, k, v = [rearrange(x, "b s ... -> (b s) ...") for x in [q, k, v]]
|
| 195 |
cu_seqlens_q = torch.arange(
|
| 196 |
0,
|
|
@@ -200,14 +89,13 @@ class FlashSelfAttention(torch.nn.Module):
|
|
| 200 |
device=q.device,
|
| 201 |
)
|
| 202 |
|
| 203 |
-
if
|
| 204 |
-
|
| 205 |
-
|
| 206 |
-
|
| 207 |
-
|
| 208 |
-
seqlen_q = seqlen_k
|
| 209 |
-
v = v[indices_k]
|
| 210 |
else:
|
|
|
|
| 211 |
cu_seqlens_k = torch.arange(
|
| 212 |
0,
|
| 213 |
(batch_size + 1) * seqlen_k,
|
|
@@ -215,15 +103,7 @@ class FlashSelfAttention(torch.nn.Module):
|
|
| 215 |
dtype=torch.int32,
|
| 216 |
device=q.device,
|
| 217 |
)
|
| 218 |
-
|
| 219 |
-
if self.training:
|
| 220 |
-
assert seqlen_k == seqlen_q
|
| 221 |
-
is_causal = self.causal
|
| 222 |
-
dropout_p = self.dropout_p
|
| 223 |
-
else:
|
| 224 |
-
is_causal = seqlen_q == seqlen_k
|
| 225 |
-
dropout_p = 0
|
| 226 |
-
|
| 227 |
output = flash_attn_unpadded_func(
|
| 228 |
q,
|
| 229 |
k,
|
|
@@ -232,23 +112,30 @@ class FlashSelfAttention(torch.nn.Module):
|
|
| 232 |
cu_seqlens_k,
|
| 233 |
seqlen_q,
|
| 234 |
seqlen_k,
|
| 235 |
-
dropout_p,
|
| 236 |
softmax_scale=self.softmax_scale,
|
| 237 |
causal=is_causal,
|
| 238 |
)
|
| 239 |
-
|
| 240 |
-
|
| 241 |
-
else:
|
| 242 |
-
new_shape = (batch_size, output.shape[0] // batch_size) + output.shape[1:]
|
| 243 |
-
output = output.view(new_shape)
|
| 244 |
return output
|
| 245 |
|
| 246 |
|
| 247 |
class QWenAttention(nn.Module):
|
| 248 |
-
def __init__(self, config):
|
| 249 |
super().__init__()
|
| 250 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 251 |
self.register_buffer("masked_bias", torch.tensor(-1e4), persistent=False)
|
|
|
|
|
|
|
| 252 |
self.seq_length = config.seq_length
|
| 253 |
|
| 254 |
self.hidden_size = config.hidden_size
|
|
@@ -259,6 +146,8 @@ class QWenAttention(nn.Module):
|
|
| 259 |
self.use_flash_attn = config.use_flash_attn
|
| 260 |
self.scale_attn_weights = True
|
| 261 |
|
|
|
|
|
|
|
| 262 |
self.projection_size = config.kv_channels * config.num_attention_heads
|
| 263 |
|
| 264 |
assert self.projection_size % config.num_attention_heads == 0
|
|
@@ -279,10 +168,25 @@ class QWenAttention(nn.Module):
|
|
| 279 |
and not self.is_fp32
|
| 280 |
):
|
| 281 |
self.core_attention_flash = FlashSelfAttention(
|
| 282 |
-
causal=True, attention_dropout=config.
|
| 283 |
)
|
|
|
|
| 284 |
self.bf16 = config.bf16
|
| 285 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 286 |
self.use_dynamic_ntk = config.use_dynamic_ntk
|
| 287 |
self.use_logn_attn = config.use_logn_attn
|
| 288 |
|
|
@@ -290,104 +194,100 @@ class QWenAttention(nn.Module):
|
|
| 290 |
math.log(i, self.seq_length) if i > self.seq_length else 1
|
| 291 |
for i in range(1, 32768)
|
| 292 |
]
|
| 293 |
-
logn_tensor = torch.
|
| 294 |
-
self.
|
| 295 |
-
|
| 296 |
-
self.attn_dropout = nn.Dropout(config.
|
| 297 |
-
|
| 298 |
-
|
| 299 |
-
|
| 300 |
-
cache_dtype = torch.float
|
| 301 |
-
if self.bf16:
|
| 302 |
-
cache_dtype=torch.bfloat16
|
| 303 |
-
elif config.fp16:
|
| 304 |
-
cache_dtype = torch.float16
|
| 305 |
-
self.cache_qmax = torch.tensor(torch.iinfo(torch.uint8).max, dtype=cache_dtype)
|
| 306 |
-
self.cache_qmin = torch.tensor(torch.iinfo(torch.uint8).min, dtype=cache_dtype)
|
| 307 |
-
|
| 308 |
-
if config.use_cache_quantization and config.use_cache_kernel:
|
| 309 |
-
# pre check if the support files existing
|
| 310 |
-
module_root = pathlib.Path(__file__).parent
|
| 311 |
-
src_files = ("cache_autogptq_cuda_256.cpp", "cache_autogptq_cuda_kernel_256.cu")
|
| 312 |
-
if any(not (module_root/src).is_file() for src in src_files):
|
| 313 |
-
warnings.warn("KV cache kernel source files (.cpp and .cu) not found.")
|
| 314 |
-
self.cache_kernels = None
|
| 315 |
-
else:
|
| 316 |
-
try:
|
| 317 |
-
from .cpp_kernels import cache_autogptq_cuda_256
|
| 318 |
-
self.cache_kernels = cache_autogptq_cuda_256
|
| 319 |
-
except ImportError:
|
| 320 |
-
warnings.warn("Failed to import KV cache kernels.")
|
| 321 |
-
self.cache_kernels = None
|
| 322 |
-
|
| 323 |
-
def _attn(self, query, key, value, causal_mask=None, attention_mask=None, head_mask=None):
|
| 324 |
-
device = query.device
|
| 325 |
-
if self.use_cache_quantization:
|
| 326 |
-
qk, qk_scale, qk_zero = key
|
| 327 |
-
if self.use_cache_kernel and self.cache_kernels is not None:
|
| 328 |
-
shape = query.shape[:-1] + (qk.shape[-2],)
|
| 329 |
-
attn_weights = torch.zeros(shape, dtype=torch.float16, device=device)
|
| 330 |
-
self.cache_kernels.vecquant8matmul_batched_faster_old(
|
| 331 |
-
query.contiguous() if query.dtype == torch.float16 else query.to(torch.float16).contiguous(),
|
| 332 |
-
qk.transpose(-1, -2).contiguous(),
|
| 333 |
-
attn_weights,
|
| 334 |
-
qk_scale.contiguous() if qk_scale.dtype == torch.float16 else qk_scale.to(torch.float16).contiguous(),
|
| 335 |
-
qk_zero.contiguous()if qk_zero.dtype == torch.float16 else qk_zero.to(torch.float16).contiguous())
|
| 336 |
-
# attn_weights = attn_weights.to(query.dtype).contiguous()
|
| 337 |
-
else:
|
| 338 |
-
key = dequantize_cache_torch(qk, qk_scale, qk_zero)
|
| 339 |
-
attn_weights = torch.matmul(query, key.transpose(-1, -2))
|
| 340 |
-
else:
|
| 341 |
-
attn_weights = torch.matmul(query, key.transpose(-1, -2))
|
| 342 |
|
| 343 |
if self.scale_attn_weights:
|
| 344 |
-
|
| 345 |
-
|
| 346 |
-
|
| 347 |
-
|
| 348 |
-
|
|
|
|
| 349 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 350 |
mask_value = torch.finfo(attn_weights.dtype).min
|
| 351 |
-
|
| 352 |
-
attn_weights
|
| 353 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 354 |
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 355 |
|
| 356 |
if attention_mask is not None:
|
| 357 |
attn_weights = attn_weights + attention_mask
|
| 358 |
|
| 359 |
-
|
| 360 |
-
attn_weights = nn.functional.softmax(attn_weights.float(), dim=-1)
|
| 361 |
-
else:
|
| 362 |
-
attn_weights = nn.functional.softmax(attn_weights, dim=-1)
|
| 363 |
|
| 364 |
-
attn_weights
|
|
|
|
|
|
|
|
|
|
|
|
|
| 365 |
attn_weights = self.attn_dropout(attn_weights)
|
| 366 |
|
| 367 |
if head_mask is not None:
|
| 368 |
attn_weights = attn_weights * head_mask
|
| 369 |
|
| 370 |
-
|
| 371 |
-
qv, qv_scale, qv_zero = value
|
| 372 |
-
if self.use_cache_kernel and self.cache_kernels is not None:
|
| 373 |
-
shape = attn_weights.shape[:-1] + (query.shape[-1],)
|
| 374 |
-
attn_output = torch.zeros(shape, dtype=torch.float16, device=device)
|
| 375 |
-
self.cache_kernels.vecquant8matmul_batched_column_compression_faster_old(
|
| 376 |
-
attn_weights.contiguous() if attn_weights.dtype == torch.float16 else attn_weights.to(torch.float16).contiguous(),
|
| 377 |
-
qv.contiguous(), # dtype: int32
|
| 378 |
-
attn_output,
|
| 379 |
-
qv_scale.contiguous() if qv_scale.dtype == torch.float16 else qv_scale.to(torch.float16).contiguous(),
|
| 380 |
-
qv_zero.contiguous() if qv_zero.dtype == torch.float16 else qv_zero.to(torch.float16).contiguous())
|
| 381 |
-
if attn_output.dtype != query.dtype:
|
| 382 |
-
attn_output = attn_output.to(query.dtype)
|
| 383 |
-
attn_weights = attn_weights.to(query.dtype)
|
| 384 |
-
else:
|
| 385 |
-
value = dequantize_cache_torch(qv, qv_scale, qv_zero)
|
| 386 |
-
attn_output = torch.matmul(attn_weights, value)
|
| 387 |
-
else:
|
| 388 |
-
attn_output = torch.matmul(attn_weights, value)
|
| 389 |
-
|
| 390 |
-
attn_output = attn_output.transpose(1, 2)
|
| 391 |
|
| 392 |
return attn_output, attn_weights
|
| 393 |
|
|
@@ -404,7 +304,6 @@ class QWenAttention(nn.Module):
|
|
| 404 |
def forward(
|
| 405 |
self,
|
| 406 |
hidden_states: Optional[Tuple[torch.FloatTensor]],
|
| 407 |
-
rotary_pos_emb_list: Optional[List[List[torch.Tensor]]] = None,
|
| 408 |
layer_past: Optional[Tuple[torch.Tensor]] = None,
|
| 409 |
attention_mask: Optional[torch.FloatTensor] = None,
|
| 410 |
head_mask: Optional[torch.FloatTensor] = None,
|
|
@@ -413,80 +312,64 @@ class QWenAttention(nn.Module):
|
|
| 413 |
output_attentions: Optional[bool] = False,
|
| 414 |
use_cache: Optional[bool] = False,
|
| 415 |
):
|
| 416 |
-
mixed_x_layer = self.c_attn(hidden_states)
|
| 417 |
|
|
|
|
| 418 |
query, key, value = mixed_x_layer.split(self.split_size, dim=2)
|
| 419 |
|
| 420 |
query = self._split_heads(query, self.num_heads, self.head_dim)
|
| 421 |
key = self._split_heads(key, self.num_heads, self.head_dim)
|
| 422 |
value = self._split_heads(value, self.num_heads, self.head_dim)
|
| 423 |
|
| 424 |
-
|
| 425 |
-
|
| 426 |
-
|
| 427 |
-
|
| 428 |
-
|
| 429 |
-
|
| 430 |
-
|
| 431 |
-
|
| 432 |
-
|
| 433 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 434 |
else:
|
| 435 |
-
|
| 436 |
-
key_list = []
|
| 437 |
-
for i, rotary_pos_emb in enumerate(rotary_pos_emb_list):
|
| 438 |
-
rotary_pos_emb = [i[:, -cur_len:, :, :] for i in rotary_pos_emb]
|
| 439 |
-
rotary_pos_emb = (rotary_pos_emb,) * 2
|
| 440 |
-
q_pos_emb, k_pos_emb = rotary_pos_emb
|
| 441 |
-
# Slice the pos emb for current inference
|
| 442 |
-
query_list += [apply_rotary_pos_emb(query[i:i+1, :, :], q_pos_emb)]
|
| 443 |
-
key_list += [apply_rotary_pos_emb(key[i:i+1, :, :], k_pos_emb)]
|
| 444 |
-
query = torch.cat(query_list, dim=0)
|
| 445 |
-
key = torch.cat(key_list, dim=0)
|
| 446 |
-
|
| 447 |
-
if self.use_cache_quantization:
|
| 448 |
-
key = quantize_cache_v(key.permute(0, 2, 1, 3),
|
| 449 |
-
bits=8,
|
| 450 |
-
qmin=self.cache_qmin,
|
| 451 |
-
qmax=self.cache_qmax)
|
| 452 |
-
value = quantize_cache_v(value.permute(0, 2, 1, 3),
|
| 453 |
-
bits=8,
|
| 454 |
-
qmin=self.cache_qmin,
|
| 455 |
-
qmax=self.cache_qmax)
|
| 456 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 457 |
|
| 458 |
if layer_past is not None:
|
| 459 |
past_key, past_value = layer_past[0], layer_past[1]
|
| 460 |
-
|
| 461 |
-
|
| 462 |
-
# present=((q_key,key_scale,key_zero_point),
|
| 463 |
-
# (q_value,value_scale,value_zero_point))
|
| 464 |
-
key = (torch.cat((past_key[0], key[0]), dim=2),
|
| 465 |
-
torch.cat((past_key[1], key[1]), dim=2),
|
| 466 |
-
torch.cat((past_key[2], key[2]), dim=2))
|
| 467 |
-
value = (torch.cat((past_value[0], value[0]), dim=2),
|
| 468 |
-
torch.cat((past_value[1], value[1]), dim=2),
|
| 469 |
-
torch.cat((past_value[2], value[2]), dim=2))
|
| 470 |
-
else:
|
| 471 |
-
# not use_cache_quantization:
|
| 472 |
-
# present=(key,value)
|
| 473 |
-
key = torch.cat((past_key, key), dim=1)
|
| 474 |
-
value = torch.cat((past_value, value), dim=1)
|
| 475 |
|
| 476 |
if use_cache:
|
| 477 |
present = (key, value)
|
| 478 |
else:
|
| 479 |
present = None
|
| 480 |
|
| 481 |
-
|
| 482 |
-
|
| 483 |
-
|
| 484 |
-
|
| 485 |
-
|
| 486 |
-
|
| 487 |
-
seq_start = key.size(1) - query.size(1)
|
| 488 |
-
seq_end = key.size(1)
|
| 489 |
-
logn_tensor = self.logn_tensor[:, seq_start:seq_end, :, :].type_as(query)
|
| 490 |
query = query * logn_tensor.expand_as(query)
|
| 491 |
|
| 492 |
if (
|
|
@@ -496,49 +379,23 @@ class QWenAttention(nn.Module):
|
|
| 496 |
and query.is_cuda
|
| 497 |
):
|
| 498 |
q, k, v = query, key, value
|
| 499 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 500 |
else:
|
| 501 |
-
key_size = key[0].size(2) if self.use_cache_quantization else key.size(1)
|
| 502 |
-
if query.size(1) == key_size:
|
| 503 |
-
causal_mask = torch.tril(
|
| 504 |
-
torch.ones((key_size, key_size), dtype=torch.bool, device=query.device)
|
| 505 |
-
).view(1, 1, key_size, key_size)
|
| 506 |
-
else:
|
| 507 |
-
causal_mask = None
|
| 508 |
query = query.permute(0, 2, 1, 3)
|
| 509 |
-
|
| 510 |
-
|
| 511 |
-
|
| 512 |
-
|
| 513 |
-
|
| 514 |
-
|
| 515 |
-
|
| 516 |
-
|
| 517 |
-
and not query.is_cuda
|
| 518 |
-
):
|
| 519 |
-
raise Exception(_ERROR_INPUT_CPU_QUERY_WITH_FLASH_ATTN_ACTIVATED)
|
| 520 |
-
|
| 521 |
-
if not self.use_cache_quantization and SUPPORT_TORCH2:
|
| 522 |
-
if attention_mask is not None:
|
| 523 |
-
attention_mask = attention_mask.expand(-1, -1, query.size(2), -1)
|
| 524 |
-
if causal_mask is not None:
|
| 525 |
-
attention_mask = attention_mask.masked_fill(~causal_mask, torch.finfo(query.dtype).min)
|
| 526 |
-
else:
|
| 527 |
-
attention_mask = causal_mask
|
| 528 |
-
attn_output = F.scaled_dot_product_attention(
|
| 529 |
-
query, key, value, attn_mask=attention_mask
|
| 530 |
-
).transpose(1, 2)
|
| 531 |
-
attn_weight = None
|
| 532 |
-
else:
|
| 533 |
-
attn_output, attn_weight = self._attn(
|
| 534 |
-
query, key, value, causal_mask, attention_mask, head_mask
|
| 535 |
-
)
|
| 536 |
-
context_layer = self._merge_heads(
|
| 537 |
-
attn_output, self.num_heads, self.head_dim
|
| 538 |
-
)
|
| 539 |
|
| 540 |
attn_output = self.c_proj(context_layer)
|
| 541 |
-
|
| 542 |
outputs = (attn_output, present)
|
| 543 |
if output_attentions:
|
| 544 |
if (
|
|
@@ -547,8 +404,6 @@ class QWenAttention(nn.Module):
|
|
| 547 |
and not self.is_fp32
|
| 548 |
):
|
| 549 |
raise ValueError("Cannot output attentions while using flash-attn")
|
| 550 |
-
elif not self.use_cache_quantization and SUPPORT_TORCH2:
|
| 551 |
-
raise ValueError("Cannot output attentions while using scaled_dot_product_attention")
|
| 552 |
else:
|
| 553 |
outputs += (attn_weight,)
|
| 554 |
|
|
@@ -559,12 +414,12 @@ class QWenMLP(nn.Module):
|
|
| 559 |
def __init__(self, config):
|
| 560 |
super().__init__()
|
| 561 |
self.w1 = nn.Linear(
|
| 562 |
-
config.hidden_size, config.
|
| 563 |
)
|
| 564 |
self.w2 = nn.Linear(
|
| 565 |
-
config.hidden_size, config.
|
| 566 |
)
|
| 567 |
-
ff_dim_in = config.
|
| 568 |
self.c_proj = nn.Linear(ff_dim_in, config.hidden_size, bias=not config.no_bias)
|
| 569 |
|
| 570 |
def forward(self, hidden_states):
|
|
@@ -576,16 +431,24 @@ class QWenMLP(nn.Module):
|
|
| 576 |
|
| 577 |
|
| 578 |
class QWenBlock(nn.Module):
|
| 579 |
-
def __init__(self, config):
|
| 580 |
super().__init__()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 581 |
hidden_size = config.hidden_size
|
|
|
|
|
|
|
|
|
|
| 582 |
self.bf16 = config.bf16
|
| 583 |
|
| 584 |
self.ln_1 = RMSNorm(
|
| 585 |
hidden_size,
|
| 586 |
eps=config.layer_norm_epsilon,
|
| 587 |
)
|
| 588 |
-
self.attn = QWenAttention(config)
|
| 589 |
self.ln_2 = RMSNorm(
|
| 590 |
hidden_size,
|
| 591 |
eps=config.layer_norm_epsilon,
|
|
@@ -596,7 +459,6 @@ class QWenBlock(nn.Module):
|
|
| 596 |
def forward(
|
| 597 |
self,
|
| 598 |
hidden_states: Optional[Tuple[torch.FloatTensor]],
|
| 599 |
-
rotary_pos_emb_list: Optional[List[List[torch.Tensor]]] = None,
|
| 600 |
layer_past: Optional[Tuple[torch.Tensor]] = None,
|
| 601 |
attention_mask: Optional[torch.FloatTensor] = None,
|
| 602 |
head_mask: Optional[torch.FloatTensor] = None,
|
|
@@ -609,7 +471,6 @@ class QWenBlock(nn.Module):
|
|
| 609 |
|
| 610 |
attn_outputs = self.attn(
|
| 611 |
layernorm_output,
|
| 612 |
-
rotary_pos_emb_list,
|
| 613 |
layer_past=layer_past,
|
| 614 |
attention_mask=attention_mask,
|
| 615 |
head_mask=head_mask,
|
|
@@ -620,12 +481,19 @@ class QWenBlock(nn.Module):
|
|
| 620 |
|
| 621 |
outputs = attn_outputs[1:]
|
| 622 |
|
| 623 |
-
|
|
|
|
|
|
|
|
|
|
| 624 |
layernorm_input = attn_output + residual
|
| 625 |
|
| 626 |
layernorm_output = self.ln_2(layernorm_input)
|
| 627 |
|
| 628 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 629 |
mlp_output = self.mlp(layernorm_output)
|
| 630 |
hidden_states = residual + mlp_output
|
| 631 |
|
|
@@ -643,7 +511,6 @@ class QWenPreTrainedModel(PreTrainedModel):
|
|
| 643 |
is_parallelizable = False
|
| 644 |
supports_gradient_checkpointing = True
|
| 645 |
_no_split_modules = ["QWenBlock"]
|
| 646 |
-
_skip_keys_device_placement = "past_key_values"
|
| 647 |
|
| 648 |
def __init__(self, *inputs, **kwargs):
|
| 649 |
super().__init__(*inputs, **kwargs)
|
|
@@ -667,7 +534,7 @@ class QWenPreTrainedModel(PreTrainedModel):
|
|
| 667 |
mean=0.0,
|
| 668 |
std=(
|
| 669 |
self.config.initializer_range
|
| 670 |
-
/ math.sqrt(2 * self.config.
|
| 671 |
),
|
| 672 |
)
|
| 673 |
|
|
@@ -681,40 +548,31 @@ class QWenModel(QWenPreTrainedModel):
|
|
| 681 |
|
| 682 |
def __init__(self, config):
|
| 683 |
super().__init__(config)
|
| 684 |
-
self.vocab_size = config.
|
| 685 |
self.num_hidden_layers = config.num_hidden_layers
|
| 686 |
self.embed_dim = config.hidden_size
|
| 687 |
-
self.use_cache_quantization = self.config.use_cache_quantization if hasattr(self.config, 'use_cache_quantization') else False
|
| 688 |
|
|
|
|
|
|
|
| 689 |
self.gradient_checkpointing = False
|
| 690 |
-
self.use_dynamic_ntk = config.use_dynamic_ntk
|
| 691 |
-
self.seq_length = config.seq_length
|
| 692 |
-
|
| 693 |
-
self.wte = nn.Embedding(self.vocab_size, self.embed_dim)
|
| 694 |
-
|
| 695 |
-
self.drop = nn.Dropout(config.emb_dropout_prob)
|
| 696 |
|
| 697 |
-
if
|
| 698 |
-
self.
|
|
|
|
|
|
|
|
|
|
| 699 |
else:
|
| 700 |
-
|
| 701 |
-
self.
|
| 702 |
-
config.kv_channels * config.rotary_pct
|
| 703 |
-
)
|
| 704 |
-
dim = (
|
| 705 |
-
self.rotary_ndims
|
| 706 |
-
if self.rotary_ndims is not None
|
| 707 |
-
else config.kv_channels
|
| 708 |
-
)
|
| 709 |
-
self.rotary_emb = RotaryEmbedding(dim, base=config.rotary_emb_base)
|
| 710 |
|
| 711 |
-
self.
|
| 712 |
-
self.is_fp32 = not (config.bf16 or config.fp16)
|
| 713 |
|
|
|
|
| 714 |
self.h = nn.ModuleList(
|
| 715 |
[
|
| 716 |
QWenBlock(
|
| 717 |
-
config
|
|
|
|
| 718 |
)
|
| 719 |
for i in range(config.num_hidden_layers)
|
| 720 |
]
|
|
@@ -732,12 +590,6 @@ class QWenModel(QWenPreTrainedModel):
|
|
| 732 |
def set_input_embeddings(self, new_embeddings):
|
| 733 |
self.wte = new_embeddings
|
| 734 |
|
| 735 |
-
def get_ntk_alpha(self, true_seq_len):
|
| 736 |
-
context_value = math.log(true_seq_len / self.seq_length, 2) + 1
|
| 737 |
-
ntk_alpha = 2 ** math.ceil(context_value) - 1
|
| 738 |
-
ntk_alpha = max(ntk_alpha, 1)
|
| 739 |
-
return ntk_alpha
|
| 740 |
-
|
| 741 |
def forward(
|
| 742 |
self,
|
| 743 |
input_ids: Optional[torch.LongTensor] = None,
|
|
@@ -794,10 +646,8 @@ class QWenModel(QWenPreTrainedModel):
|
|
| 794 |
past_length = 0
|
| 795 |
past_key_values = tuple([None] * len(self.h))
|
| 796 |
else:
|
| 797 |
-
|
| 798 |
-
|
| 799 |
-
else:
|
| 800 |
-
past_length = past_key_values[0][0].size(-2)
|
| 801 |
if position_ids is None:
|
| 802 |
position_ids = torch.arange(
|
| 803 |
past_length,
|
|
@@ -816,39 +666,14 @@ class QWenModel(QWenPreTrainedModel):
|
|
| 816 |
attention_mask = (1.0 - attention_mask) * torch.finfo(self.dtype).min
|
| 817 |
|
| 818 |
encoder_attention_mask = None
|
| 819 |
-
head_mask = self.get_head_mask(head_mask, self.config.
|
| 820 |
|
| 821 |
if inputs_embeds is None:
|
| 822 |
inputs_embeds = self.wte(input_ids)
|
| 823 |
hidden_states = inputs_embeds
|
| 824 |
-
|
| 825 |
-
|
| 826 |
-
|
| 827 |
-
# past key values[0][0] shape: bs * seq_len * head_num * dim
|
| 828 |
-
if self.use_cache_quantization:
|
| 829 |
-
kv_seq_len += past_key_values[0][0][0].shape[2]
|
| 830 |
-
else:
|
| 831 |
-
kv_seq_len += past_key_values[0][0].shape[1]
|
| 832 |
-
|
| 833 |
-
if self.training or not self.use_dynamic_ntk:
|
| 834 |
-
ntk_alpha_list = [1.0]
|
| 835 |
-
elif kv_seq_len != hidden_states.size()[1]:
|
| 836 |
-
ntk_alpha_list = self.rotary_emb._ntk_alpha_cached_list
|
| 837 |
-
else:
|
| 838 |
-
ntk_alpha_list = []
|
| 839 |
-
if attention_mask is not None and kv_seq_len > self.seq_length:
|
| 840 |
-
true_seq_lens = attention_mask.squeeze(1).squeeze(1).eq(0).sum(dim=-1, dtype=torch.int32)
|
| 841 |
-
for i in range(hidden_states.size()[0]):
|
| 842 |
-
true_seq_len = true_seq_lens[i].item()
|
| 843 |
-
ntk_alpha = self.get_ntk_alpha(true_seq_len)
|
| 844 |
-
ntk_alpha_list.append(ntk_alpha)
|
| 845 |
-
else:
|
| 846 |
-
ntk_alpha = self.get_ntk_alpha(kv_seq_len)
|
| 847 |
-
ntk_alpha_list.append(ntk_alpha)
|
| 848 |
-
self.rotary_emb._ntk_alpha_cached_list = ntk_alpha_list
|
| 849 |
-
rotary_pos_emb_list = [
|
| 850 |
-
self.rotary_emb(kv_seq_len, ntk_alpha=ntk_alpha) for ntk_alpha in ntk_alpha_list
|
| 851 |
-
]
|
| 852 |
|
| 853 |
hidden_states = self.drop(hidden_states)
|
| 854 |
output_shape = input_shape + (hidden_states.size(-1),)
|
|
@@ -880,7 +705,6 @@ class QWenModel(QWenPreTrainedModel):
|
|
| 880 |
outputs = torch.utils.checkpoint.checkpoint(
|
| 881 |
create_custom_forward(block),
|
| 882 |
hidden_states,
|
| 883 |
-
rotary_pos_emb_list,
|
| 884 |
None,
|
| 885 |
attention_mask,
|
| 886 |
head_mask[i],
|
|
@@ -891,7 +715,6 @@ class QWenModel(QWenPreTrainedModel):
|
|
| 891 |
outputs = block(
|
| 892 |
hidden_states,
|
| 893 |
layer_past=layer_past,
|
| 894 |
-
rotary_pos_emb_list=rotary_pos_emb_list,
|
| 895 |
attention_mask=attention_mask,
|
| 896 |
head_mask=head_mask[i],
|
| 897 |
encoder_hidden_states=encoder_hidden_states,
|
|
@@ -902,16 +725,13 @@ class QWenModel(QWenPreTrainedModel):
|
|
| 902 |
|
| 903 |
hidden_states = outputs[0]
|
| 904 |
if use_cache is True:
|
| 905 |
-
presents = presents + (outputs[1],)
|
| 906 |
|
| 907 |
if output_attentions:
|
| 908 |
-
all_self_attentions = all_self_attentions + (outputs[
|
| 909 |
|
| 910 |
hidden_states = self.ln_f(hidden_states)
|
| 911 |
hidden_states = hidden_states.view(output_shape)
|
| 912 |
-
# Add last hidden state
|
| 913 |
-
if output_hidden_states:
|
| 914 |
-
all_hidden_states = all_hidden_states + (hidden_states,)
|
| 915 |
|
| 916 |
if not return_dict:
|
| 917 |
return tuple(
|
|
@@ -963,7 +783,7 @@ class QWenLMHeadModel(QWenPreTrainedModel):
|
|
| 963 |
logger.warn("Your device support faster inference by passing bf16=True in \"AutoModelForCausalLM.from_pretrained\".")
|
| 964 |
elif SUPPORT_FP16:
|
| 965 |
logger.warn("Your device support faster inference by passing fp16=True in \"AutoModelForCausalLM.from_pretrained\".")
|
| 966 |
-
|
| 967 |
if config.use_flash_attn == "auto":
|
| 968 |
if config.bf16 or config.fp16:
|
| 969 |
logger.warn("Try importing flash-attention for faster inference...")
|
|
@@ -974,10 +794,36 @@ class QWenLMHeadModel(QWenPreTrainedModel):
|
|
| 974 |
logger.warn("Flash attention will be disabled because it does NOT support fp32.")
|
| 975 |
|
| 976 |
if config.use_flash_attn:
|
| 977 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 978 |
|
| 979 |
self.transformer = QWenModel(config)
|
| 980 |
-
self.lm_head = nn.Linear(config.
|
| 981 |
|
| 982 |
if config.bf16:
|
| 983 |
self.transformer.bfloat16()
|
|
@@ -996,13 +842,22 @@ class QWenLMHeadModel(QWenPreTrainedModel):
|
|
| 996 |
def prepare_inputs_for_generation(
|
| 997 |
self, input_ids, past_key_values=None, inputs_embeds=None, **kwargs
|
| 998 |
):
|
|
|
|
| 999 |
if past_key_values:
|
| 1000 |
input_ids = input_ids[:, -1].unsqueeze(-1)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1001 |
|
| 1002 |
-
if
|
| 1003 |
-
|
|
|
|
|
|
|
|
|
|
| 1004 |
else:
|
| 1005 |
-
|
| 1006 |
|
| 1007 |
if inputs_embeds is not None and past_key_values is None:
|
| 1008 |
model_inputs = {"inputs_embeds": inputs_embeds}
|
|
@@ -1013,7 +868,9 @@ class QWenLMHeadModel(QWenPreTrainedModel):
|
|
| 1013 |
{
|
| 1014 |
"past_key_values": past_key_values,
|
| 1015 |
"use_cache": kwargs.get("use_cache"),
|
|
|
|
| 1016 |
"attention_mask": attention_mask,
|
|
|
|
| 1017 |
}
|
| 1018 |
)
|
| 1019 |
return model_inputs
|
|
@@ -1100,129 +957,67 @@ class QWenLMHeadModel(QWenPreTrainedModel):
|
|
| 1100 |
query: str,
|
| 1101 |
history: Optional[HistoryType],
|
| 1102 |
system: str = "You are a helpful assistant.",
|
| 1103 |
-
|
|
|
|
| 1104 |
stop_words_ids: Optional[List[List[int]]] = None,
|
| 1105 |
-
generation_config: Optional[GenerationConfig] = None,
|
| 1106 |
**kwargs,
|
| 1107 |
) -> Tuple[str, HistoryType]:
|
| 1108 |
-
generation_config = generation_config if generation_config is not None else self.generation_config
|
| 1109 |
-
|
| 1110 |
-
assert stream is _SENTINEL, _ERROR_STREAM_IN_CHAT
|
| 1111 |
-
assert generation_config.chat_format == 'chatml', _ERROR_BAD_CHAT_FORMAT
|
| 1112 |
if history is None:
|
| 1113 |
history = []
|
| 1114 |
-
else:
|
| 1115 |
-
# make a copy of the user's input such that is is left untouched
|
| 1116 |
-
history = copy.deepcopy(history)
|
| 1117 |
-
|
| 1118 |
if stop_words_ids is None:
|
| 1119 |
stop_words_ids = []
|
| 1120 |
|
| 1121 |
-
max_window_size = kwargs.get('max_window_size', None)
|
| 1122 |
-
if max_window_size is None:
|
| 1123 |
-
max_window_size = generation_config.max_window_size
|
| 1124 |
raw_text, context_tokens = make_context(
|
| 1125 |
tokenizer,
|
| 1126 |
query,
|
| 1127 |
history=history,
|
| 1128 |
system=system,
|
| 1129 |
-
max_window_size=
|
| 1130 |
-
chat_format=generation_config.chat_format,
|
| 1131 |
)
|
| 1132 |
|
| 1133 |
stop_words_ids.extend(get_stop_words_ids(
|
| 1134 |
-
generation_config.chat_format, tokenizer
|
| 1135 |
))
|
| 1136 |
input_ids = torch.tensor([context_tokens]).to(self.device)
|
| 1137 |
-
|
| 1138 |
-
|
| 1139 |
-
|
| 1140 |
-
|
| 1141 |
-
|
| 1142 |
-
|
| 1143 |
-
|
| 1144 |
-
|
| 1145 |
-
|
| 1146 |
-
|
| 1147 |
-
|
| 1148 |
-
|
| 1149 |
-
|
| 1150 |
-
|
| 1151 |
-
|
| 1152 |
-
|
| 1153 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1154 |
|
| 1155 |
-
|
| 1156 |
-
|
| 1157 |
-
# separating input history and output history also enables the user
|
| 1158 |
-
# to implement more complex history management
|
| 1159 |
-
history.append((query, response))
|
| 1160 |
|
| 1161 |
return response, history
|
| 1162 |
|
| 1163 |
-
def chat_stream(
|
| 1164 |
-
self,
|
| 1165 |
-
tokenizer: PreTrainedTokenizer,
|
| 1166 |
-
query: str,
|
| 1167 |
-
history: Optional[HistoryType],
|
| 1168 |
-
system: str = "You are a helpful assistant.",
|
| 1169 |
-
stop_words_ids: Optional[List[List[int]]] = None,
|
| 1170 |
-
logits_processor: Optional[LogitsProcessorList] = None,
|
| 1171 |
-
generation_config: Optional[GenerationConfig] = None,
|
| 1172 |
-
**kwargs,
|
| 1173 |
-
) -> Generator[str, Any, None]:
|
| 1174 |
-
generation_config = generation_config if generation_config is not None else self.generation_config
|
| 1175 |
-
assert generation_config.chat_format == 'chatml', _ERROR_BAD_CHAT_FORMAT
|
| 1176 |
-
if history is None:
|
| 1177 |
-
history = []
|
| 1178 |
-
if stop_words_ids is None:
|
| 1179 |
-
stop_words_ids = []
|
| 1180 |
-
|
| 1181 |
-
max_window_size = kwargs.get('max_window_size', None)
|
| 1182 |
-
if max_window_size is None:
|
| 1183 |
-
max_window_size = generation_config.max_window_size
|
| 1184 |
-
raw_text, context_tokens = make_context(
|
| 1185 |
-
tokenizer,
|
| 1186 |
-
query,
|
| 1187 |
-
history=history,
|
| 1188 |
-
system=system,
|
| 1189 |
-
max_window_size=max_window_size,
|
| 1190 |
-
chat_format=generation_config.chat_format,
|
| 1191 |
-
)
|
| 1192 |
-
|
| 1193 |
-
stop_words_ids.extend(get_stop_words_ids(
|
| 1194 |
-
generation_config.chat_format, tokenizer
|
| 1195 |
-
))
|
| 1196 |
-
if stop_words_ids is not None:
|
| 1197 |
-
stop_words_logits_processor = StopWordsLogitsProcessor(
|
| 1198 |
-
stop_words_ids=stop_words_ids,
|
| 1199 |
-
eos_token_id=generation_config.eos_token_id,
|
| 1200 |
-
)
|
| 1201 |
-
if logits_processor is None:
|
| 1202 |
-
logits_processor = LogitsProcessorList([stop_words_logits_processor])
|
| 1203 |
-
else:
|
| 1204 |
-
logits_processor.append(stop_words_logits_processor)
|
| 1205 |
-
input_ids = torch.tensor([context_tokens]).to(self.device)
|
| 1206 |
-
|
| 1207 |
-
from transformers_stream_generator.main import NewGenerationMixin, StreamGenerationConfig
|
| 1208 |
-
self.__class__.generate_stream = NewGenerationMixin.generate
|
| 1209 |
-
self.__class__.sample_stream = NewGenerationMixin.sample_stream
|
| 1210 |
-
stream_config = StreamGenerationConfig(**generation_config.to_dict(), do_stream=True)
|
| 1211 |
-
|
| 1212 |
-
def stream_generator():
|
| 1213 |
-
outputs = []
|
| 1214 |
-
for token in self.generate_stream(
|
| 1215 |
-
input_ids,
|
| 1216 |
-
return_dict_in_generate=False,
|
| 1217 |
-
generation_config=stream_config,
|
| 1218 |
-
logits_processor=logits_processor,
|
| 1219 |
-
seed=-1,
|
| 1220 |
-
**kwargs):
|
| 1221 |
-
outputs.append(token.item())
|
| 1222 |
-
yield tokenizer.decode(outputs, skip_special_tokens=True, errors='ignore')
|
| 1223 |
-
|
| 1224 |
-
return stream_generator()
|
| 1225 |
-
|
| 1226 |
def generate(
|
| 1227 |
self,
|
| 1228 |
inputs: Optional[torch.Tensor] = None,
|
|
@@ -1233,23 +1028,20 @@ class QWenLMHeadModel(QWenPreTrainedModel):
|
|
| 1233 |
Callable[[int, torch.Tensor], List[int]]
|
| 1234 |
] = None,
|
| 1235 |
synced_gpus: Optional[bool] = None,
|
| 1236 |
-
assistant_model: Optional["PreTrainedModel"] = None,
|
| 1237 |
streamer: Optional["BaseStreamer"] = None,
|
| 1238 |
**kwargs,
|
| 1239 |
) -> Union[GenerateOutput, torch.LongTensor]:
|
| 1240 |
-
generation_config = generation_config if generation_config is not None else self.generation_config
|
| 1241 |
-
|
| 1242 |
# Process stop_words_ids.
|
| 1243 |
stop_words_ids = kwargs.pop("stop_words_ids", None)
|
| 1244 |
if stop_words_ids is None and generation_config is not None:
|
| 1245 |
stop_words_ids = getattr(generation_config, "stop_words_ids", None)
|
| 1246 |
if stop_words_ids is None:
|
| 1247 |
-
stop_words_ids = getattr(generation_config, "stop_words_ids", None)
|
| 1248 |
|
| 1249 |
if stop_words_ids is not None:
|
| 1250 |
stop_words_logits_processor = StopWordsLogitsProcessor(
|
| 1251 |
stop_words_ids=stop_words_ids,
|
| 1252 |
-
eos_token_id=generation_config.eos_token_id,
|
| 1253 |
)
|
| 1254 |
if logits_processor is None:
|
| 1255 |
logits_processor = LogitsProcessorList([stop_words_logits_processor])
|
|
@@ -1258,13 +1050,12 @@ class QWenLMHeadModel(QWenPreTrainedModel):
|
|
| 1258 |
|
| 1259 |
return super().generate(
|
| 1260 |
inputs,
|
| 1261 |
-
generation_config
|
| 1262 |
-
logits_processor
|
| 1263 |
-
stopping_criteria
|
| 1264 |
-
prefix_allowed_tokens_fn
|
| 1265 |
-
synced_gpus
|
| 1266 |
-
|
| 1267 |
-
streamer=streamer,
|
| 1268 |
**kwargs,
|
| 1269 |
)
|
| 1270 |
|
|
@@ -1274,17 +1065,16 @@ class RotaryEmbedding(torch.nn.Module):
|
|
| 1274 |
super().__init__()
|
| 1275 |
self.dim = dim
|
| 1276 |
self.base = base
|
| 1277 |
-
inv_freq = 1.0 / (base ** (torch.arange(0, dim, 2).float() / dim))
|
| 1278 |
-
self.register_buffer("inv_freq", inv_freq, persistent=False)
|
| 1279 |
if importlib.util.find_spec("einops") is None:
|
| 1280 |
raise RuntimeError("einops is required for Rotary Embedding")
|
| 1281 |
|
| 1282 |
self._rotary_pos_emb_cache = None
|
| 1283 |
self._seq_len_cached = 0
|
| 1284 |
self._ntk_alpha_cached = 1.0
|
| 1285 |
-
self._ntk_alpha_cached_list = [1.0]
|
| 1286 |
|
| 1287 |
-
def update_rotary_pos_emb_cache(self,
|
|
|
|
| 1288 |
if seqlen > self._seq_len_cached or ntk_alpha != self._ntk_alpha_cached:
|
| 1289 |
base = self.base * ntk_alpha ** (self.dim / (self.dim - 2))
|
| 1290 |
self.inv_freq = 1.0 / (
|
|
@@ -1294,23 +1084,18 @@ class RotaryEmbedding(torch.nn.Module):
|
|
| 1294 |
/ self.dim
|
| 1295 |
)
|
| 1296 |
)
|
| 1297 |
-
self._seq_len_cached =
|
| 1298 |
self._ntk_alpha_cached = ntk_alpha
|
| 1299 |
-
seq = torch.arange(
|
| 1300 |
freqs = torch.outer(seq.type_as(self.inv_freq), self.inv_freq)
|
| 1301 |
-
|
| 1302 |
emb = torch.cat((freqs, freqs), dim=-1)
|
| 1303 |
from einops import rearrange
|
| 1304 |
|
| 1305 |
-
|
| 1306 |
-
|
| 1307 |
-
cos, sin = emb.cos(), emb.sin()
|
| 1308 |
-
self._rotary_pos_emb_cache = [cos, sin]
|
| 1309 |
|
| 1310 |
-
def forward(self, max_seq_len, ntk_alpha=1.0):
|
| 1311 |
-
self.update_rotary_pos_emb_cache(max_seq_len, ntk_alpha)
|
| 1312 |
-
|
| 1313 |
-
return [cos[:, :max_seq_len], sin[:, :max_seq_len]]
|
| 1314 |
|
| 1315 |
|
| 1316 |
def _rotate_half(x):
|
|
@@ -1322,28 +1107,20 @@ def _rotate_half(x):
|
|
| 1322 |
|
| 1323 |
|
| 1324 |
def apply_rotary_pos_emb(t, freqs):
|
| 1325 |
-
|
| 1326 |
-
|
| 1327 |
-
|
| 1328 |
-
|
| 1329 |
-
|
| 1330 |
-
|
| 1331 |
-
|
| 1332 |
-
"""
|
| 1333 |
-
rot_dim = freqs[0].shape[-1]
|
| 1334 |
-
cos, sin = freqs
|
| 1335 |
-
t_float = t.float()
|
| 1336 |
-
if apply_rotary_emb_func is not None and t.is_cuda:
|
| 1337 |
-
# apply_rotary_emb in flash_attn requires cos/sin to be of
|
| 1338 |
-
# shape (seqlen, rotary_dim / 2) and apply rotary embedding
|
| 1339 |
-
# to the first rotary_dim of the input
|
| 1340 |
-
cos = cos.squeeze(0).squeeze(1)[:, : rot_dim // 2]
|
| 1341 |
-
sin = sin.squeeze(0).squeeze(1)[:, : rot_dim // 2]
|
| 1342 |
-
return apply_rotary_emb_func(t_float, cos, sin).type_as(t)
|
| 1343 |
else:
|
| 1344 |
-
|
| 1345 |
-
|
| 1346 |
-
|
|
|
|
|
|
|
|
|
|
| 1347 |
|
| 1348 |
|
| 1349 |
class RMSNorm(torch.nn.Module):
|
|
|
|
| 3 |
# This source code is licensed under the license found in the
|
| 4 |
# LICENSE file in the root directory of this source tree.
|
| 5 |
|
|
|
|
| 6 |
import importlib
|
| 7 |
import math
|
| 8 |
+
from typing import TYPE_CHECKING, Optional, Tuple, Union, Callable, List
|
|
|
|
| 9 |
|
| 10 |
import torch
|
| 11 |
import torch.nn.functional as F
|
| 12 |
import torch.utils.checkpoint
|
| 13 |
+
from torch.cuda.amp import autocast
|
| 14 |
|
| 15 |
from torch.nn import CrossEntropyLoss
|
| 16 |
from transformers import PreTrainedTokenizer, GenerationConfig, StoppingCriteriaList
|
|
|
|
| 35 |
SUPPORT_CUDA = torch.cuda.is_available()
|
| 36 |
SUPPORT_BF16 = SUPPORT_CUDA and torch.cuda.is_bf16_supported()
|
| 37 |
SUPPORT_FP16 = SUPPORT_CUDA and torch.cuda.get_device_capability(0)[0] >= 7
|
|
|
|
| 38 |
|
| 39 |
+
apply_rotary_emb_func = None
|
| 40 |
+
rms_norm = None
|
| 41 |
+
flash_attn_unpadded_func = None
|
| 42 |
|
| 43 |
from .configuration_qwen import QWenConfig
|
| 44 |
from .qwen_generation_utils import (
|
|
|
|
| 57 |
|
| 58 |
QWen_PRETRAINED_MODEL_ARCHIVE_LIST = ["qwen-7b"]
|
| 59 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 60 |
class FlashSelfAttention(torch.nn.Module):
|
| 61 |
def __init__(
|
| 62 |
self,
|
|
|
|
| 75 |
self.softmax_scale = softmax_scale
|
| 76 |
self.dropout_p = attention_dropout
|
| 77 |
|
| 78 |
+
def forward(self, q, k, v):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 79 |
assert all((i.dtype in [torch.float16, torch.bfloat16] for i in (q, k, v)))
|
| 80 |
assert all((i.is_cuda for i in (q, k, v)))
|
| 81 |
batch_size, seqlen_q = q.shape[0], q.shape[1]
|
| 82 |
seqlen_k = k.shape[1]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 83 |
q, k, v = [rearrange(x, "b s ... -> (b s) ...") for x in [q, k, v]]
|
| 84 |
cu_seqlens_q = torch.arange(
|
| 85 |
0,
|
|
|
|
| 89 |
device=q.device,
|
| 90 |
)
|
| 91 |
|
| 92 |
+
if self.training:
|
| 93 |
+
assert seqlen_k == seqlen_q
|
| 94 |
+
|
| 95 |
+
is_causal = self.causal
|
| 96 |
+
cu_seqlens_k = cu_seqlens_q
|
|
|
|
|
|
|
| 97 |
else:
|
| 98 |
+
is_causal = seqlen_q == seqlen_k
|
| 99 |
cu_seqlens_k = torch.arange(
|
| 100 |
0,
|
| 101 |
(batch_size + 1) * seqlen_k,
|
|
|
|
| 103 |
dtype=torch.int32,
|
| 104 |
device=q.device,
|
| 105 |
)
|
| 106 |
+
self.dropout_p = 0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 107 |
output = flash_attn_unpadded_func(
|
| 108 |
q,
|
| 109 |
k,
|
|
|
|
| 112 |
cu_seqlens_k,
|
| 113 |
seqlen_q,
|
| 114 |
seqlen_k,
|
| 115 |
+
self.dropout_p,
|
| 116 |
softmax_scale=self.softmax_scale,
|
| 117 |
causal=is_causal,
|
| 118 |
)
|
| 119 |
+
|
| 120 |
+
output = rearrange(output, "(b s) ... -> b s ...", b=batch_size)
|
|
|
|
|
|
|
|
|
|
| 121 |
return output
|
| 122 |
|
| 123 |
|
| 124 |
class QWenAttention(nn.Module):
|
| 125 |
+
def __init__(self, config, layer_number=None):
|
| 126 |
super().__init__()
|
| 127 |
|
| 128 |
+
max_positions = config.max_position_embeddings
|
| 129 |
+
self.register_buffer(
|
| 130 |
+
"bias",
|
| 131 |
+
torch.tril(
|
| 132 |
+
torch.ones((max_positions, max_positions), dtype=torch.bool)
|
| 133 |
+
).view(1, 1, max_positions, max_positions),
|
| 134 |
+
persistent=False,
|
| 135 |
+
)
|
| 136 |
self.register_buffer("masked_bias", torch.tensor(-1e4), persistent=False)
|
| 137 |
+
self.layer_number = max(1, layer_number)
|
| 138 |
+
self.params_dtype = config.params_dtype
|
| 139 |
self.seq_length = config.seq_length
|
| 140 |
|
| 141 |
self.hidden_size = config.hidden_size
|
|
|
|
| 146 |
self.use_flash_attn = config.use_flash_attn
|
| 147 |
self.scale_attn_weights = True
|
| 148 |
|
| 149 |
+
self.layer_idx = None
|
| 150 |
+
|
| 151 |
self.projection_size = config.kv_channels * config.num_attention_heads
|
| 152 |
|
| 153 |
assert self.projection_size % config.num_attention_heads == 0
|
|
|
|
| 168 |
and not self.is_fp32
|
| 169 |
):
|
| 170 |
self.core_attention_flash = FlashSelfAttention(
|
| 171 |
+
causal=True, attention_dropout=config.attn_pdrop
|
| 172 |
)
|
| 173 |
+
|
| 174 |
self.bf16 = config.bf16
|
| 175 |
|
| 176 |
+
if config.rotary_pct == 1.0:
|
| 177 |
+
self.rotary_ndims = None
|
| 178 |
+
else:
|
| 179 |
+
assert config.rotary_pct < 1
|
| 180 |
+
self.rotary_ndims = int(
|
| 181 |
+
self.hidden_size_per_attention_head * config.rotary_pct
|
| 182 |
+
)
|
| 183 |
+
dim = (
|
| 184 |
+
self.rotary_ndims
|
| 185 |
+
if self.rotary_ndims is not None
|
| 186 |
+
else self.hidden_size_per_attention_head
|
| 187 |
+
)
|
| 188 |
+
self.rotary_emb = RotaryEmbedding(dim, base=config.rotary_emb_base)
|
| 189 |
+
|
| 190 |
self.use_dynamic_ntk = config.use_dynamic_ntk
|
| 191 |
self.use_logn_attn = config.use_logn_attn
|
| 192 |
|
|
|
|
| 194 |
math.log(i, self.seq_length) if i > self.seq_length else 1
|
| 195 |
for i in range(1, 32768)
|
| 196 |
]
|
| 197 |
+
self.logn_tensor = torch.Tensor(logn_list)[None, :, None, None]
|
| 198 |
+
self._ntk_cached = 1.0
|
| 199 |
+
|
| 200 |
+
self.attn_dropout = nn.Dropout(config.attn_pdrop)
|
| 201 |
+
|
| 202 |
+
def _attn(self, query, key, value, attention_mask=None, head_mask=None):
|
| 203 |
+
attn_weights = torch.matmul(query, key.transpose(-1, -2))
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 204 |
|
| 205 |
if self.scale_attn_weights:
|
| 206 |
+
attn_weights = attn_weights / torch.full(
|
| 207 |
+
[],
|
| 208 |
+
value.size(-1) ** 0.5,
|
| 209 |
+
dtype=attn_weights.dtype,
|
| 210 |
+
device=attn_weights.device,
|
| 211 |
+
)
|
| 212 |
|
| 213 |
+
query_length, key_length = query.size(-2), key.size(-2)
|
| 214 |
+
causal_mask = self.bias[
|
| 215 |
+
:, :, key_length - query_length : key_length, :key_length
|
| 216 |
+
]
|
| 217 |
mask_value = torch.finfo(attn_weights.dtype).min
|
| 218 |
+
mask_value = torch.full([], mask_value, dtype=attn_weights.dtype).to(
|
| 219 |
+
attn_weights.device
|
| 220 |
+
)
|
| 221 |
+
attn_weights = torch.where(
|
| 222 |
+
causal_mask, attn_weights.to(attn_weights.dtype), mask_value
|
| 223 |
+
)
|
| 224 |
+
|
| 225 |
+
attn_weights = nn.functional.softmax(attn_weights, dim=-1)
|
| 226 |
+
|
| 227 |
+
attn_weights = attn_weights.type(value.dtype)
|
| 228 |
+
attn_weights = self.attn_dropout(attn_weights)
|
| 229 |
+
|
| 230 |
+
if head_mask is not None:
|
| 231 |
+
attn_weights = attn_weights * head_mask
|
| 232 |
+
|
| 233 |
+
attn_output = torch.matmul(attn_weights, value)
|
| 234 |
+
attn_output = attn_output.transpose(1, 2)
|
| 235 |
+
|
| 236 |
+
return attn_output, attn_weights
|
| 237 |
+
|
| 238 |
+
def _upcast_and_reordered_attn(
|
| 239 |
+
self, query, key, value, attention_mask=None, head_mask=None
|
| 240 |
+
):
|
| 241 |
+
bsz, num_heads, q_seq_len, dk = query.size()
|
| 242 |
+
_, _, k_seq_len, _ = key.size()
|
| 243 |
+
|
| 244 |
+
attn_weights = torch.empty(
|
| 245 |
+
bsz * num_heads,
|
| 246 |
+
q_seq_len,
|
| 247 |
+
k_seq_len,
|
| 248 |
+
dtype=torch.float32,
|
| 249 |
+
device=query.device,
|
| 250 |
+
)
|
| 251 |
+
|
| 252 |
+
scale_factor = 1.0
|
| 253 |
+
if self.scale_attn_weights:
|
| 254 |
+
scale_factor /= float(value.size(-1)) ** 0.5
|
| 255 |
+
|
| 256 |
+
with autocast(enabled=False):
|
| 257 |
+
q, k = query.reshape(-1, q_seq_len, dk), key.transpose(-1, -2).reshape(
|
| 258 |
+
-1, dk, k_seq_len
|
| 259 |
)
|
| 260 |
+
attn_weights = torch.baddbmm(
|
| 261 |
+
attn_weights, q.float(), k.float(), beta=0, alpha=scale_factor
|
| 262 |
+
)
|
| 263 |
+
attn_weights = attn_weights.reshape(bsz, num_heads, q_seq_len, k_seq_len)
|
| 264 |
+
|
| 265 |
+
query_length, key_length = query.size(-2), key.size(-2)
|
| 266 |
+
causal_mask = self.bias[
|
| 267 |
+
:, :, key_length - query_length : key_length, :key_length
|
| 268 |
+
]
|
| 269 |
+
mask_value = torch.finfo(attn_weights.dtype).min
|
| 270 |
+
mask_value = torch.tensor(mask_value, dtype=attn_weights.dtype).to(
|
| 271 |
+
attn_weights.device
|
| 272 |
+
)
|
| 273 |
+
attn_weights = torch.where(causal_mask, attn_weights, mask_value)
|
| 274 |
|
| 275 |
if attention_mask is not None:
|
| 276 |
attn_weights = attn_weights + attention_mask
|
| 277 |
|
| 278 |
+
attn_weights = nn.functional.softmax(attn_weights, dim=-1)
|
|
|
|
|
|
|
|
|
|
| 279 |
|
| 280 |
+
if attn_weights.dtype != torch.float32:
|
| 281 |
+
raise RuntimeError(
|
| 282 |
+
"Error with upcasting, attn_weights does not have dtype torch.float32"
|
| 283 |
+
)
|
| 284 |
+
attn_weights = attn_weights.type(value.dtype)
|
| 285 |
attn_weights = self.attn_dropout(attn_weights)
|
| 286 |
|
| 287 |
if head_mask is not None:
|
| 288 |
attn_weights = attn_weights * head_mask
|
| 289 |
|
| 290 |
+
attn_output = torch.matmul(attn_weights, value)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 291 |
|
| 292 |
return attn_output, attn_weights
|
| 293 |
|
|
|
|
| 304 |
def forward(
|
| 305 |
self,
|
| 306 |
hidden_states: Optional[Tuple[torch.FloatTensor]],
|
|
|
|
| 307 |
layer_past: Optional[Tuple[torch.Tensor]] = None,
|
| 308 |
attention_mask: Optional[torch.FloatTensor] = None,
|
| 309 |
head_mask: Optional[torch.FloatTensor] = None,
|
|
|
|
| 312 |
output_attentions: Optional[bool] = False,
|
| 313 |
use_cache: Optional[bool] = False,
|
| 314 |
):
|
|
|
|
| 315 |
|
| 316 |
+
mixed_x_layer = self.c_attn(hidden_states)
|
| 317 |
query, key, value = mixed_x_layer.split(self.split_size, dim=2)
|
| 318 |
|
| 319 |
query = self._split_heads(query, self.num_heads, self.head_dim)
|
| 320 |
key = self._split_heads(key, self.num_heads, self.head_dim)
|
| 321 |
value = self._split_heads(value, self.num_heads, self.head_dim)
|
| 322 |
|
| 323 |
+
kv_seq_len = hidden_states.size()[1]
|
| 324 |
+
if layer_past:
|
| 325 |
+
# layer past[0] shape: bs * seq_len * head_num * dim
|
| 326 |
+
kv_seq_len += layer_past[0].shape[1]
|
| 327 |
+
if (
|
| 328 |
+
self.use_dynamic_ntk
|
| 329 |
+
and kv_seq_len == hidden_states.size()[1]
|
| 330 |
+
and not self.training
|
| 331 |
+
):
|
| 332 |
+
context_value = math.log(kv_seq_len / self.seq_length, 2) + 1
|
| 333 |
+
ntk_alpha = 2 ** math.ceil(context_value) - 1
|
| 334 |
+
ntk_alpha = max(ntk_alpha, 1)
|
| 335 |
+
self._ntk_cached = ntk_alpha
|
| 336 |
+
else:
|
| 337 |
+
ntk_alpha = self._ntk_cached
|
| 338 |
+
rotary_pos_emb = self.rotary_emb(kv_seq_len, ntk_alpha=ntk_alpha).to(
|
| 339 |
+
hidden_states.device
|
| 340 |
+
)
|
| 341 |
+
|
| 342 |
+
if rotary_pos_emb is not None:
|
| 343 |
+
if isinstance(rotary_pos_emb, tuple):
|
| 344 |
+
rotary_pos_emb = rotary_pos_emb
|
| 345 |
else:
|
| 346 |
+
rotary_pos_emb = (rotary_pos_emb,) * 2
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 347 |
|
| 348 |
+
if rotary_pos_emb is not None:
|
| 349 |
+
q_pos_emb, k_pos_emb = rotary_pos_emb
|
| 350 |
+
# Slice the pos emb for current inference
|
| 351 |
+
cur_len = query.shape[1]
|
| 352 |
+
q_pos_emb = q_pos_emb[:, -cur_len:, :, :]
|
| 353 |
+
k_pos_emb = k_pos_emb[:, -cur_len:, :, :]
|
| 354 |
+
query = apply_rotary_pos_emb(query, q_pos_emb)
|
| 355 |
+
key = apply_rotary_pos_emb(key, k_pos_emb)
|
| 356 |
|
| 357 |
if layer_past is not None:
|
| 358 |
past_key, past_value = layer_past[0], layer_past[1]
|
| 359 |
+
key = torch.cat((past_key, key), dim=1)
|
| 360 |
+
value = torch.cat((past_value, value), dim=1)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 361 |
|
| 362 |
if use_cache:
|
| 363 |
present = (key, value)
|
| 364 |
else:
|
| 365 |
present = None
|
| 366 |
|
| 367 |
+
if self.use_logn_attn and not self.training:
|
| 368 |
+
if self.logn_tensor.device != query.device or self.logn_tensor.dtype != query.dtype:
|
| 369 |
+
self.logn_tensor = self.logn_tensor.to(query.device).type_as(query)
|
| 370 |
+
seq_start = key.size(1) - query.size(1)
|
| 371 |
+
seq_end = key.size(1)
|
| 372 |
+
logn_tensor = self.logn_tensor[:, seq_start:seq_end, :, :]
|
|
|
|
|
|
|
|
|
|
| 373 |
query = query * logn_tensor.expand_as(query)
|
| 374 |
|
| 375 |
if (
|
|
|
|
| 379 |
and query.is_cuda
|
| 380 |
):
|
| 381 |
q, k, v = query, key, value
|
| 382 |
+
context_layer = self.core_attention_flash(q, k, v)
|
| 383 |
+
|
| 384 |
+
context_layer = rearrange(
|
| 385 |
+
context_layer, "b s h d -> b s (h d)"
|
| 386 |
+
).contiguous()
|
| 387 |
else:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 388 |
query = query.permute(0, 2, 1, 3)
|
| 389 |
+
key = key.permute(0, 2, 1, 3)
|
| 390 |
+
value = value.permute(0, 2, 1, 3)
|
| 391 |
+
attn_output, attn_weight = self._attn(
|
| 392 |
+
query, key, value, attention_mask, head_mask
|
| 393 |
+
)
|
| 394 |
+
context_layer = self._merge_heads(
|
| 395 |
+
attn_output, self.num_heads, self.head_dim
|
| 396 |
+
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 397 |
|
| 398 |
attn_output = self.c_proj(context_layer)
|
|
|
|
| 399 |
outputs = (attn_output, present)
|
| 400 |
if output_attentions:
|
| 401 |
if (
|
|
|
|
| 404 |
and not self.is_fp32
|
| 405 |
):
|
| 406 |
raise ValueError("Cannot output attentions while using flash-attn")
|
|
|
|
|
|
|
| 407 |
else:
|
| 408 |
outputs += (attn_weight,)
|
| 409 |
|
|
|
|
| 414 |
def __init__(self, config):
|
| 415 |
super().__init__()
|
| 416 |
self.w1 = nn.Linear(
|
| 417 |
+
config.hidden_size, config.ffn_hidden_size // 2, bias=not config.no_bias
|
| 418 |
)
|
| 419 |
self.w2 = nn.Linear(
|
| 420 |
+
config.hidden_size, config.ffn_hidden_size // 2, bias=not config.no_bias
|
| 421 |
)
|
| 422 |
+
ff_dim_in = config.ffn_hidden_size // 2
|
| 423 |
self.c_proj = nn.Linear(ff_dim_in, config.hidden_size, bias=not config.no_bias)
|
| 424 |
|
| 425 |
def forward(self, hidden_states):
|
|
|
|
| 431 |
|
| 432 |
|
| 433 |
class QWenBlock(nn.Module):
|
| 434 |
+
def __init__(self, config, layer_idx=None, num_expert=1):
|
| 435 |
super().__init__()
|
| 436 |
+
self.num_expert = num_expert
|
| 437 |
+
self.layer_number = layer_idx
|
| 438 |
+
self.apply_residual_connection_post_layernorm = (
|
| 439 |
+
config.apply_residual_connection_post_layernorm
|
| 440 |
+
)
|
| 441 |
hidden_size = config.hidden_size
|
| 442 |
+
self.apply_residual_connection_post_layernorm = (
|
| 443 |
+
config.apply_residual_connection_post_layernorm
|
| 444 |
+
)
|
| 445 |
self.bf16 = config.bf16
|
| 446 |
|
| 447 |
self.ln_1 = RMSNorm(
|
| 448 |
hidden_size,
|
| 449 |
eps=config.layer_norm_epsilon,
|
| 450 |
)
|
| 451 |
+
self.attn = QWenAttention(config, layer_number=layer_idx)
|
| 452 |
self.ln_2 = RMSNorm(
|
| 453 |
hidden_size,
|
| 454 |
eps=config.layer_norm_epsilon,
|
|
|
|
| 459 |
def forward(
|
| 460 |
self,
|
| 461 |
hidden_states: Optional[Tuple[torch.FloatTensor]],
|
|
|
|
| 462 |
layer_past: Optional[Tuple[torch.Tensor]] = None,
|
| 463 |
attention_mask: Optional[torch.FloatTensor] = None,
|
| 464 |
head_mask: Optional[torch.FloatTensor] = None,
|
|
|
|
| 471 |
|
| 472 |
attn_outputs = self.attn(
|
| 473 |
layernorm_output,
|
|
|
|
| 474 |
layer_past=layer_past,
|
| 475 |
attention_mask=attention_mask,
|
| 476 |
head_mask=head_mask,
|
|
|
|
| 481 |
|
| 482 |
outputs = attn_outputs[1:]
|
| 483 |
|
| 484 |
+
if self.apply_residual_connection_post_layernorm:
|
| 485 |
+
residual = layernorm_output
|
| 486 |
+
else:
|
| 487 |
+
residual = hidden_states
|
| 488 |
layernorm_input = attn_output + residual
|
| 489 |
|
| 490 |
layernorm_output = self.ln_2(layernorm_input)
|
| 491 |
|
| 492 |
+
if self.apply_residual_connection_post_layernorm:
|
| 493 |
+
residual = layernorm_output
|
| 494 |
+
else:
|
| 495 |
+
residual = layernorm_input
|
| 496 |
+
|
| 497 |
mlp_output = self.mlp(layernorm_output)
|
| 498 |
hidden_states = residual + mlp_output
|
| 499 |
|
|
|
|
| 511 |
is_parallelizable = False
|
| 512 |
supports_gradient_checkpointing = True
|
| 513 |
_no_split_modules = ["QWenBlock"]
|
|
|
|
| 514 |
|
| 515 |
def __init__(self, *inputs, **kwargs):
|
| 516 |
super().__init__(*inputs, **kwargs)
|
|
|
|
| 534 |
mean=0.0,
|
| 535 |
std=(
|
| 536 |
self.config.initializer_range
|
| 537 |
+
/ math.sqrt(2 * self.config.n_layer)
|
| 538 |
),
|
| 539 |
)
|
| 540 |
|
|
|
|
| 548 |
|
| 549 |
def __init__(self, config):
|
| 550 |
super().__init__(config)
|
| 551 |
+
self.vocab_size = config.padded_vocab_size
|
| 552 |
self.num_hidden_layers = config.num_hidden_layers
|
| 553 |
self.embed_dim = config.hidden_size
|
|
|
|
| 554 |
|
| 555 |
+
max_sequence_length = config.max_position_embeddings
|
| 556 |
+
self.position_embedding_type = config.pos_emb
|
| 557 |
self.gradient_checkpointing = False
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 558 |
|
| 559 |
+
if self.position_embedding_type == "learned":
|
| 560 |
+
self.wpe = nn.Embedding(max_sequence_length, self.embed_dim)
|
| 561 |
+
self.init_method(self.position_embeddings.weight)
|
| 562 |
+
self._position_embeddings_key = "position_embeddings"
|
| 563 |
+
self.init_method(self.position_embeddings.weight)
|
| 564 |
else:
|
| 565 |
+
self.wpe = None
|
| 566 |
+
self._position_embeddings_key = ""
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 567 |
|
| 568 |
+
self.wte = nn.Embedding(self.vocab_size, self.embed_dim)
|
|
|
|
| 569 |
|
| 570 |
+
self.drop = nn.Dropout(config.embd_pdrop)
|
| 571 |
self.h = nn.ModuleList(
|
| 572 |
[
|
| 573 |
QWenBlock(
|
| 574 |
+
config,
|
| 575 |
+
layer_idx=i,
|
| 576 |
)
|
| 577 |
for i in range(config.num_hidden_layers)
|
| 578 |
]
|
|
|
|
| 590 |
def set_input_embeddings(self, new_embeddings):
|
| 591 |
self.wte = new_embeddings
|
| 592 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 593 |
def forward(
|
| 594 |
self,
|
| 595 |
input_ids: Optional[torch.LongTensor] = None,
|
|
|
|
| 646 |
past_length = 0
|
| 647 |
past_key_values = tuple([None] * len(self.h))
|
| 648 |
else:
|
| 649 |
+
past_length = past_key_values[0][0].size(-2)
|
| 650 |
+
|
|
|
|
|
|
|
| 651 |
if position_ids is None:
|
| 652 |
position_ids = torch.arange(
|
| 653 |
past_length,
|
|
|
|
| 666 |
attention_mask = (1.0 - attention_mask) * torch.finfo(self.dtype).min
|
| 667 |
|
| 668 |
encoder_attention_mask = None
|
| 669 |
+
head_mask = self.get_head_mask(head_mask, self.config.n_layer)
|
| 670 |
|
| 671 |
if inputs_embeds is None:
|
| 672 |
inputs_embeds = self.wte(input_ids)
|
| 673 |
hidden_states = inputs_embeds
|
| 674 |
+
if self.wpe is not None:
|
| 675 |
+
position_embeds = self.wpe(position_ids)
|
| 676 |
+
hidden_states = hidden_states + position_embeds
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 677 |
|
| 678 |
hidden_states = self.drop(hidden_states)
|
| 679 |
output_shape = input_shape + (hidden_states.size(-1),)
|
|
|
|
| 705 |
outputs = torch.utils.checkpoint.checkpoint(
|
| 706 |
create_custom_forward(block),
|
| 707 |
hidden_states,
|
|
|
|
| 708 |
None,
|
| 709 |
attention_mask,
|
| 710 |
head_mask[i],
|
|
|
|
| 715 |
outputs = block(
|
| 716 |
hidden_states,
|
| 717 |
layer_past=layer_past,
|
|
|
|
| 718 |
attention_mask=attention_mask,
|
| 719 |
head_mask=head_mask[i],
|
| 720 |
encoder_hidden_states=encoder_hidden_states,
|
|
|
|
| 725 |
|
| 726 |
hidden_states = outputs[0]
|
| 727 |
if use_cache is True:
|
| 728 |
+
presents = presents + (outputs[2 if output_attentions else 1],)
|
| 729 |
|
| 730 |
if output_attentions:
|
| 731 |
+
all_self_attentions = all_self_attentions + (outputs[1],)
|
| 732 |
|
| 733 |
hidden_states = self.ln_f(hidden_states)
|
| 734 |
hidden_states = hidden_states.view(output_shape)
|
|
|
|
|
|
|
|
|
|
| 735 |
|
| 736 |
if not return_dict:
|
| 737 |
return tuple(
|
|
|
|
| 783 |
logger.warn("Your device support faster inference by passing bf16=True in \"AutoModelForCausalLM.from_pretrained\".")
|
| 784 |
elif SUPPORT_FP16:
|
| 785 |
logger.warn("Your device support faster inference by passing fp16=True in \"AutoModelForCausalLM.from_pretrained\".")
|
| 786 |
+
|
| 787 |
if config.use_flash_attn == "auto":
|
| 788 |
if config.bf16 or config.fp16:
|
| 789 |
logger.warn("Try importing flash-attention for faster inference...")
|
|
|
|
| 794 |
logger.warn("Flash attention will be disabled because it does NOT support fp32.")
|
| 795 |
|
| 796 |
if config.use_flash_attn:
|
| 797 |
+
global apply_rotary_emb_func, rms_norm, flash_attn_unpadded_func
|
| 798 |
+
try:
|
| 799 |
+
from flash_attn.layers.rotary import apply_rotary_emb_func as __apply_rotary_emb_func
|
| 800 |
+
apply_rotary_emb_func = __apply_rotary_emb_func
|
| 801 |
+
except ImportError:
|
| 802 |
+
logger.warn(
|
| 803 |
+
"Warning: import flash_attn rotary fail, please install FlashAttention rotary to get higher efficiency "
|
| 804 |
+
"https://github.com/Dao-AILab/flash-attention/tree/main/csrc/rotary"
|
| 805 |
+
)
|
| 806 |
+
|
| 807 |
+
try:
|
| 808 |
+
from flash_attn.ops.rms_norm import rms_norm as __rms_norm
|
| 809 |
+
rms_norm = __rms_norm
|
| 810 |
+
except ImportError:
|
| 811 |
+
logger.warn(
|
| 812 |
+
"Warning: import flash_attn rms_norm fail, please install FlashAttention layer_norm to get higher efficiency "
|
| 813 |
+
"https://github.com/Dao-AILab/flash-attention/tree/main/csrc/layer_norm"
|
| 814 |
+
)
|
| 815 |
+
|
| 816 |
+
try:
|
| 817 |
+
from flash_attn.flash_attn_interface import flash_attn_unpadded_func as __flash_attn_unpadded_func
|
| 818 |
+
flash_attn_unpadded_func = __flash_attn_unpadded_func
|
| 819 |
+
except ImportError:
|
| 820 |
+
logger.warn(
|
| 821 |
+
"Warning: import flash_attn fail, please install FlashAttention to get higher efficiency "
|
| 822 |
+
"https://github.com/Dao-AILab/flash-attention"
|
| 823 |
+
)
|
| 824 |
|
| 825 |
self.transformer = QWenModel(config)
|
| 826 |
+
self.lm_head = nn.Linear(config.n_embd, config.vocab_size, bias=False)
|
| 827 |
|
| 828 |
if config.bf16:
|
| 829 |
self.transformer.bfloat16()
|
|
|
|
| 842 |
def prepare_inputs_for_generation(
|
| 843 |
self, input_ids, past_key_values=None, inputs_embeds=None, **kwargs
|
| 844 |
):
|
| 845 |
+
token_type_ids = kwargs.get("token_type_ids", None)
|
| 846 |
if past_key_values:
|
| 847 |
input_ids = input_ids[:, -1].unsqueeze(-1)
|
| 848 |
+
if token_type_ids is not None:
|
| 849 |
+
token_type_ids = token_type_ids[:, -1].unsqueeze(-1)
|
| 850 |
+
|
| 851 |
+
attention_mask = kwargs.get("attention_mask", None)
|
| 852 |
+
position_ids = kwargs.get("position_ids", None)
|
| 853 |
|
| 854 |
+
if attention_mask is not None and position_ids is None:
|
| 855 |
+
position_ids = attention_mask.long().cumsum(-1) - 1
|
| 856 |
+
position_ids.masked_fill_(attention_mask == 0, 1)
|
| 857 |
+
if past_key_values:
|
| 858 |
+
position_ids = position_ids[:, -1].unsqueeze(-1)
|
| 859 |
else:
|
| 860 |
+
position_ids = None
|
| 861 |
|
| 862 |
if inputs_embeds is not None and past_key_values is None:
|
| 863 |
model_inputs = {"inputs_embeds": inputs_embeds}
|
|
|
|
| 868 |
{
|
| 869 |
"past_key_values": past_key_values,
|
| 870 |
"use_cache": kwargs.get("use_cache"),
|
| 871 |
+
"position_ids": position_ids,
|
| 872 |
"attention_mask": attention_mask,
|
| 873 |
+
"token_type_ids": token_type_ids,
|
| 874 |
}
|
| 875 |
)
|
| 876 |
return model_inputs
|
|
|
|
| 957 |
query: str,
|
| 958 |
history: Optional[HistoryType],
|
| 959 |
system: str = "You are a helpful assistant.",
|
| 960 |
+
append_history: bool = True,
|
| 961 |
+
stream: Optional[bool] = False,
|
| 962 |
stop_words_ids: Optional[List[List[int]]] = None,
|
|
|
|
| 963 |
**kwargs,
|
| 964 |
) -> Tuple[str, HistoryType]:
|
|
|
|
|
|
|
|
|
|
|
|
|
| 965 |
if history is None:
|
| 966 |
history = []
|
|
|
|
|
|
|
|
|
|
|
|
|
| 967 |
if stop_words_ids is None:
|
| 968 |
stop_words_ids = []
|
| 969 |
|
|
|
|
|
|
|
|
|
|
| 970 |
raw_text, context_tokens = make_context(
|
| 971 |
tokenizer,
|
| 972 |
query,
|
| 973 |
history=history,
|
| 974 |
system=system,
|
| 975 |
+
max_window_size=6144,
|
| 976 |
+
chat_format=self.generation_config.chat_format,
|
| 977 |
)
|
| 978 |
|
| 979 |
stop_words_ids.extend(get_stop_words_ids(
|
| 980 |
+
self.generation_config.chat_format, tokenizer
|
| 981 |
))
|
| 982 |
input_ids = torch.tensor([context_tokens]).to(self.device)
|
| 983 |
+
if stream:
|
| 984 |
+
assert self.generation_config.chat_format == 'chatml'
|
| 985 |
+
from transformers_stream_generator.main import NewGenerationMixin, StreamGenerationConfig
|
| 986 |
+
self.__class__.generate = NewGenerationMixin.generate
|
| 987 |
+
self.__class__.sample_stream = NewGenerationMixin.sample_stream
|
| 988 |
+
stream_config = StreamGenerationConfig(**self.generation_config.to_dict(), do_stream=True)
|
| 989 |
+
def stream_generator():
|
| 990 |
+
outputs = []
|
| 991 |
+
for token in self.generate(
|
| 992 |
+
input_ids, return_dict_in_generate=False, generation_config=stream_config, **kwargs):
|
| 993 |
+
outputs.append(token.item())
|
| 994 |
+
if outputs[-1] in (tokenizer.im_end_id, tokenizer.im_start_id):
|
| 995 |
+
break
|
| 996 |
+
yield tokenizer.decode(outputs, skip_special_tokens=True)
|
| 997 |
+
|
| 998 |
+
return stream_generator()
|
| 999 |
+
else:
|
| 1000 |
+
outputs = self.generate(
|
| 1001 |
+
input_ids,
|
| 1002 |
+
stop_words_ids = stop_words_ids,
|
| 1003 |
+
return_dict_in_generate = False,
|
| 1004 |
+
**kwargs,
|
| 1005 |
+
)
|
| 1006 |
+
|
| 1007 |
+
response = decode_tokens(
|
| 1008 |
+
outputs[0],
|
| 1009 |
+
tokenizer,
|
| 1010 |
+
raw_text_len=len(raw_text),
|
| 1011 |
+
context_length=len(context_tokens),
|
| 1012 |
+
chat_format=self.generation_config.chat_format,
|
| 1013 |
+
verbose=False,
|
| 1014 |
+
)
|
| 1015 |
|
| 1016 |
+
if append_history:
|
| 1017 |
+
history.append((query, response))
|
|
|
|
|
|
|
|
|
|
| 1018 |
|
| 1019 |
return response, history
|
| 1020 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1021 |
def generate(
|
| 1022 |
self,
|
| 1023 |
inputs: Optional[torch.Tensor] = None,
|
|
|
|
| 1028 |
Callable[[int, torch.Tensor], List[int]]
|
| 1029 |
] = None,
|
| 1030 |
synced_gpus: Optional[bool] = None,
|
|
|
|
| 1031 |
streamer: Optional["BaseStreamer"] = None,
|
| 1032 |
**kwargs,
|
| 1033 |
) -> Union[GenerateOutput, torch.LongTensor]:
|
|
|
|
|
|
|
| 1034 |
# Process stop_words_ids.
|
| 1035 |
stop_words_ids = kwargs.pop("stop_words_ids", None)
|
| 1036 |
if stop_words_ids is None and generation_config is not None:
|
| 1037 |
stop_words_ids = getattr(generation_config, "stop_words_ids", None)
|
| 1038 |
if stop_words_ids is None:
|
| 1039 |
+
stop_words_ids = getattr(self.generation_config, "stop_words_ids", None)
|
| 1040 |
|
| 1041 |
if stop_words_ids is not None:
|
| 1042 |
stop_words_logits_processor = StopWordsLogitsProcessor(
|
| 1043 |
stop_words_ids=stop_words_ids,
|
| 1044 |
+
eos_token_id=self.generation_config.eos_token_id,
|
| 1045 |
)
|
| 1046 |
if logits_processor is None:
|
| 1047 |
logits_processor = LogitsProcessorList([stop_words_logits_processor])
|
|
|
|
| 1050 |
|
| 1051 |
return super().generate(
|
| 1052 |
inputs,
|
| 1053 |
+
generation_config,
|
| 1054 |
+
logits_processor,
|
| 1055 |
+
stopping_criteria,
|
| 1056 |
+
prefix_allowed_tokens_fn,
|
| 1057 |
+
synced_gpus,
|
| 1058 |
+
streamer,
|
|
|
|
| 1059 |
**kwargs,
|
| 1060 |
)
|
| 1061 |
|
|
|
|
| 1065 |
super().__init__()
|
| 1066 |
self.dim = dim
|
| 1067 |
self.base = base
|
| 1068 |
+
self.inv_freq = 1.0 / (base ** (torch.arange(0, dim, 2).float() / dim))
|
|
|
|
| 1069 |
if importlib.util.find_spec("einops") is None:
|
| 1070 |
raise RuntimeError("einops is required for Rotary Embedding")
|
| 1071 |
|
| 1072 |
self._rotary_pos_emb_cache = None
|
| 1073 |
self._seq_len_cached = 0
|
| 1074 |
self._ntk_alpha_cached = 1.0
|
|
|
|
| 1075 |
|
| 1076 |
+
def update_rotary_pos_emb_cache(self, max_seq_len, offset=0, ntk_alpha=1.0):
|
| 1077 |
+
seqlen = max_seq_len + offset
|
| 1078 |
if seqlen > self._seq_len_cached or ntk_alpha != self._ntk_alpha_cached:
|
| 1079 |
base = self.base * ntk_alpha ** (self.dim / (self.dim - 2))
|
| 1080 |
self.inv_freq = 1.0 / (
|
|
|
|
| 1084 |
/ self.dim
|
| 1085 |
)
|
| 1086 |
)
|
| 1087 |
+
self._seq_len_cached = seqlen
|
| 1088 |
self._ntk_alpha_cached = ntk_alpha
|
| 1089 |
+
seq = torch.arange(seqlen, device=self.inv_freq.device)
|
| 1090 |
freqs = torch.outer(seq.type_as(self.inv_freq), self.inv_freq)
|
|
|
|
| 1091 |
emb = torch.cat((freqs, freqs), dim=-1)
|
| 1092 |
from einops import rearrange
|
| 1093 |
|
| 1094 |
+
self._rotary_pos_emb_cache = rearrange(emb, "n d -> 1 n 1 d")
|
|
|
|
|
|
|
|
|
|
| 1095 |
|
| 1096 |
+
def forward(self, max_seq_len, offset=0, ntk_alpha=1.0):
|
| 1097 |
+
self.update_rotary_pos_emb_cache(max_seq_len, offset, ntk_alpha)
|
| 1098 |
+
return self._rotary_pos_emb_cache[:, offset : offset + max_seq_len]
|
|
|
|
| 1099 |
|
| 1100 |
|
| 1101 |
def _rotate_half(x):
|
|
|
|
| 1107 |
|
| 1108 |
|
| 1109 |
def apply_rotary_pos_emb(t, freqs):
|
| 1110 |
+
if apply_rotary_emb_func is not None:
|
| 1111 |
+
t_ = t.float()
|
| 1112 |
+
freqs = freqs.squeeze(0).squeeze(1)
|
| 1113 |
+
cos = freqs[:, : freqs.shape[-1] // 2].cos()
|
| 1114 |
+
sin = freqs[:, : freqs.shape[-1] // 2].sin()
|
| 1115 |
+
output = apply_rotary_emb_func(t_, cos, sin).type_as(t)
|
| 1116 |
+
return output
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1117 |
else:
|
| 1118 |
+
rot_dim = freqs.shape[-1]
|
| 1119 |
+
t_, t_pass_ = t[..., :rot_dim], t[..., rot_dim:]
|
| 1120 |
+
t_ = t_.float()
|
| 1121 |
+
t_pass_ = t_pass_.float()
|
| 1122 |
+
t_ = (t_ * freqs.cos()) + (_rotate_half(t_) * freqs.sin())
|
| 1123 |
+
return torch.cat((t_, t_pass_), dim=-1).type_as(t)
|
| 1124 |
|
| 1125 |
|
| 1126 |
class RMSNorm(torch.nn.Module):
|
model-00001-of-00008.safetensors → pytorch_model.bin
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7361b298a82b284129276f586dec570b5d41259130d190960321cc0db92d958f
|
| 3 |
+
size 15442733145
|
qwen_generation_utils.py
CHANGED
|
@@ -198,9 +198,8 @@ def _decode_default(
|
|
| 198 |
raw_text_len: int,
|
| 199 |
verbose: bool = False,
|
| 200 |
return_end_reason: bool = False,
|
| 201 |
-
errors: str='replace',
|
| 202 |
):
|
| 203 |
-
trim_decode_tokens = tokenizer.decode(tokens
|
| 204 |
if verbose:
|
| 205 |
print("\nRaw Generate: ", trim_decode_tokens)
|
| 206 |
|
|
@@ -232,7 +231,6 @@ def _decode_chatml(
|
|
| 232 |
context_length: int,
|
| 233 |
verbose: bool = False,
|
| 234 |
return_end_reason: bool = False,
|
| 235 |
-
errors: str='replace'
|
| 236 |
):
|
| 237 |
end_reason = f"Gen length {len(tokens)}"
|
| 238 |
eod_token_idx = context_length
|
|
@@ -241,9 +239,9 @@ def _decode_chatml(
|
|
| 241 |
end_reason = f"Gen {tokenizer.decode([tokens[eod_token_idx]])!r}"
|
| 242 |
break
|
| 243 |
|
| 244 |
-
trim_decode_tokens = tokenizer.decode(tokens[:eod_token_idx]
|
| 245 |
if verbose:
|
| 246 |
-
print("\nRaw Generate w/o EOD:", tokenizer.decode(tokens
|
| 247 |
print("\nRaw Generate:", trim_decode_tokens)
|
| 248 |
print("\nEnd Reason:", end_reason)
|
| 249 |
for stop_word in stop_words:
|
|
@@ -266,7 +264,6 @@ def decode_tokens(
|
|
| 266 |
chat_format: str,
|
| 267 |
verbose: bool = False,
|
| 268 |
return_end_reason: bool = False,
|
| 269 |
-
errors: str="replace",
|
| 270 |
) -> str:
|
| 271 |
if torch.is_tensor(tokens):
|
| 272 |
tokens = tokens.cpu().numpy().tolist()
|
|
@@ -281,7 +278,6 @@ def decode_tokens(
|
|
| 281 |
context_length=context_length,
|
| 282 |
verbose=verbose,
|
| 283 |
return_end_reason=return_end_reason,
|
| 284 |
-
errors=errors,
|
| 285 |
)
|
| 286 |
elif chat_format == "raw":
|
| 287 |
return _decode_default(
|
|
@@ -292,7 +288,6 @@ def decode_tokens(
|
|
| 292 |
raw_text_len=raw_text_len,
|
| 293 |
verbose=verbose,
|
| 294 |
return_end_reason=return_end_reason,
|
| 295 |
-
errors=errors,
|
| 296 |
)
|
| 297 |
else:
|
| 298 |
raise NotImplementedError(f"Unknown chat format {chat_format!r}")
|
|
|
|
| 198 |
raw_text_len: int,
|
| 199 |
verbose: bool = False,
|
| 200 |
return_end_reason: bool = False,
|
|
|
|
| 201 |
):
|
| 202 |
+
trim_decode_tokens = tokenizer.decode(tokens)[raw_text_len:]
|
| 203 |
if verbose:
|
| 204 |
print("\nRaw Generate: ", trim_decode_tokens)
|
| 205 |
|
|
|
|
| 231 |
context_length: int,
|
| 232 |
verbose: bool = False,
|
| 233 |
return_end_reason: bool = False,
|
|
|
|
| 234 |
):
|
| 235 |
end_reason = f"Gen length {len(tokens)}"
|
| 236 |
eod_token_idx = context_length
|
|
|
|
| 239 |
end_reason = f"Gen {tokenizer.decode([tokens[eod_token_idx]])!r}"
|
| 240 |
break
|
| 241 |
|
| 242 |
+
trim_decode_tokens = tokenizer.decode(tokens[:eod_token_idx])[raw_text_len:]
|
| 243 |
if verbose:
|
| 244 |
+
print("\nRaw Generate w/o EOD:", tokenizer.decode(tokens)[raw_text_len:])
|
| 245 |
print("\nRaw Generate:", trim_decode_tokens)
|
| 246 |
print("\nEnd Reason:", end_reason)
|
| 247 |
for stop_word in stop_words:
|
|
|
|
| 264 |
chat_format: str,
|
| 265 |
verbose: bool = False,
|
| 266 |
return_end_reason: bool = False,
|
|
|
|
| 267 |
) -> str:
|
| 268 |
if torch.is_tensor(tokens):
|
| 269 |
tokens = tokens.cpu().numpy().tolist()
|
|
|
|
| 278 |
context_length=context_length,
|
| 279 |
verbose=verbose,
|
| 280 |
return_end_reason=return_end_reason,
|
|
|
|
| 281 |
)
|
| 282 |
elif chat_format == "raw":
|
| 283 |
return _decode_default(
|
|
|
|
| 288 |
raw_text_len=raw_text_len,
|
| 289 |
verbose=verbose,
|
| 290 |
return_end_reason=return_end_reason,
|
|
|
|
| 291 |
)
|
| 292 |
else:
|
| 293 |
raise NotImplementedError(f"Unknown chat format {chat_format!r}")
|
special_tokens_map.json
ADDED
|
@@ -0,0 +1,30 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"bos_token": {
|
| 3 |
+
"content": "<|im_start|>",
|
| 4 |
+
"lstrip": false,
|
| 5 |
+
"normalized": true,
|
| 6 |
+
"rstrip": false,
|
| 7 |
+
"single_word": true
|
| 8 |
+
},
|
| 9 |
+
"eos_token": {
|
| 10 |
+
"content": "<|endoftext|>",
|
| 11 |
+
"lstrip": false,
|
| 12 |
+
"normalized": true,
|
| 13 |
+
"rstrip": false,
|
| 14 |
+
"single_word": true
|
| 15 |
+
},
|
| 16 |
+
"unk_token": {
|
| 17 |
+
"content": "<|im_end|>",
|
| 18 |
+
"lstrip": false,
|
| 19 |
+
"normalized": true,
|
| 20 |
+
"rstrip": false,
|
| 21 |
+
"single_word": true
|
| 22 |
+
},
|
| 23 |
+
"pad_token": {
|
| 24 |
+
"content": "<|im_end|>",
|
| 25 |
+
"lstrip": false,
|
| 26 |
+
"normalized": true,
|
| 27 |
+
"rstrip": false,
|
| 28 |
+
"single_word": true
|
| 29 |
+
}
|
| 30 |
+
}
|
tokenization_qwen.py
CHANGED
|
@@ -27,33 +27,20 @@ IMEND = "<|im_end|>"
|
|
| 27 |
# regular texts, the surface forms of special tokens need to be
|
| 28 |
# as different as possible to minimize the impact
|
| 29 |
EXTRAS = tuple((f"<|extra_{i}|>" for i in range(205)))
|
| 30 |
-
|
| 31 |
-
|
| 32 |
-
|
| 33 |
-
|
| 34 |
-
|
| 35 |
-
(
|
| 36 |
-
ENDOFTEXT,
|
| 37 |
-
IMSTART,
|
| 38 |
-
IMEND,
|
| 39 |
-
)
|
| 40 |
-
+ EXTRAS
|
| 41 |
-
),
|
| 42 |
-
start=SPECIAL_START_ID,
|
| 43 |
-
)
|
| 44 |
-
)
|
| 45 |
-
SPECIAL_TOKENS_SET = set(t for i, t in SPECIAL_TOKENS)
|
| 46 |
|
| 47 |
|
| 48 |
def _load_tiktoken_bpe(tiktoken_bpe_file: str) -> Dict[bytes, int]:
|
| 49 |
-
|
| 50 |
-
contents = f.read()
|
| 51 |
return {
|
| 52 |
base64.b64decode(token): int(rank)
|
| 53 |
for token, rank in (line.split() for line in contents.splitlines() if line)
|
| 54 |
}
|
| 55 |
|
| 56 |
-
|
| 57 |
class QWenTokenizer(PreTrainedTokenizer):
|
| 58 |
"""QWen tokenizer."""
|
| 59 |
|
|
@@ -63,35 +50,20 @@ class QWenTokenizer(PreTrainedTokenizer):
|
|
| 63 |
self,
|
| 64 |
vocab_file,
|
| 65 |
errors="replace",
|
| 66 |
-
extra_vocab_file=None,
|
| 67 |
**kwargs,
|
| 68 |
):
|
| 69 |
super().__init__(**kwargs)
|
| 70 |
|
| 71 |
-
# how to handle errors in decoding
|
| 72 |
-
# use ignore if you are in streaming inference
|
| 73 |
-
self.errors = errors
|
| 74 |
|
| 75 |
-
self.mergeable_ranks = _load_tiktoken_bpe(vocab_file) # type:
|
| 76 |
self.special_tokens = {
|
| 77 |
token: index
|
| 78 |
-
for index, token in
|
|
|
|
|
|
|
| 79 |
}
|
| 80 |
|
| 81 |
-
# try load extra vocab from file
|
| 82 |
-
if extra_vocab_file is not None:
|
| 83 |
-
used_ids = set(self.mergeable_ranks.values()) | set(self.special_tokens.values())
|
| 84 |
-
extra_mergeable_ranks = _load_tiktoken_bpe(extra_vocab_file)
|
| 85 |
-
for token, index in extra_mergeable_ranks.items():
|
| 86 |
-
if token in self.mergeable_ranks:
|
| 87 |
-
logger.info(f"extra token {token} exists, skipping")
|
| 88 |
-
continue
|
| 89 |
-
if index in used_ids:
|
| 90 |
-
logger.info(f'the index {index} for extra token {token} exists, skipping')
|
| 91 |
-
continue
|
| 92 |
-
self.mergeable_ranks[token] = index
|
| 93 |
-
# the index may be sparse after this, but don't worry tiktoken.Encoding will handle this
|
| 94 |
-
|
| 95 |
enc = tiktoken.Encoding(
|
| 96 |
"Qwen",
|
| 97 |
pat_str=PAT_STR,
|
|
@@ -113,23 +85,6 @@ class QWenTokenizer(PreTrainedTokenizer):
|
|
| 113 |
self.im_start_id = self.special_tokens[IMSTART]
|
| 114 |
self.im_end_id = self.special_tokens[IMEND]
|
| 115 |
|
| 116 |
-
def __getstate__(self):
|
| 117 |
-
# for pickle lovers
|
| 118 |
-
state = self.__dict__.copy()
|
| 119 |
-
del state["tokenizer"]
|
| 120 |
-
return state
|
| 121 |
-
|
| 122 |
-
def __setstate__(self, state):
|
| 123 |
-
# tokenizer is not python native; don't pass it; rebuild it
|
| 124 |
-
self.__dict__.update(state)
|
| 125 |
-
enc = tiktoken.Encoding(
|
| 126 |
-
"Qwen",
|
| 127 |
-
pat_str=PAT_STR,
|
| 128 |
-
mergeable_ranks=self.mergeable_ranks,
|
| 129 |
-
special_tokens=self.special_tokens,
|
| 130 |
-
)
|
| 131 |
-
self.tokenizer = enc
|
| 132 |
-
|
| 133 |
def __len__(self) -> int:
|
| 134 |
return self.tokenizer.n_vocab
|
| 135 |
|
|
@@ -152,17 +107,13 @@ class QWenTokenizer(PreTrainedTokenizer):
|
|
| 152 |
ids.append(self.mergeable_ranks.get(token))
|
| 153 |
return ids
|
| 154 |
|
| 155 |
-
def _add_tokens(
|
| 156 |
-
self,
|
| 157 |
-
new_tokens: Union[List[str], List[AddedToken]],
|
| 158 |
-
special_tokens: bool = False,
|
| 159 |
-
) -> int:
|
| 160 |
if not special_tokens and new_tokens:
|
| 161 |
-
raise ValueError(
|
| 162 |
for token in new_tokens:
|
| 163 |
surface_form = token.content if isinstance(token, AddedToken) else token
|
| 164 |
-
if surface_form not in
|
| 165 |
-
raise ValueError(
|
| 166 |
return 0
|
| 167 |
|
| 168 |
def save_vocabulary(self, save_directory: str, **kwargs) -> Tuple[str]:
|
|
@@ -266,11 +217,10 @@ class QWenTokenizer(PreTrainedTokenizer):
|
|
| 266 |
self,
|
| 267 |
token_ids: Union[int, List[int]],
|
| 268 |
skip_special_tokens: bool = False,
|
| 269 |
-
errors: str = None,
|
| 270 |
**kwargs,
|
| 271 |
) -> str:
|
| 272 |
if isinstance(token_ids, int):
|
| 273 |
token_ids = [token_ids]
|
| 274 |
if skip_special_tokens:
|
| 275 |
token_ids = [i for i in token_ids if i < self.eod_id]
|
| 276 |
-
return self.tokenizer.decode(token_ids, errors=
|
|
|
|
| 27 |
# regular texts, the surface forms of special tokens need to be
|
| 28 |
# as different as possible to minimize the impact
|
| 29 |
EXTRAS = tuple((f"<|extra_{i}|>" for i in range(205)))
|
| 30 |
+
SPECIAL_TOKENS = (
|
| 31 |
+
ENDOFTEXT,
|
| 32 |
+
IMSTART,
|
| 33 |
+
IMEND,
|
| 34 |
+
) + EXTRAS
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 35 |
|
| 36 |
|
| 37 |
def _load_tiktoken_bpe(tiktoken_bpe_file: str) -> Dict[bytes, int]:
|
| 38 |
+
contents = open(tiktoken_bpe_file, "rb").read()
|
|
|
|
| 39 |
return {
|
| 40 |
base64.b64decode(token): int(rank)
|
| 41 |
for token, rank in (line.split() for line in contents.splitlines() if line)
|
| 42 |
}
|
| 43 |
|
|
|
|
| 44 |
class QWenTokenizer(PreTrainedTokenizer):
|
| 45 |
"""QWen tokenizer."""
|
| 46 |
|
|
|
|
| 50 |
self,
|
| 51 |
vocab_file,
|
| 52 |
errors="replace",
|
|
|
|
| 53 |
**kwargs,
|
| 54 |
):
|
| 55 |
super().__init__(**kwargs)
|
| 56 |
|
| 57 |
+
self.errors = errors # how to handle errors in decoding
|
|
|
|
|
|
|
| 58 |
|
| 59 |
+
self.mergeable_ranks = _load_tiktoken_bpe(vocab_file) # type: dict[bytes, int]
|
| 60 |
self.special_tokens = {
|
| 61 |
token: index
|
| 62 |
+
for index, token in enumerate(
|
| 63 |
+
SPECIAL_TOKENS, start=len(self.mergeable_ranks)
|
| 64 |
+
)
|
| 65 |
}
|
| 66 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 67 |
enc = tiktoken.Encoding(
|
| 68 |
"Qwen",
|
| 69 |
pat_str=PAT_STR,
|
|
|
|
| 85 |
self.im_start_id = self.special_tokens[IMSTART]
|
| 86 |
self.im_end_id = self.special_tokens[IMEND]
|
| 87 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 88 |
def __len__(self) -> int:
|
| 89 |
return self.tokenizer.n_vocab
|
| 90 |
|
|
|
|
| 107 |
ids.append(self.mergeable_ranks.get(token))
|
| 108 |
return ids
|
| 109 |
|
| 110 |
+
def _add_tokens(self, new_tokens: Union[List[str], List[AddedToken]], special_tokens: bool = False) -> int:
|
|
|
|
|
|
|
|
|
|
|
|
|
| 111 |
if not special_tokens and new_tokens:
|
| 112 |
+
raise ValueError('Adding regular tokens is not supported')
|
| 113 |
for token in new_tokens:
|
| 114 |
surface_form = token.content if isinstance(token, AddedToken) else token
|
| 115 |
+
if surface_form not in SPECIAL_TOKENS:
|
| 116 |
+
raise ValueError('Adding unknown special tokens is not supported')
|
| 117 |
return 0
|
| 118 |
|
| 119 |
def save_vocabulary(self, save_directory: str, **kwargs) -> Tuple[str]:
|
|
|
|
| 217 |
self,
|
| 218 |
token_ids: Union[int, List[int]],
|
| 219 |
skip_special_tokens: bool = False,
|
|
|
|
| 220 |
**kwargs,
|
| 221 |
) -> str:
|
| 222 |
if isinstance(token_ids, int):
|
| 223 |
token_ids = [token_ids]
|
| 224 |
if skip_special_tokens:
|
| 225 |
token_ids = [i for i in token_ids if i < self.eod_id]
|
| 226 |
+
return self.tokenizer.decode(token_ids, errors=self.errors)
|
tokenizer_config.json
CHANGED
|
@@ -1,5 +1,5 @@
|
|
| 1 |
{
|
| 2 |
-
"model_max_length":
|
| 3 |
"tokenizer_class": "QWenTokenizer",
|
| 4 |
"auto_map": {
|
| 5 |
"AutoTokenizer": [
|
|
|
|
| 1 |
{
|
| 2 |
+
"model_max_length": 8192,
|
| 3 |
"tokenizer_class": "QWenTokenizer",
|
| 4 |
"auto_map": {
|
| 5 |
"AutoTokenizer": [
|