gptqmodel 1.0.1__tar.gz

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (147) hide show
  1. gptqmodel-1.0.1/LICENSE +201 -0
  2. gptqmodel-1.0.1/MANIFEST.in +6 -0
  3. gptqmodel-1.0.1/PKG-INFO +320 -0
  4. gptqmodel-1.0.1/README.md +291 -0
  5. gptqmodel-1.0.1/gptqmodel/__init__.py +5 -0
  6. gptqmodel-1.0.1/gptqmodel/eval_tasks/__init__.py +3 -0
  7. gptqmodel-1.0.1/gptqmodel/eval_tasks/_base.py +65 -0
  8. gptqmodel-1.0.1/gptqmodel/eval_tasks/_utils/__init__.py +0 -0
  9. gptqmodel-1.0.1/gptqmodel/eval_tasks/_utils/classification_utils.py +42 -0
  10. gptqmodel-1.0.1/gptqmodel/eval_tasks/_utils/data_utils.py +263 -0
  11. gptqmodel-1.0.1/gptqmodel/eval_tasks/_utils/generation_utils.py +34 -0
  12. gptqmodel-1.0.1/gptqmodel/eval_tasks/language_modeling_task.py +45 -0
  13. gptqmodel-1.0.1/gptqmodel/eval_tasks/sequence_classification_task.py +101 -0
  14. gptqmodel-1.0.1/gptqmodel/eval_tasks/text_summarization_task.py +70 -0
  15. gptqmodel-1.0.1/gptqmodel/integration/__init__.py +1 -0
  16. gptqmodel-1.0.1/gptqmodel/integration/optimum/__init__.py +2 -0
  17. gptqmodel-1.0.1/gptqmodel/integration/optimum/constants.py +23 -0
  18. gptqmodel-1.0.1/gptqmodel/integration/optimum/data.py +257 -0
  19. gptqmodel-1.0.1/gptqmodel/integration/optimum/hf_quantizer_gptq.py +88 -0
  20. gptqmodel-1.0.1/gptqmodel/integration/optimum/quantizer.py +883 -0
  21. gptqmodel-1.0.1/gptqmodel/integration/optimum/utils.py +153 -0
  22. gptqmodel-1.0.1/gptqmodel/models/__init__.py +37 -0
  23. gptqmodel-1.0.1/gptqmodel/models/_const.py +64 -0
  24. gptqmodel-1.0.1/gptqmodel/models/auto.py +143 -0
  25. gptqmodel-1.0.1/gptqmodel/models/baichuan.py +16 -0
  26. gptqmodel-1.0.1/gptqmodel/models/base.py +1495 -0
  27. gptqmodel-1.0.1/gptqmodel/models/bloom.py +20 -0
  28. gptqmodel-1.0.1/gptqmodel/models/chatglm.py +14 -0
  29. gptqmodel-1.0.1/gptqmodel/models/codegen.py +14 -0
  30. gptqmodel-1.0.1/gptqmodel/models/cohere.py +14 -0
  31. gptqmodel-1.0.1/gptqmodel/models/dbrx.py +7 -0
  32. gptqmodel-1.0.1/gptqmodel/models/dbrx_converted.py +51 -0
  33. gptqmodel-1.0.1/gptqmodel/models/decilm.py +14 -0
  34. gptqmodel-1.0.1/gptqmodel/models/deepseek_v2.py +42 -0
  35. gptqmodel-1.0.1/gptqmodel/models/exaone.py +17 -0
  36. gptqmodel-1.0.1/gptqmodel/models/gemma.py +14 -0
  37. gptqmodel-1.0.1/gptqmodel/models/gemma2.py +55 -0
  38. gptqmodel-1.0.1/gptqmodel/models/gpt2.py +14 -0
  39. gptqmodel-1.0.1/gptqmodel/models/gpt_bigcode.py +14 -0
  40. gptqmodel-1.0.1/gptqmodel/models/gpt_neox.py +16 -0
  41. gptqmodel-1.0.1/gptqmodel/models/gptj.py +14 -0
  42. gptqmodel-1.0.1/gptqmodel/models/internlm.py +14 -0
  43. gptqmodel-1.0.1/gptqmodel/models/internlm2.py +14 -0
  44. gptqmodel-1.0.1/gptqmodel/models/llama.py +22 -0
  45. gptqmodel-1.0.1/gptqmodel/models/longllama.py +14 -0
  46. gptqmodel-1.0.1/gptqmodel/models/minicpm.py +16 -0
  47. gptqmodel-1.0.1/gptqmodel/models/mistral.py +14 -0
  48. gptqmodel-1.0.1/gptqmodel/models/mixtral.py +40 -0
  49. gptqmodel-1.0.1/gptqmodel/models/moss.py +14 -0
  50. gptqmodel-1.0.1/gptqmodel/models/mpt.py +14 -0
  51. gptqmodel-1.0.1/gptqmodel/models/opt.py +20 -0
  52. gptqmodel-1.0.1/gptqmodel/models/phi.py +16 -0
  53. gptqmodel-1.0.1/gptqmodel/models/phi3.py +14 -0
  54. gptqmodel-1.0.1/gptqmodel/models/qwen.py +19 -0
  55. gptqmodel-1.0.1/gptqmodel/models/qwen2.py +14 -0
  56. gptqmodel-1.0.1/gptqmodel/models/qwen2_moe.py +23 -0
  57. gptqmodel-1.0.1/gptqmodel/models/rw.py +14 -0
  58. gptqmodel-1.0.1/gptqmodel/models/stablelmepoch.py +14 -0
  59. gptqmodel-1.0.1/gptqmodel/models/starcoder2.py +14 -0
  60. gptqmodel-1.0.1/gptqmodel/models/xverse.py +14 -0
  61. gptqmodel-1.0.1/gptqmodel/models/yi.py +14 -0
  62. gptqmodel-1.0.1/gptqmodel/nn_modules/__init__.py +0 -0
  63. gptqmodel-1.0.1/gptqmodel/nn_modules/qlinear/__init__.py +94 -0
  64. gptqmodel-1.0.1/gptqmodel/nn_modules/qlinear/bitblas_target_detector.py +106 -0
  65. gptqmodel-1.0.1/gptqmodel/nn_modules/qlinear/qlinear_bitblas.py +379 -0
  66. gptqmodel-1.0.1/gptqmodel/nn_modules/qlinear/qlinear_exllama.py +192 -0
  67. gptqmodel-1.0.1/gptqmodel/nn_modules/qlinear/qlinear_exllamav2.py +236 -0
  68. gptqmodel-1.0.1/gptqmodel/nn_modules/qlinear/qlinear_marlin.py +268 -0
  69. gptqmodel-1.0.1/gptqmodel/nn_modules/qlinear/qlinear_marlin_inference.py +315 -0
  70. gptqmodel-1.0.1/gptqmodel/nn_modules/qlinear/qlinear_qbits.py +334 -0
  71. gptqmodel-1.0.1/gptqmodel/nn_modules/qlinear/qlinear_tritonv2.py +139 -0
  72. gptqmodel-1.0.1/gptqmodel/nn_modules/triton_utils/__init__.py +0 -0
  73. gptqmodel-1.0.1/gptqmodel/nn_modules/triton_utils/custom_autotune.py +209 -0
  74. gptqmodel-1.0.1/gptqmodel/nn_modules/triton_utils/dequant.py +138 -0
  75. gptqmodel-1.0.1/gptqmodel/nn_modules/triton_utils/kernels.py +461 -0
  76. gptqmodel-1.0.1/gptqmodel/nn_modules/triton_utils/mixin.py +4 -0
  77. gptqmodel-1.0.1/gptqmodel/quantization/__init__.py +4 -0
  78. gptqmodel-1.0.1/gptqmodel/quantization/config.py +387 -0
  79. gptqmodel-1.0.1/gptqmodel/quantization/gptq.py +224 -0
  80. gptqmodel-1.0.1/gptqmodel/quantization/quantizer.py +142 -0
  81. gptqmodel-1.0.1/gptqmodel/utils/__init__.py +2 -0
  82. gptqmodel-1.0.1/gptqmodel/utils/backend.py +21 -0
  83. gptqmodel-1.0.1/gptqmodel/utils/bitblas.py +137 -0
  84. gptqmodel-1.0.1/gptqmodel/utils/data.py +260 -0
  85. gptqmodel-1.0.1/gptqmodel/utils/device.py +13 -0
  86. gptqmodel-1.0.1/gptqmodel/utils/exllama.py +79 -0
  87. gptqmodel-1.0.1/gptqmodel/utils/importer.py +75 -0
  88. gptqmodel-1.0.1/gptqmodel/utils/marlin.py +204 -0
  89. gptqmodel-1.0.1/gptqmodel/utils/model.py +696 -0
  90. gptqmodel-1.0.1/gptqmodel/utils/perplexity.py +227 -0
  91. gptqmodel-1.0.1/gptqmodel/utils/sglang.py +56 -0
  92. gptqmodel-1.0.1/gptqmodel/utils/vllm.py +67 -0
  93. gptqmodel-1.0.1/gptqmodel/version.py +1 -0
  94. gptqmodel-1.0.1/gptqmodel.egg-info/PKG-INFO +320 -0
  95. gptqmodel-1.0.1/gptqmodel.egg-info/SOURCES.txt +145 -0
  96. gptqmodel-1.0.1/gptqmodel.egg-info/dependency_links.txt +1 -0
  97. gptqmodel-1.0.1/gptqmodel.egg-info/requires.txt +41 -0
  98. gptqmodel-1.0.1/gptqmodel.egg-info/top_level.txt +5 -0
  99. gptqmodel-1.0.1/gptqmodel_ext/exllama/cu_compat.cuh +58 -0
  100. gptqmodel-1.0.1/gptqmodel_ext/exllama/cuda_buffers.cu +75 -0
  101. gptqmodel-1.0.1/gptqmodel_ext/exllama/cuda_buffers.cuh +55 -0
  102. gptqmodel-1.0.1/gptqmodel_ext/exllama/cuda_func/column_remap.cu +63 -0
  103. gptqmodel-1.0.1/gptqmodel_ext/exllama/cuda_func/column_remap.cuh +19 -0
  104. gptqmodel-1.0.1/gptqmodel_ext/exllama/cuda_func/q4_matmul.cu +260 -0
  105. gptqmodel-1.0.1/gptqmodel_ext/exllama/cuda_func/q4_matmul.cuh +43 -0
  106. gptqmodel-1.0.1/gptqmodel_ext/exllama/cuda_func/q4_matrix.cu +225 -0
  107. gptqmodel-1.0.1/gptqmodel_ext/exllama/cuda_func/q4_matrix.cuh +53 -0
  108. gptqmodel-1.0.1/gptqmodel_ext/exllama/exllama_ext.cpp +260 -0
  109. gptqmodel-1.0.1/gptqmodel_ext/exllama/hip_compat.cuh +53 -0
  110. gptqmodel-1.0.1/gptqmodel_ext/exllama/matrix.cuh +294 -0
  111. gptqmodel-1.0.1/gptqmodel_ext/exllama/tuning.h +13 -0
  112. gptqmodel-1.0.1/gptqmodel_ext/exllama/util.cuh +33 -0
  113. gptqmodel-1.0.1/gptqmodel_ext/exllamav2/config.h +13 -0
  114. gptqmodel-1.0.1/gptqmodel_ext/exllamav2/cpp/util.h +12 -0
  115. gptqmodel-1.0.1/gptqmodel_ext/exllamav2/cuda/compat.cuh +56 -0
  116. gptqmodel-1.0.1/gptqmodel_ext/exllamav2/cuda/compat_gemm.cuh +38 -0
  117. gptqmodel-1.0.1/gptqmodel_ext/exllamav2/cuda/matrix_view.cuh +121 -0
  118. gptqmodel-1.0.1/gptqmodel_ext/exllamav2/cuda/q_gemm.cu +211 -0
  119. gptqmodel-1.0.1/gptqmodel_ext/exllamav2/cuda/q_gemm.cuh +33 -0
  120. gptqmodel-1.0.1/gptqmodel_ext/exllamav2/cuda/q_gemm_kernel.cuh +487 -0
  121. gptqmodel-1.0.1/gptqmodel_ext/exllamav2/cuda/q_gemm_kernel_gptq.cuh +223 -0
  122. gptqmodel-1.0.1/gptqmodel_ext/exllamav2/cuda/q_matrix.cu +627 -0
  123. gptqmodel-1.0.1/gptqmodel_ext/exllamav2/cuda/q_matrix.cuh +73 -0
  124. gptqmodel-1.0.1/gptqmodel_ext/exllamav2/cuda/quant/qdq_2.cuh +103 -0
  125. gptqmodel-1.0.1/gptqmodel_ext/exllamav2/cuda/quant/qdq_3.cuh +169 -0
  126. gptqmodel-1.0.1/gptqmodel_ext/exllamav2/cuda/quant/qdq_4.cuh +227 -0
  127. gptqmodel-1.0.1/gptqmodel_ext/exllamav2/cuda/quant/qdq_5.cuh +207 -0
  128. gptqmodel-1.0.1/gptqmodel_ext/exllamav2/cuda/quant/qdq_6.cuh +44 -0
  129. gptqmodel-1.0.1/gptqmodel_ext/exllamav2/cuda/quant/qdq_8.cuh +38 -0
  130. gptqmodel-1.0.1/gptqmodel_ext/exllamav2/cuda/quant/qdq_util.cuh +51 -0
  131. gptqmodel-1.0.1/gptqmodel_ext/exllamav2/cuda/util.cuh +42 -0
  132. gptqmodel-1.0.1/gptqmodel_ext/exllamav2/ext.cpp +134 -0
  133. gptqmodel-1.0.1/gptqmodel_ext/marlin/marlin_cuda.cpp +80 -0
  134. gptqmodel-1.0.1/gptqmodel_ext/marlin/marlin_cuda_kernel.cu +849 -0
  135. gptqmodel-1.0.1/gptqmodel_ext/marlin/marlin_cuda_kernel.cuh +20 -0
  136. gptqmodel-1.0.1/gptqmodel_ext/marlin/marlin_repack.cu +93 -0
  137. gptqmodel-1.0.1/gptqmodel_ext/marlin/marlin_repack.cuh +12 -0
  138. gptqmodel-1.0.1/gptqmodel_ext/marlin_inference/marlin.cuh +90 -0
  139. gptqmodel-1.0.1/gptqmodel_ext/marlin_inference/marlin_cuda.cpp +20 -0
  140. gptqmodel-1.0.1/gptqmodel_ext/marlin_inference/marlin_cuda_kernel.cu +2185 -0
  141. gptqmodel-1.0.1/gptqmodel_ext/marlin_inference/marlin_cuda_kernel.cuh +8 -0
  142. gptqmodel-1.0.1/gptqmodel_ext/marlin_inference/marlin_dtypes.cuh +81 -0
  143. gptqmodel-1.0.1/gptqmodel_ext/marlin_inference/marlin_repack.cu +347 -0
  144. gptqmodel-1.0.1/gptqmodel_ext/marlin_inference/marlin_repack.cuh +16 -0
  145. gptqmodel-1.0.1/requirements.txt +19 -0
  146. gptqmodel-1.0.1/setup.cfg +4 -0
  147. gptqmodel-1.0.1/setup.py +186 -0
@@ -0,0 +1,201 @@
1
+ Apache License
2
+ Version 2.0, January 2004
3
+ http://www.apache.org/licenses/
4
+
5
+ TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
6
+
7
+ 1. Definitions.
8
+
9
+ "License" shall mean the terms and conditions for use, reproduction,
10
+ and distribution as defined by Sections 1 through 9 of this document.
11
+
12
+ "Licensor" shall mean the copyright owner or entity authorized by
13
+ the copyright owner that is granting the License.
14
+
15
+ "Legal Entity" shall mean the union of the acting entity and all
16
+ other entities that control, are controlled by, or are under common
17
+ control with that entity. For the purposes of this definition,
18
+ "control" means (i) the power, direct or indirect, to cause the
19
+ direction or management of such entity, whether by contract or
20
+ otherwise, or (ii) ownership of fifty percent (50%) or more of the
21
+ outstanding shares, or (iii) beneficial ownership of such entity.
22
+
23
+ "You" (or "Your") shall mean an individual or Legal Entity
24
+ exercising permissions granted by this License.
25
+
26
+ "Source" form shall mean the preferred form for making modifications,
27
+ including but not limited to software source code, documentation
28
+ source, and configuration files.
29
+
30
+ "Object" form shall mean any form resulting from mechanical
31
+ transformation or translation of a Source form, including but
32
+ not limited to compiled object code, generated documentation,
33
+ and conversions to other media types.
34
+
35
+ "Work" shall mean the work of authorship, whether in Source or
36
+ Object form, made available under the License, as indicated by a
37
+ copyright notice that is included in or attached to the work
38
+ (an example is provided in the Appendix below).
39
+
40
+ "Derivative Works" shall mean any work, whether in Source or Object
41
+ form, that is based on (or derived from) the Work and for which the
42
+ editorial revisions, annotations, elaborations, or other modifications
43
+ represent, as a whole, an original work of authorship. For the purposes
44
+ of this License, Derivative Works shall not include works that remain
45
+ separable from, or merely link (or bind by name) to the interfaces of,
46
+ the Work and Derivative Works thereof.
47
+
48
+ "Contribution" shall mean any work of authorship, including
49
+ the original version of the Work and any modifications or additions
50
+ to that Work or Derivative Works thereof, that is intentionally
51
+ submitted to Licensor for inclusion in the Work by the copyright owner
52
+ or by an individual or Legal Entity authorized to submit on behalf of
53
+ the copyright owner. For the purposes of this definition, "submitted"
54
+ means any form of electronic, verbal, or written communication sent
55
+ to the Licensor or its representatives, including but not limited to
56
+ communication on electronic mailing lists, source code control systems,
57
+ and issue tracking systems that are managed by, or on behalf of, the
58
+ Licensor for the purpose of discussing and improving the Work, but
59
+ excluding communication that is conspicuously marked or otherwise
60
+ designated in writing by the copyright owner as "Not a Contribution."
61
+
62
+ "Contributor" shall mean Licensor and any individual or Legal Entity
63
+ on behalf of whom a Contribution has been received by Licensor and
64
+ subsequently incorporated within the Work.
65
+
66
+ 2. Grant of Copyright License. Subject to the terms and conditions of
67
+ this License, each Contributor hereby grants to You a perpetual,
68
+ worldwide, non-exclusive, no-charge, royalty-free, irrevocable
69
+ copyright license to reproduce, prepare Derivative Works of,
70
+ publicly display, publicly perform, sublicense, and distribute the
71
+ Work and such Derivative Works in Source or Object form.
72
+
73
+ 3. Grant of Patent License. Subject to the terms and conditions of
74
+ this License, each Contributor hereby grants to You a perpetual,
75
+ worldwide, non-exclusive, no-charge, royalty-free, irrevocable
76
+ (except as stated in this section) patent license to make, have made,
77
+ use, offer to sell, sell, import, and otherwise transfer the Work,
78
+ where such license applies only to those patent claims licensable
79
+ by such Contributor that are necessarily infringed by their
80
+ Contribution(s) alone or by combination of their Contribution(s)
81
+ with the Work to which such Contribution(s) was submitted. If You
82
+ institute patent litigation against any entity (including a
83
+ cross-claim or counterclaim in a lawsuit) alleging that the Work
84
+ or a Contribution incorporated within the Work constitutes direct
85
+ or contributory patent infringement, then any patent licenses
86
+ granted to You under this License for that Work shall terminate
87
+ as of the date such litigation is filed.
88
+
89
+ 4. Redistribution. You may reproduce and distribute copies of the
90
+ Work or Derivative Works thereof in any medium, with or without
91
+ modifications, and in Source or Object form, provided that You
92
+ meet the following conditions:
93
+
94
+ (a) You must give any other recipients of the Work or
95
+ Derivative Works a copy of this License; and
96
+
97
+ (b) You must cause any modified files to carry prominent notices
98
+ stating that You changed the files; and
99
+
100
+ (c) You must retain, in the Source form of any Derivative Works
101
+ that You distribute, all copyright, patent, trademark, and
102
+ attribution notices from the Source form of the Work,
103
+ excluding those notices that do not pertain to any part of
104
+ the Derivative Works; and
105
+
106
+ (d) If the Work includes a "NOTICE" text file as part of its
107
+ distribution, then any Derivative Works that You distribute must
108
+ include a readable copy of the attribution notices contained
109
+ within such NOTICE file, excluding those notices that do not
110
+ pertain to any part of the Derivative Works, in at least one
111
+ of the following places: within a NOTICE text file distributed
112
+ as part of the Derivative Works; within the Source form or
113
+ documentation, if provided along with the Derivative Works; or,
114
+ within a display generated by the Derivative Works, if and
115
+ wherever such third-party notices normally appear. The contents
116
+ of the NOTICE file are for informational purposes only and
117
+ do not modify the License. You may add Your own attribution
118
+ notices within Derivative Works that You distribute, alongside
119
+ or as an addendum to the NOTICE text from the Work, provided
120
+ that such additional attribution notices cannot be construed
121
+ as modifying the License.
122
+
123
+ You may add Your own copyright statement to Your modifications and
124
+ may provide additional or different license terms and conditions
125
+ for use, reproduction, or distribution of Your modifications, or
126
+ for any such Derivative Works as a whole, provided Your use,
127
+ reproduction, and distribution of the Work otherwise complies with
128
+ the conditions stated in this License.
129
+
130
+ 5. Submission of Contributions. Unless You explicitly state otherwise,
131
+ any Contribution intentionally submitted for inclusion in the Work
132
+ by You to the Licensor shall be under the terms and conditions of
133
+ this License, without any additional terms or conditions.
134
+ Notwithstanding the above, nothing herein shall supersede or modify
135
+ the terms of any separate license agreement you may have executed
136
+ with Licensor regarding such Contributions.
137
+
138
+ 6. Trademarks. This License does not grant permission to use the trade
139
+ names, trademarks, service marks, or product names of the Licensor,
140
+ except as required for reasonable and customary use in describing the
141
+ origin of the Work and reproducing the content of the NOTICE file.
142
+
143
+ 7. Disclaimer of Warranty. Unless required by applicable law or
144
+ agreed to in writing, Licensor provides the Work (and each
145
+ Contributor provides its Contributions) on an "AS IS" BASIS,
146
+ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
147
+ implied, including, without limitation, any warranties or conditions
148
+ of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
149
+ PARTICULAR PURPOSE. You are solely responsible for determining the
150
+ appropriateness of using or redistributing the Work and assume any
151
+ risks associated with Your exercise of permissions under this License.
152
+
153
+ 8. Limitation of Liability. In no event and under no legal theory,
154
+ whether in tort (including negligence), contract, or otherwise,
155
+ unless required by applicable law (such as deliberate and grossly
156
+ negligent acts) or agreed to in writing, shall any Contributor be
157
+ liable to You for damages, including any direct, indirect, special,
158
+ incidental, or consequential damages of any character arising as a
159
+ result of this License or out of the use or inability to use the
160
+ Work (including but not limited to damages for loss of goodwill,
161
+ work stoppage, computer failure or malfunction, or any and all
162
+ other commercial damages or losses), even if such Contributor
163
+ has been advised of the possibility of such damages.
164
+
165
+ 9. Accepting Warranty or Additional Liability. While redistributing
166
+ the Work or Derivative Works thereof, You may choose to offer,
167
+ and charge a fee for, acceptance of support, warranty, indemnity,
168
+ or other liability obligations and/or rights consistent with this
169
+ License. However, in accepting such obligations, You may act only
170
+ on Your own behalf and on Your sole responsibility, not on behalf
171
+ of any other Contributor, and only if You agree to indemnify,
172
+ defend, and hold each Contributor harmless for any liability
173
+ incurred by, or claims asserted against, such Contributor by reason
174
+ of your accepting any such warranty or additional liability.
175
+
176
+ END OF TERMS AND CONDITIONS
177
+
178
+ APPENDIX: How to apply the Apache License to your work.
179
+
180
+ To apply the Apache License to your work, attach the following
181
+ boilerplate notice, with the fields enclosed by brackets "[]"
182
+ replaced with your own identifying information. (Don't include
183
+ the brackets!) The text should be enclosed in the appropriate
184
+ comment syntax for the file format. We also recommend that a
185
+ file or class name and description of purpose be included on the
186
+ same "printed page" as the copyright notice for easier
187
+ identification within third-party archives.
188
+
189
+ Copyright [yyyy] [name of copyright owner]
190
+
191
+ Licensed under the Apache License, Version 2.0 (the "License");
192
+ you may not use this file except in compliance with the License.
193
+ You may obtain a copy of the License at
194
+
195
+ http://www.apache.org/licenses/LICENSE-2.0
196
+
197
+ Unless required by applicable law or agreed to in writing, software
198
+ distributed under the License is distributed on an "AS IS" BASIS,
199
+ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
200
+ See the License for the specific language governing permissions and
201
+ limitations under the License.
@@ -0,0 +1,6 @@
1
+ global-include gptqmodel_ext/**/*.cuh
2
+ global-include gptqmodel_ext/**/*.h
3
+ global-include gptqmodel_ext/**/*.cpp
4
+ global-include gptqmodel_ext/**/*.cu
5
+ global-include gptqmodel_ext/**/*.py
6
+ include requirements.txt
@@ -0,0 +1,320 @@
1
+ Metadata-Version: 2.1
2
+ Name: gptqmodel
3
+ Version: 1.0.1
4
+ Summary: A LLM quantization package with user-friendly apis. Based on GPTQ algorithm.
5
+ Home-page: https://github.com/ModelCloud/GPTQModel
6
+ Author: ModelCloud
7
+ Author-email: qubitium@modelcloud.ai
8
+ Keywords: gptq,quantization,large-language-models,transformers,4bit,llm
9
+ Platform: linux
10
+ Classifier: License :: OSI Approved :: Apache Software License
11
+ Classifier: Programming Language :: Python :: 3
12
+ Classifier: Programming Language :: Python :: 3.9
13
+ Classifier: Programming Language :: Python :: 3.10
14
+ Classifier: Programming Language :: Python :: 3.11
15
+ Classifier: Programming Language :: C++
16
+ Classifier: Intended Audience :: Developers
17
+ Classifier: Intended Audience :: Education
18
+ Classifier: Intended Audience :: Science/Research
19
+ Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
20
+ Requires-Python: >=3.8.0
21
+ Description-Content-Type: text/markdown
22
+ Provides-Extra: test
23
+ Provides-Extra: quality
24
+ Provides-Extra: vllm
25
+ Provides-Extra: sglang
26
+ Provides-Extra: bitblas
27
+ Provides-Extra: hf
28
+ License-File: LICENSE
29
+
30
+ <h1 align="center">GPTQModel</h1>
31
+ <p align="center">An easy-to-use LLM quantization and inference toolkit based on GPTQ algorithm (weight-only quantization).</p>
32
+ <p align="center">
33
+ <a href="https://github.com/ModelCloud/GPTQModel/releases">
34
+ <img alt="GitHub release" src="https://img.shields.io/github/release/ModelCloud/GPTQModel.svg">
35
+ </a>
36
+ <a href="https://pypi.org/project/model-gptq/">
37
+ <img alt="PyPI - Downloads" src="https://img.shields.io/pypi/dd/gptq-model">
38
+ </a>
39
+ </p>
40
+
41
+ ## News
42
+ * 08/14/2024 ✨✨ [v1.0.0](https://github.com/ModelCloud/GPTQModel/releases/tag/v1.0.0) 40% faster `packing`, Fixed Python 3.9 compat, added `lm_eval` api.
43
+ * 08/10/2024 🚀🚀🚀 [v0.9.11](https://github.com/ModelCloud/GPTQModel/releases/tag/v0.9.11) Added LG EXAONE 3.0 model support. New `dynamic` per layer/module flexible quantization where each layer/module may have different bits/params. Added proper sharding support to `backend.BITBLAS`. Auto-heal quantization errors due to small damp values.
44
+ * 07/31/2024 🚀🚀 [v0.9.10](https://github.com/ModelCloud/GPTQModel/releases/tag/v0.9.10) Ported vllm/nm `gptq_marlin` inference kernel with expanded bits (8bits), group_size (64,32), and desc_act support for all GPTQ models with `FORMAT.GPTQ`. Auto calculate auto-round nsamples/seglen parameters based on calibration dataset. Fixed save_quantized() called on pre-quantized models with non-supported backends. HF transformers depend updated to ensure Llama 3.1 fixes are correctly applied to both quant and inference.
45
+ * 07/25/2024 🚀 [v0.9.9](https://github.com/ModelCloud/GPTQModel/releases/tag/v0.9.9): Added Llama-3.1 support, Gemma2 27B quant inference support via vLLM, auto pad_token normalization, fixed auto-round quant compat for vLLM/SGLang, and more.
46
+ * 07/13/2024 🚀🚀🚀 [v0.9.8](https://github.com/ModelCloud/GPTQModel/releases/tag/v0.9.8):
47
+ Run quantized models directly using GPTQModel using fast `vLLM` or `SGLang` backend! Both vLLM and SGLang are optimized for dyanamic batching inference for maximum `TPS` (check usage under examples). Marlin backend also
48
+ got full end-to-end in/out features padding to enhance current/future model compatibility.
49
+ <details>
50
+
51
+ <summary>Archived News:</summary>
52
+
53
+ * 07/08/2024 🚀 [v0.9.7](https://github.com/ModelCloud/GPTQModel/releases/tag/v0.9.7): InternLM 2.5 model support added.
54
+
55
+ * 07/08/2024 🚀🚀 [v0.9.6](https://github.com/ModelCloud/GPTQModel/releases/tag/v0.9.6): [Intel/AutoRound](https://github.com/intel/auto-round) QUANT_METHOD support added for a potentially higher quality quantization with `lm_head` module quantization support for even more vram reduction: format export to `FORMAT.GPTQ` for max inference compatibility.
56
+
57
+ * 07/05/2024 🚀🚀 [v0.9.5](https://github.com/ModelCloud/GPTQModel/releases/tag/v0.9.5): [Intel/QBits](https://github.com/intel/intel-extension-for-transformers) support added for [2,3,4,8] bit quantization/inference on CPU. Cuda kernels have been fully deprecated in favor of Exllama(v1/v2)/Marlin/Triton.
58
+
59
+ * 07/03/2024 🚀 [v0.9.4](https://github.com/ModelCloud/GPTQModel/releases/tag/v0.9.4): HF Transformers integration added and bug fixed Gemma 2 support.
60
+
61
+ * 07/02/2024 🚀 [v0.9.3](https://github.com/ModelCloud/GPTQModel/releases/tag/v0.9.3): Added Gemma 2 support, faster PPL calculations on gpu, and more code/arg refractor.
62
+
63
+ * 06/30/2024 🚀 [v0.9.2](https://github.com/ModelCloud/GPTQModel/releases/tag/v0.9.2): Added auto-padding of model in/out-features for exllama and exllama v2.
64
+ Fixed quantization of OPT and DeepSeek V2-Lite models. Fixed inference for DeepSeek V2-Lite.
65
+
66
+ * 06/29/2024 🚀🚀🚀 [v0.9.1](https://github.com/ModelCloud/GPTQModel/releases/tag/v0.9.1): With 3 new models (DeepSeek-V2, DeepSeek-V2-Lite, DBRX Converted), BITBLAS new format/kernel, proper batching of calibration dataset resulting > 50% quantization speedup, security hash check of loaded model weights, tons of refractor/usability improvements, bugs fixes and much more.
67
+
68
+ * 06/20/2924 ✨ GPTQModel [v0.9.0](https://github.com/ModelCloud/GPTQModel/releases/tag/v0.9.0): Thanks for all the work from ModelCloud team and the opensource ML community for their contributions!
69
+ </details>
70
+
71
+ ## Mission Statement
72
+
73
+ We want GPTQModel to be highly focused on GPTQ based quantization and target inference compatibility with HF Transformers, vLLM, and SGLang.
74
+
75
+ ## How is GPTQModel different from AutoGPTQ?
76
+
77
+ GPTQModel is an opinionated fork/refactor of AutoGPTQ with latest bug fixes, more model support, faster quant inference, faster quantization, better quants (as measured in PPL) and a pledge from the ModelCloud team and that we, along with the open-source ML community, will take every effort to bring the library up-to-date with latest advancements, model support, and bug fixes.
78
+
79
+ We will backport bug fixes to AutoGPTQ on a case-by-case basis.
80
+
81
+ ## Major Changes (Advantages) vs AutoGPTQ
82
+ * 🚀🚀🚀🚀 Extensive model support for: `EXAONE 3.0`, `InternLM 2.5`, `Gemma 2`, `DeepSeek-V2`, `DeepSeek-V2-Lite`, `ChatGLM`, `MiniCPM`, `Phi-3`, `Qwen2MoE`, `DBRX` (Converted).
83
+ * 🚀🚀 vLLM inference integration for quantized model where format = `FORMAT.GPTQ`
84
+ * 🚀🚀 SGLang inference integration for quantized model where format = `FORMAT.GPTQ`
85
+ * 🚀 [Intel/AutoRound](https://github.com/intel/auto-round) QUANT_METHOD support added for a potentially higher quality quantization with `lm_head` module quantization support for even more vram reduction: format export to `FORMAT.GPTQ` for max inference compatibility.
86
+ * 🚀 [Intel/QBits](https://github.com/intel/intel-extension-for-transformers) support added for [2,3,4,8] bit quantization/inference on CPU.
87
+ * 🚀 [BITBLAS](https://github.com/microsoft/BitBLAS) format/inference support from Microsoft
88
+ * 🚀`Sym=False` Support. AutoGPTQ has unusable `sym=false`. (Re-quant required)
89
+ * 🚀`lm_head` module quant inference support for further VRAM reduction.
90
+ * 🚀 Faster quantization: More than 50% faster for TinyLlama + 4090 with batching and large calibration dataset.
91
+ * 🚀 Better quality quants as measured by PPL. (Test config: defaults + `sym=True` + `FORMAT.GPTQ`, TinyLlama)
92
+ * 🚀 Model weights sharding support
93
+ * 🚀 Security: hash check of model weights on load
94
+ * 🚀 Over 50% faster PPL calculations (OPT)
95
+ * 🚀 Over 40% faster `packing` stage in quantization (Llama 3.1 8B)
96
+ * ✨ Alert users of sub-optimal calibration data. Most new users get this part horribly wrong.
97
+ * ✨ Increased compatibility with newest models with auto-padding of in/out-features for [ Exllama, Exllama V2 ] backends.
98
+ * 👾 Removed non-working, partially working, or fully deprecated features: Peft, ROCM, AWQ Gemm inference, Triton v1 (replaced by v2), Fused Attention (Replaced by Marlin/Exllama).
99
+ * 👾 <del>Fixed packing Performance regression on high core-count systems.</del> Backported to AutoGPTQ
100
+ * 👾 <del>Fixed crash on H100.</del> Backported to AutoGPTQ
101
+ * ✨ 10s of thousands of lines of refactor/cleanup.
102
+ * ✨ Over 8+ overly complex api args removed/merged into simple human-readable args.
103
+ * ✨ Added CI workflow for validation of future PRs and prevent code regressions.
104
+ * ✨ Added perplexity unit-test to prevent against model quant quality regressions.
105
+ * 👾 De-bloated 271K lines of which 250K was caused by a single dataset used only by an example.
106
+ * 👾 De-bloat the number of args presented in public `.from_quantized()`/`.from_pretrained()` api
107
+ * ✨ Shorter and more concise public api/internal vars. No need to mimic HF style for verbose class names.
108
+ * ✨ Everything that did not pass unit-tests have been removed from repo.
109
+
110
+ ## Model Support ( 🚀 GPTQModel only )
111
+ [Ready to deply quantized models](https://hf.co/ModelCloud)
112
+
113
+ | Model | | | | | | | |
114
+ |------------------|----|-----------------------|----|-----------|----|------------|----|
115
+ | Baichuan | ✅ | EXAONE 3.0 | 🚀 | LongLLaMA | ✅ | Qwen2MoE | 🚀 |
116
+ | Bloom | ✅ | Falon | ✅ | MiniCPM | 🚀 | RefinedWeb | ✅ |
117
+ | ChatGLM | 🚀 | Gemma 2 | 🚀 | Mistral | ✅ | StableLM | ✅ |
118
+ | CodeGen | ✅ | GPTBigCod | ✅ | Mixtral | ✅ | StarCoder2 | ✅ |
119
+ | Cohere | ✅ | GPTNeoX | ✅ | MOSS | ✅ | XVERSE | ✅ |
120
+ | DBRX Converted | 🚀 | GPT-2 | ✅ | MPT | ✅ | Yi | ✅ |
121
+ | Deci | ✅ | GPT-J | ✅ | OPT | ✅ | | |
122
+ | DeepSeek-V2 | 🚀 | InternLM/InternLM 2.5 | 🚀 | Phi/Phi-3 | 🚀 | | |
123
+ | DeepSeek-V2-Lite | 🚀 | Llama | ✅ | Qwen | ✅ | | |
124
+
125
+ ## Compatiblity
126
+
127
+ We aim for 100% compatibility with models quanted by AutoGPTQ <= 0.7.1 and will consider syncing future compatibilty on a case-by-case basis.
128
+
129
+ ## Platform/GPU Requirements
130
+
131
+ GPTQModel is currently Linux only and requires CUDA capability >= 6.0 Nvidia GPU.
132
+
133
+ WSL on Windows should work as well.
134
+
135
+ ROCM/AMD support will be re-added in a future version after everything on ROCM has been validated. Only fully validated features will be re-added from the original AutoGPTQ repo.
136
+
137
+ ## Install
138
+
139
+ ### Install from source
140
+
141
+ ```bash
142
+ # clone repo
143
+ git clone https://github.com/ModelCloud/GPTQModel.git && cd GPTQModel
144
+
145
+ # compile and install
146
+ # You can optionally include specific modules like vllm, sglang, or bitblas by adding them in brackets. Example: pip install -vvv --no-build-isolation .[vllm,sglang,bitblas]
147
+ pip install -vvv --no-build-isolation .
148
+
149
+ # If you have `uv` package version 0.1.16 or higher, you can use `uv pip` for potentially better dependency management
150
+ # Include modules as needed: uv pip install -vvv --no-build-isolation .[vllm,sglang,bitblas]
151
+ uv pip install -vvv --no-build-isolation .
152
+ ```
153
+
154
+ ### Script installation
155
+ ```bash
156
+ # You can pass modules as arguments, e.g., --vllm --sglang --bitblas. Example: bash install.sh --vllm --sglang --bitblas
157
+ bash install.sh
158
+ ```
159
+
160
+ ### PIP (PENDING RELEASE)
161
+
162
+ ```bash
163
+ # Include any specific modules needed using brackets. Example: pip install gptqmodel[sglang,vllm,bitblas] --no-build-isolation
164
+ pip install gptqmodel --no-build-isolation
165
+ ```
166
+
167
+ ### Quantization and Inference
168
+
169
+ > warning: this is just a showcase of the usage of basic apis in GPTQModel, which uses only one sample to quantize a much small model, quality of quantized model using such little samples may not good.
170
+
171
+ Below is an example for the simplest use of `gptqmodel` to quantize a model and inference after quantization:
172
+
173
+ ```py
174
+ from transformers import AutoTokenizer
175
+ from gptqmodel import GPTQModel, QuantizeConfig
176
+
177
+ pretrained_model_dir = "facebook/opt-125m"
178
+ quant_output_dir = "opt-125m-4bit"
179
+
180
+ tokenizer = AutoTokenizer.from_pretrained(pretrained_model_dir, use_fast=True)
181
+ calibration_dataset = [
182
+ tokenizer(
183
+ "The world is a wonderful place full of beauty and love."
184
+ )
185
+ ]
186
+
187
+ quant_config = QuantizeConfig(
188
+ bits=4, # 4-bit
189
+ group_size=128, # 128 is good balance between quality and performance
190
+ )
191
+
192
+ # load un-quantized model, by default, the model will always be loaded into CPU memory
193
+ model = GPTQModel.from_pretrained(pretrained_model_dir, quant_config)
194
+
195
+ # quantize model, the calibration_dataset should be list of dict whose keys can only be "input_ids" and "attention_mask"
196
+ model.quantize(calibration_dataset)
197
+
198
+ # save quantized model
199
+ model.save_quantized(quant_output_dir)
200
+
201
+ # load quantized model to the first GPU
202
+ model = GPTQModel.from_quantized(quant_output_dir)
203
+
204
+ # inference with model.generate
205
+ print(tokenizer.decode(model.generate(**tokenizer("gptqmodel is", return_tensors="pt").to(model.device))[0]))
206
+ ```
207
+
208
+ For more advanced features of model quantization, please reference to [this script](examples/quantization/quant_with_alpaca.py)
209
+
210
+ ### How to Add Support for a New Model
211
+
212
+ Read the [`gptqmodel/models/llama.py`](https://github.com/ModelCloud/GPTQModel/blob/5627f5ffeb3f19b1a2a97e3b6de6fbe668b0dc42/gptqmodel/models/llama.py) code which explains in detail via comments how the model support is defined. Use it as guide to PR for to new models. Most models follow the same pattern.
213
+
214
+ ### Evaluation on Downstream Tasks
215
+
216
+ You can use tasks defined in `gptqmodel.eval_tasks` to evaluate model's performance on specific down-stream task before and after quantization.
217
+
218
+ The predefined tasks support all causal-language-models implemented in [🤗 transformers](https://github.com/huggingface/transformers) and in this project.
219
+
220
+ <details>
221
+
222
+ <summary>Below is an example to evaluate `EleutherAI/gpt-j-6b` on sequence-classification task using `cardiffnlp/tweet_sentiment_multilingual` dataset:</summary>
223
+
224
+ ```python
225
+ from functools import partial
226
+
227
+ import datasets
228
+ from transformers import AutoTokenizer, AutoModelForCausalLM, GenerationConfig
229
+ from gptqmodel import GPTQModel, QuantizeConfig
230
+ from gptqmodel.eval_tasks import SequenceClassificationTask
231
+
232
+ MODEL = "EleutherAI/gpt-j-6b"
233
+ DATASET = "cardiffnlp/tweet_sentiment_multilingual"
234
+ TEMPLATE = "Question:What's the sentiment of the given text? Choices are {labels}.\nText: {text}\nAnswer:"
235
+ ID2LABEL = {
236
+ 0: "negative",
237
+ 1: "neutral",
238
+ 2: "positive"
239
+ }
240
+ LABELS = list(ID2LABEL.values())
241
+
242
+
243
+ def ds_refactor_fn(samples):
244
+ text_data = samples["text"]
245
+ label_data = samples["label"]
246
+
247
+ new_samples = {"prompt": [], "label": []}
248
+ for text, label in zip(text_data, label_data):
249
+ prompt = TEMPLATE.format(labels=LABELS, text=text)
250
+ new_samples["prompt"].append(prompt)
251
+ new_samples["label"].append(ID2LABEL[label])
252
+
253
+ return new_samples
254
+
255
+
256
+ # model = AutoModelForCausalLM.from_pretrained(MODEL).eval().half().to("cuda:0")
257
+ model = GPTQModel.from_pretrained(MODEL, QuantizeConfig())
258
+ tokenizer = AutoTokenizer.from_pretrained(MODEL)
259
+
260
+ task = SequenceClassificationTask(
261
+ model=model,
262
+ tokenizer=tokenizer,
263
+ classes=LABELS,
264
+ data_name_or_path=DATASET,
265
+ prompt_col_name="prompt",
266
+ label_col_name="label",
267
+ **{
268
+ "num_samples": 1000, # how many samples will be sampled to evaluation
269
+ "sample_max_len": 1024, # max tokens for each sample
270
+ "block_max_len": 2048, # max tokens for each data block
271
+ # function to load dataset, one must only accept data_name_or_path as input
272
+ # and return datasets.Dataset
273
+ "load_fn": partial(datasets.load_dataset, name="english"),
274
+ # function to preprocess dataset, which is used for datasets.Dataset.map,
275
+ # must return Dict[str, list] with only two keys: [prompt_col_name, label_col_name]
276
+ "preprocess_fn": ds_refactor_fn,
277
+ # truncate label when sample's length exceed sample_max_len
278
+ "truncate_prompt": False
279
+ }
280
+ )
281
+
282
+ # note that max_new_tokens will be automatically specified internally based on given classes
283
+ print(task.run())
284
+
285
+ # self-consistency
286
+ print(
287
+ task.run(
288
+ generation_config=GenerationConfig(
289
+ num_beams=3,
290
+ num_return_sequences=3,
291
+ do_sample=True
292
+ )
293
+ )
294
+ )
295
+ ```
296
+
297
+ </details>
298
+
299
+ ## Learn More
300
+
301
+ [tutorials](docs/tutorial) provide step-by-step guidance to integrate `gptqmodel` with your own project and some best practice principles.
302
+
303
+ [examples](examples/README.md) provide plenty of example scripts to use `gptqmodel` in different ways.
304
+
305
+ ## Supported Evaluation Tasks
306
+
307
+ Currently, `gptqmodel` supports: `LanguageModelingTask`, `SequenceClassificationTask` and `TextSummarizationTask`; more Tasks will come soon!
308
+
309
+ ### Which kernel is used by default?
310
+
311
+ GPTQModel will use Marlin, Exllama v2, Triton kernels in that order for maximum inference performance.
312
+
313
+ # Acknowledgements
314
+
315
+ * **Elias Frantar**, **Saleh Ashkboos**, **Torsten Hoefler** and **Dan Alistarh**: for creating [GPTQ](https://github.com/IST-DASLab/gptq) and [Marlin](https://github.com/IST-DASLab/marlin).
316
+ * **PanQiWei**: for creation of [AutoGPTQ](https://github.com/autogptq/AutoGPTQ) which this project code is based upon.
317
+ * **FXMarty**: for maintaining and support of [AutoGPTQ](https://github.com/autogptq/AutoGPTQ).
318
+ * **Qwopqwop200**: for quantization code used in this project adapted from [GPTQ-for-LLaMa](https://github.com/qwopqwop200/GPTQ-for-LLaMa/tree/cuda).
319
+ * **Turboderp**: for releasing [Exllama v1](https://github.com/turboderp/exllama) and [Exllama v2](https://github.com/turboderp/exllamav2) kernels adapted for use in this project.
320
+ * **FpgaMiner**: for [GPTQ-Triton](https://github.com/fpgaminer/GPTQ-triton) kernels used in [GPTQ-for-LLaMa](https://github.com/qwopqwop200/GPTQ-for-LLaMa/tree/cuda) which is adapted into this project.