Commit ·
8f3e832
0
Parent(s):
Publish MLX 4-bit release
Browse files- .gitattributes +35 -0
- README.md +39 -0
- SHA256SUMS.txt +1 -0
- SPARK_MLX_LLM_LICENSE +201 -0
- assets/benchmark.png +0 -0
- chat_template.jinja +110 -0
- config.json +93 -0
- configuration_spark.py +118 -0
- generation_config.json +14 -0
- model.safetensors +3 -0
- model.safetensors.index.json +732 -0
- modeling_spark.py +483 -0
- tokenizer.json +0 -0
- tokenizer_config.json +14 -0
.gitattributes
ADDED
|
@@ -0,0 +1,35 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
*.7z filter=lfs diff=lfs merge=lfs -text
|
| 2 |
+
*.arrow filter=lfs diff=lfs merge=lfs -text
|
| 3 |
+
*.bin filter=lfs diff=lfs merge=lfs -text
|
| 4 |
+
*.bz2 filter=lfs diff=lfs merge=lfs -text
|
| 5 |
+
*.ckpt filter=lfs diff=lfs merge=lfs -text
|
| 6 |
+
*.ftz filter=lfs diff=lfs merge=lfs -text
|
| 7 |
+
*.gz filter=lfs diff=lfs merge=lfs -text
|
| 8 |
+
*.h5 filter=lfs diff=lfs merge=lfs -text
|
| 9 |
+
*.joblib filter=lfs diff=lfs merge=lfs -text
|
| 10 |
+
*.lfs.* filter=lfs diff=lfs merge=lfs -text
|
| 11 |
+
*.mlmodel filter=lfs diff=lfs merge=lfs -text
|
| 12 |
+
*.model filter=lfs diff=lfs merge=lfs -text
|
| 13 |
+
*.msgpack filter=lfs diff=lfs merge=lfs -text
|
| 14 |
+
*.npy filter=lfs diff=lfs merge=lfs -text
|
| 15 |
+
*.npz filter=lfs diff=lfs merge=lfs -text
|
| 16 |
+
*.onnx filter=lfs diff=lfs merge=lfs -text
|
| 17 |
+
*.ot filter=lfs diff=lfs merge=lfs -text
|
| 18 |
+
*.parquet filter=lfs diff=lfs merge=lfs -text
|
| 19 |
+
*.pb filter=lfs diff=lfs merge=lfs -text
|
| 20 |
+
*.pickle filter=lfs diff=lfs merge=lfs -text
|
| 21 |
+
*.pkl filter=lfs diff=lfs merge=lfs -text
|
| 22 |
+
*.pt filter=lfs diff=lfs merge=lfs -text
|
| 23 |
+
*.pth filter=lfs diff=lfs merge=lfs -text
|
| 24 |
+
*.rar filter=lfs diff=lfs merge=lfs -text
|
| 25 |
+
*.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 26 |
+
saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
| 27 |
+
*.tar.* filter=lfs diff=lfs merge=lfs -text
|
| 28 |
+
*.tar filter=lfs diff=lfs merge=lfs -text
|
| 29 |
+
*.tflite filter=lfs diff=lfs merge=lfs -text
|
| 30 |
+
*.tgz filter=lfs diff=lfs merge=lfs -text
|
| 31 |
+
*.wasm filter=lfs diff=lfs merge=lfs -text
|
| 32 |
+
*.xz filter=lfs diff=lfs merge=lfs -text
|
| 33 |
+
*.zip filter=lfs diff=lfs merge=lfs -text
|
| 34 |
+
*.zst filter=lfs diff=lfs merge=lfs -text
|
| 35 |
+
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
README.md
ADDED
|
@@ -0,0 +1,39 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
base_model:
|
| 3 |
+
- XHToken/Spark-X2.5-4B
|
| 4 |
+
license: apache-2.0
|
| 5 |
+
language:
|
| 6 |
+
- en
|
| 7 |
+
- zh
|
| 8 |
+
library_name: mlx
|
| 9 |
+
pipeline_tag: text-generation
|
| 10 |
+
tags:
|
| 11 |
+
- mlx
|
| 12 |
+
- spark-x2.5
|
| 13 |
+
- long-context
|
| 14 |
+
- 1m-context
|
| 15 |
+
---
|
| 16 |
+
|
| 17 |
+
# Spark-X2.5-4B MLX 4-bit
|
| 18 |
+
|
| 19 |
+
MLX 4-bit quantization of [XHToken/Spark-X2.5-4B](https://huggingface.co/XHToken/Spark-X2.5-4B), a 4B general-purpose language model for reasoning, coding, tool use, and agentic workflows. Native context: **1,048,576 tokens (1M)**.
|
| 20 |
+
|
| 21 |
+
## Benchmarks
|
| 22 |
+
|
| 23 |
+
|
| 24 |
+
|
| 25 |
+
*Benchmark results reported by XHToken for Spark-X2.5-4B in thinking mode.*
|
| 26 |
+
|
| 27 |
+
## Files
|
| 28 |
+
|
| 29 |
+
| Format | Weights | Size |
|
| 30 |
+
| --- | --- | ---: |
|
| 31 |
+
| MLX 4-bit | [model.safetensors](model.safetensors) | 2.31 GB |
|
| 32 |
+
|
| 33 |
+
Includes the upstream `chat_template.jinja`. Checksums: [SHA256SUMS.txt](SHA256SUMS.txt).
|
| 34 |
+
|
| 35 |
+
## Source
|
| 36 |
+
|
| 37 |
+
- Source model: [XHToken/Spark-X2.5-4B](https://huggingface.co/XHToken/Spark-X2.5-4B)
|
| 38 |
+
- Source revision: [`ea14618d20e76b5b093d3ee20a5b9d733bb12410`](https://huggingface.co/XHToken/Spark-X2.5-4B/tree/ea14618d20e76b5b093d3ee20a5b9d733bb12410)
|
| 39 |
+
- Source license: [Apache-2.0](https://huggingface.co/XHToken/Spark-X2.5-4B/blob/main/LICENSE)
|
SHA256SUMS.txt
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
1fd6370f641e7fbffb2f57562793bf0493f48d55e3e85c8fe9091bd870a87e8b model.safetensors
|
SPARK_MLX_LLM_LICENSE
ADDED
|
@@ -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.
|
assets/benchmark.png
ADDED
|
chat_template.jinja
ADDED
|
@@ -0,0 +1,110 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{#- 0826版本 -#}
|
| 2 |
+
{%- if not messages %}
|
| 3 |
+
{{- raise_exception('No messages provided.') }}
|
| 4 |
+
{%- endif %}
|
| 5 |
+
{%- set enable_thinking = enable_thinking | default(true) %}
|
| 6 |
+
|
| 7 |
+
{#- Render a string or a list of text blocks. -#}
|
| 8 |
+
{%- macro render_content(content, context_name) %}
|
| 9 |
+
{%- if content is string %}
|
| 10 |
+
{{- content }}
|
| 11 |
+
{%- elif content is none or content is undefined %}
|
| 12 |
+
{{- '' }}
|
| 13 |
+
{%- elif content is iterable and content is not mapping %}
|
| 14 |
+
{%- for block in content %}
|
| 15 |
+
{%- if block.type == 'text' %}
|
| 16 |
+
{{- block.text }}
|
| 17 |
+
{%- else %}
|
| 18 |
+
{{- raise_exception('Unsupported ' ~ context_name ~ ' content block type: ' ~ (block.type | string)) }}
|
| 19 |
+
{%- endif %}
|
| 20 |
+
{%- endfor %}
|
| 21 |
+
{%- else %}
|
| 22 |
+
{{- raise_exception(context_name ~ ' content must be a string or a list of text blocks') }}
|
| 23 |
+
{%- endif %}
|
| 24 |
+
{%- endmacro %}
|
| 25 |
+
|
| 26 |
+
{#- Default system prompt -#}
|
| 27 |
+
{%- set default_system = "you are a helpful assistant." %}
|
| 28 |
+
|
| 29 |
+
{#- The first message-level system is placed in the initial system block. -#}
|
| 30 |
+
{%- set ns = namespace(initial_system='') %}
|
| 31 |
+
{%- if messages[0].role == "system" %}
|
| 32 |
+
{%- set ns.initial_system = render_content(messages[0].content, 'system') %}
|
| 33 |
+
{%- endif %}
|
| 34 |
+
|
| 35 |
+
{#- System block -#}
|
| 36 |
+
{{- '<｜start▁of▁sentence｜><|System|>' + '\n' + default_system }}
|
| 37 |
+
{%- if tools %}
|
| 38 |
+
{{- '## Tools' + '\n' + 'You have access to the following functions:' + '\n' + '<tools>' }}
|
| 39 |
+
{%- for tool in tools %}
|
| 40 |
+
{{- '\n' + tool.function | tojson}}
|
| 41 |
+
{%- endfor %}
|
| 42 |
+
{{- '\n' + '</tools>' }}
|
| 43 |
+
{%- endif %}
|
| 44 |
+
{%- if ns.initial_system %}
|
| 45 |
+
{{- '\n\n' + ns.initial_system }}
|
| 46 |
+
{%- endif %}
|
| 47 |
+
{{- '<｜end▁of▁sentence｜>'}}
|
| 48 |
+
|
| 49 |
+
{#- Conversation turns -#}
|
| 50 |
+
{%- for message in messages %}
|
| 51 |
+
{%- if message.role == "system" %}
|
| 52 |
+
{#- The first system message was consumed by the initial block. -#}
|
| 53 |
+
{%- if not loop.first %}
|
| 54 |
+
{{- '<｜start▁of▁sentence｜><|System|>\n' + render_content(message.content, 'system') + '<｜end▁of▁sentence｜>' }}
|
| 55 |
+
{%- endif %}
|
| 56 |
+
{%- elif message.role == "user" %}
|
| 57 |
+
{{- '<｜start▁of▁sentence｜><|User|>' + render_content(message.content, 'user') + '<｜end▁of▁sentence｜>' }}
|
| 58 |
+
{%- elif message.role == "assistant" %}
|
| 59 |
+
{%- set assistant_content = render_content(message.content, 'assistant') %}
|
| 60 |
+
{%- if message.reasoning_content is defined and message.reasoning_content %}
|
| 61 |
+
{%- set reasoning_content = message.reasoning_content %}
|
| 62 |
+
{%- else %}
|
| 63 |
+
{%- set reasoning_content = '' %}
|
| 64 |
+
{%- endif %}
|
| 65 |
+
{{- '<｜start▁of▁sentence｜><|Bot|>'}}
|
| 66 |
+
{%- if reasoning_content %}
|
| 67 |
+
{{- '<think>' + reasoning_content + '</think>'}}
|
| 68 |
+
{%- else %}
|
| 69 |
+
{{- '</think>' }}
|
| 70 |
+
{%- endif %}
|
| 71 |
+
{%- if assistant_content %}
|
| 72 |
+
{{- assistant_content }}
|
| 73 |
+
{%- endif %}
|
| 74 |
+
{%- if message.tool_calls is defined and message.tool_calls is not none %}
|
| 75 |
+
{%- for tool_call in message.tool_calls %}
|
| 76 |
+
{%- if tool_call.function.arguments is not mapping %}
|
| 77 |
+
{{- raise_exception('tool_call.function.arguments must be a dictionary; normalize JSON strings before apply_chat_template') }}
|
| 78 |
+
{%- endif %}
|
| 79 |
+
{%- set args = tool_call.function.arguments %}
|
| 80 |
+
{{- '<tool_call>' + tool_call.function.name }}
|
| 81 |
+
{%- for k, v in args.items() %}
|
| 82 |
+
{{- '<arg_key>' ~ k ~ '</arg_key><arg_value>' ~ (v if v is string else v | tojson) ~ '</arg_value>' }}
|
| 83 |
+
{%- endfor %}
|
| 84 |
+
{{- '</tool_call>' }}
|
| 85 |
+
{%- endfor %}
|
| 86 |
+
{%- endif %}
|
| 87 |
+
{{- '<｜end▁of▁sentence｜>' }}
|
| 88 |
+
{%- elif message.role == "tool" %}
|
| 89 |
+
{%- if loop.previtem is undefined or loop.previtem.role != "tool" %}
|
| 90 |
+
{{- '<｜start▁of▁sentence｜><|Tool|>' }}
|
| 91 |
+
{%- endif %}
|
| 92 |
+
{{- '<tool_response>' ~ message.content ~ '</tool_response>' }}
|
| 93 |
+
{%- if loop.nextitem is undefined or loop.nextitem.role != "tool" %}
|
| 94 |
+
{{- '<｜end▁of▁sentence｜>' }}
|
| 95 |
+
{%- endif %}
|
| 96 |
+
{%- else %}
|
| 97 |
+
{{- raise_exception('Unsupported message role: ' ~ message.role) }}
|
| 98 |
+
{%- endif %}
|
| 99 |
+
{%- endfor %}
|
| 100 |
+
|
| 101 |
+
{#- Generation prompt -#}
|
| 102 |
+
{%- if add_generation_prompt %}
|
| 103 |
+
{{- '<｜start▁of▁sentence｜><|Bot|>' }}
|
| 104 |
+
{%- if enable_thinking is defined and enable_thinking %}
|
| 105 |
+
{{- '<think>' }}
|
| 106 |
+
{%- endif %}
|
| 107 |
+
{%- if enable_thinking is defined and not enable_thinking %}
|
| 108 |
+
{{- '</think>' }}
|
| 109 |
+
{%- endif %}
|
| 110 |
+
{%- endif %}
|
config.json
ADDED
|
@@ -0,0 +1,93 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"Spark2_5ForCausalLM"
|
| 4 |
+
],
|
| 5 |
+
"attention_bias": false,
|
| 6 |
+
"attention_dropout": 0.0,
|
| 7 |
+
"auto_map": {
|
| 8 |
+
"AutoConfig": "configuration_spark.Spark2_5Config",
|
| 9 |
+
"AutoModel": "modeling_spark.Spark2_5Model",
|
| 10 |
+
"AutoModelForCausalLM": "modeling_spark.Spark2_5ForCausalLM"
|
| 11 |
+
},
|
| 12 |
+
"bos_token_id": 0,
|
| 13 |
+
"dtype": "bfloat16",
|
| 14 |
+
"eos_token_id": 1,
|
| 15 |
+
"gate_attn_act_mode": "sigmoid",
|
| 16 |
+
"head_dim": 256,
|
| 17 |
+
"headwise_attn_output_gate": true,
|
| 18 |
+
"hidden_act": "gelu",
|
| 19 |
+
"hidden_size": 2560,
|
| 20 |
+
"initializer_range": 0.01976,
|
| 21 |
+
"intermediate_size": 10240,
|
| 22 |
+
"layer_types": [
|
| 23 |
+
"sliding_attention",
|
| 24 |
+
"sliding_attention",
|
| 25 |
+
"sliding_attention",
|
| 26 |
+
"full_attention",
|
| 27 |
+
"sliding_attention",
|
| 28 |
+
"sliding_attention",
|
| 29 |
+
"sliding_attention",
|
| 30 |
+
"full_attention",
|
| 31 |
+
"sliding_attention",
|
| 32 |
+
"sliding_attention",
|
| 33 |
+
"sliding_attention",
|
| 34 |
+
"full_attention",
|
| 35 |
+
"sliding_attention",
|
| 36 |
+
"sliding_attention",
|
| 37 |
+
"sliding_attention",
|
| 38 |
+
"full_attention",
|
| 39 |
+
"sliding_attention",
|
| 40 |
+
"sliding_attention",
|
| 41 |
+
"sliding_attention",
|
| 42 |
+
"full_attention",
|
| 43 |
+
"sliding_attention",
|
| 44 |
+
"sliding_attention",
|
| 45 |
+
"sliding_attention",
|
| 46 |
+
"full_attention",
|
| 47 |
+
"sliding_attention",
|
| 48 |
+
"sliding_attention",
|
| 49 |
+
"sliding_attention",
|
| 50 |
+
"full_attention",
|
| 51 |
+
"sliding_attention",
|
| 52 |
+
"sliding_attention",
|
| 53 |
+
"sliding_attention",
|
| 54 |
+
"full_attention",
|
| 55 |
+
"sliding_attention",
|
| 56 |
+
"sliding_attention",
|
| 57 |
+
"sliding_attention",
|
| 58 |
+
"full_attention"
|
| 59 |
+
],
|
| 60 |
+
"max_position_embeddings": 1048576,
|
| 61 |
+
"mlp_bias": false,
|
| 62 |
+
"model_type": "spark2_5",
|
| 63 |
+
"num_attention_heads": 16,
|
| 64 |
+
"num_hidden_layers": 36,
|
| 65 |
+
"num_key_value_heads": 4,
|
| 66 |
+
"pad_token_id": 2,
|
| 67 |
+
"quantization": {
|
| 68 |
+
"group_size": 64,
|
| 69 |
+
"bits": 4,
|
| 70 |
+
"mode": "affine"
|
| 71 |
+
},
|
| 72 |
+
"quantization_config": {
|
| 73 |
+
"group_size": 64,
|
| 74 |
+
"bits": 4,
|
| 75 |
+
"mode": "affine"
|
| 76 |
+
},
|
| 77 |
+
"rms_norm_eps": 1e-06,
|
| 78 |
+
"rope_parameters": {
|
| 79 |
+
"full_attention": {
|
| 80 |
+
"partial_rotary_factor": 0.25,
|
| 81 |
+
"rope_theta": 5000000
|
| 82 |
+
},
|
| 83 |
+
"sliding_attention": {
|
| 84 |
+
"partial_rotary_factor": 1.0,
|
| 85 |
+
"rope_theta": 10000
|
| 86 |
+
}
|
| 87 |
+
},
|
| 88 |
+
"sliding_window": 512,
|
| 89 |
+
"tie_word_embeddings": true,
|
| 90 |
+
"transformers_version": "4.57.1",
|
| 91 |
+
"use_cache": true,
|
| 92 |
+
"vocab_size": 131072
|
| 93 |
+
}
|
configuration_spark.py
ADDED
|
@@ -0,0 +1,118 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# coding=utf-8
|
| 2 |
+
# Copyright 2026 The XHToken team and the HuggingFace Inc. team. All rights reserved.
|
| 3 |
+
#
|
| 4 |
+
# Licensed under the Apache License, Version 2.0 (the "License");
|
| 5 |
+
# you may not use this file except in compliance with the License.
|
| 6 |
+
# You may obtain a copy of the License at
|
| 7 |
+
#
|
| 8 |
+
# http://www.apache.org/licenses/LICENSE-2.0
|
| 9 |
+
#
|
| 10 |
+
# Unless required by applicable law or agreed to in writing, software
|
| 11 |
+
# distributed under the License is distributed on an "AS IS" BASIS,
|
| 12 |
+
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
| 13 |
+
# See the License for the specific language governing permissions and
|
| 14 |
+
# limitations under the License.
|
| 15 |
+
|
| 16 |
+
from transformers import PretrainedConfig
|
| 17 |
+
|
| 18 |
+
|
| 19 |
+
class Spark2_5Config(PretrainedConfig):
|
| 20 |
+
model_type = "spark2_5"
|
| 21 |
+
keys_to_ignore_at_inference = ["past_key_values"]
|
| 22 |
+
|
| 23 |
+
base_model_tp_plan = {
|
| 24 |
+
"layers.*.self_attn.q_k_v_proj": "colwise",
|
| 25 |
+
"layers.*.self_attn.g_proj": "colwise",
|
| 26 |
+
"layers.*.self_attn.out_proj": "rowwise",
|
| 27 |
+
"layers.*.mlp.gate_proj": "colwise",
|
| 28 |
+
"layers.*.mlp.up_proj": "colwise",
|
| 29 |
+
"layers.*.mlp.down_proj": "rowwise",
|
| 30 |
+
}
|
| 31 |
+
base_model_pp_plan = {
|
| 32 |
+
"embedding": (["input_ids"], ["inputs_embeds"]),
|
| 33 |
+
"layers": (["hidden_states", "attention_mask"], ["hidden_states"]),
|
| 34 |
+
"norm": (["hidden_states"], ["hidden_states"]),
|
| 35 |
+
}
|
| 36 |
+
|
| 37 |
+
def __init__(
|
| 38 |
+
self,
|
| 39 |
+
vocab_size=32000,
|
| 40 |
+
hidden_size=4096,
|
| 41 |
+
intermediate_size=11008,
|
| 42 |
+
num_hidden_layers=32,
|
| 43 |
+
num_attention_heads=32,
|
| 44 |
+
num_key_value_heads=None,
|
| 45 |
+
hidden_act="gelu",
|
| 46 |
+
max_position_embeddings=2048,
|
| 47 |
+
initializer_range=0.02,
|
| 48 |
+
rms_norm_eps=1e-6,
|
| 49 |
+
use_cache=True,
|
| 50 |
+
pad_token_id=None,
|
| 51 |
+
bos_token_id=1,
|
| 52 |
+
eos_token_id=2,
|
| 53 |
+
tie_word_embeddings=False,
|
| 54 |
+
rope_parameters=None,
|
| 55 |
+
attention_bias=False,
|
| 56 |
+
attention_dropout=0.0,
|
| 57 |
+
mlp_bias=False,
|
| 58 |
+
head_dim=None,
|
| 59 |
+
headwise_attn_output_gate=False,
|
| 60 |
+
gate_attn_act_mode="sigmoid",
|
| 61 |
+
sliding_window=None,
|
| 62 |
+
layer_types=None,
|
| 63 |
+
**kwargs,
|
| 64 |
+
):
|
| 65 |
+
self.vocab_size = vocab_size
|
| 66 |
+
self.max_position_embeddings = max_position_embeddings
|
| 67 |
+
self.hidden_size = hidden_size
|
| 68 |
+
self.intermediate_size = intermediate_size
|
| 69 |
+
self.num_hidden_layers = num_hidden_layers
|
| 70 |
+
self.num_attention_heads = num_attention_heads
|
| 71 |
+
|
| 72 |
+
if num_key_value_heads is None:
|
| 73 |
+
num_key_value_heads = num_attention_heads
|
| 74 |
+
if num_attention_heads % num_key_value_heads != 0:
|
| 75 |
+
raise ValueError(
|
| 76 |
+
f"num_attention_heads ({num_attention_heads}) must be divisible by num_key_value_heads ({num_key_value_heads})"
|
| 77 |
+
)
|
| 78 |
+
self.num_key_value_heads = num_key_value_heads
|
| 79 |
+
|
| 80 |
+
self.hidden_act = hidden_act
|
| 81 |
+
self.initializer_range = initializer_range
|
| 82 |
+
self.rms_norm_eps = rms_norm_eps
|
| 83 |
+
self.use_cache = use_cache
|
| 84 |
+
self.attention_bias = attention_bias
|
| 85 |
+
self.attention_dropout = attention_dropout
|
| 86 |
+
self.mlp_bias = mlp_bias
|
| 87 |
+
self.head_dim = head_dim if head_dim is not None else self.hidden_size // self.num_attention_heads
|
| 88 |
+
self.headwise_attn_output_gate = headwise_attn_output_gate
|
| 89 |
+
self.gate_attn_act_mode = gate_attn_act_mode
|
| 90 |
+
self.sliding_window = sliding_window
|
| 91 |
+
self.rope_parameters = rope_parameters
|
| 92 |
+
|
| 93 |
+
if layer_types is None:
|
| 94 |
+
layer_types = ["full_attention"] * num_hidden_layers
|
| 95 |
+
if len(layer_types) != num_hidden_layers:
|
| 96 |
+
raise ValueError(
|
| 97 |
+
f"layer_types length ({len(layer_types)}) must match num_hidden_layers ({num_hidden_layers})"
|
| 98 |
+
)
|
| 99 |
+
self.layer_types = layer_types
|
| 100 |
+
|
| 101 |
+
super().__init__(
|
| 102 |
+
pad_token_id=pad_token_id,
|
| 103 |
+
bos_token_id=bos_token_id,
|
| 104 |
+
eos_token_id=eos_token_id,
|
| 105 |
+
tie_word_embeddings=tie_word_embeddings,
|
| 106 |
+
**kwargs,
|
| 107 |
+
)
|
| 108 |
+
|
| 109 |
+
def get_rope_theta(self, layer_type):
|
| 110 |
+
params = self.rope_parameters.get(layer_type, {})
|
| 111 |
+
return params.get("rope_theta", 10000)
|
| 112 |
+
|
| 113 |
+
def get_partial_rotary_factor(self, layer_type):
|
| 114 |
+
params = self.rope_parameters.get(layer_type, {})
|
| 115 |
+
return params.get("partial_rotary_factor", 1.0)
|
| 116 |
+
|
| 117 |
+
|
| 118 |
+
__all__ = ["Spark2_5Config"]
|
generation_config.json
ADDED
|
@@ -0,0 +1,14 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"bos_token_id": 0,
|
| 3 |
+
"eos_token_id": 1,
|
| 4 |
+
"pad_token_id": 2,
|
| 5 |
+
"max_tokens": 1048576,
|
| 6 |
+
"temperature": 1.0,
|
| 7 |
+
"top_p": 0.95,
|
| 8 |
+
"top_k": -1,
|
| 9 |
+
"repetition_penalty": 1.0,
|
| 10 |
+
"presence_penalty":0,
|
| 11 |
+
"frequency_penalty":0,
|
| 12 |
+
"do_sample": true,
|
| 13 |
+
"transformers_version": "4.57.1"
|
| 14 |
+
}
|
model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1fd6370f641e7fbffb2f57562793bf0493f48d55e3e85c8fe9091bd870a87e8b
|
| 3 |
+
size 2313395808
|
model.safetensors.index.json
ADDED
|
@@ -0,0 +1,732 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"metadata": {
|
| 3 |
+
"total_size": 2313313280,
|
| 4 |
+
"total_parameters": 4112079360
|
| 5 |
+
},
|
| 6 |
+
"weight_map": {
|
| 7 |
+
"model.embedding.biases": "model.safetensors",
|
| 8 |
+
"model.embedding.scales": "model.safetensors",
|
| 9 |
+
"model.embedding.weight": "model.safetensors",
|
| 10 |
+
"model.layers.0.input_layernorm.weight": "model.safetensors",
|
| 11 |
+
"model.layers.0.mlp.down_proj.biases": "model.safetensors",
|
| 12 |
+
"model.layers.0.mlp.down_proj.scales": "model.safetensors",
|
| 13 |
+
"model.layers.0.mlp.down_proj.weight": "model.safetensors",
|
| 14 |
+
"model.layers.0.mlp.gate_proj.biases": "model.safetensors",
|
| 15 |
+
"model.layers.0.mlp.gate_proj.scales": "model.safetensors",
|
| 16 |
+
"model.layers.0.mlp.gate_proj.weight": "model.safetensors",
|
| 17 |
+
"model.layers.0.mlp.up_proj.biases": "model.safetensors",
|
| 18 |
+
"model.layers.0.mlp.up_proj.scales": "model.safetensors",
|
| 19 |
+
"model.layers.0.mlp.up_proj.weight": "model.safetensors",
|
| 20 |
+
"model.layers.0.post_attention_layernorm.weight": "model.safetensors",
|
| 21 |
+
"model.layers.0.self_attn.g_proj.biases": "model.safetensors",
|
| 22 |
+
"model.layers.0.self_attn.g_proj.scales": "model.safetensors",
|
| 23 |
+
"model.layers.0.self_attn.g_proj.weight": "model.safetensors",
|
| 24 |
+
"model.layers.0.self_attn.out_proj.biases": "model.safetensors",
|
| 25 |
+
"model.layers.0.self_attn.out_proj.scales": "model.safetensors",
|
| 26 |
+
"model.layers.0.self_attn.out_proj.weight": "model.safetensors",
|
| 27 |
+
"model.layers.0.self_attn.q_k_v_proj.biases": "model.safetensors",
|
| 28 |
+
"model.layers.0.self_attn.q_k_v_proj.scales": "model.safetensors",
|
| 29 |
+
"model.layers.0.self_attn.q_k_v_proj.weight": "model.safetensors",
|
| 30 |
+
"model.layers.1.input_layernorm.weight": "model.safetensors",
|
| 31 |
+
"model.layers.1.mlp.down_proj.biases": "model.safetensors",
|
| 32 |
+
"model.layers.1.mlp.down_proj.scales": "model.safetensors",
|
| 33 |
+
"model.layers.1.mlp.down_proj.weight": "model.safetensors",
|
| 34 |
+
"model.layers.1.mlp.gate_proj.biases": "model.safetensors",
|
| 35 |
+
"model.layers.1.mlp.gate_proj.scales": "model.safetensors",
|
| 36 |
+
"model.layers.1.mlp.gate_proj.weight": "model.safetensors",
|
| 37 |
+
"model.layers.1.mlp.up_proj.biases": "model.safetensors",
|
| 38 |
+
"model.layers.1.mlp.up_proj.scales": "model.safetensors",
|
| 39 |
+
"model.layers.1.mlp.up_proj.weight": "model.safetensors",
|
| 40 |
+
"model.layers.1.post_attention_layernorm.weight": "model.safetensors",
|
| 41 |
+
"model.layers.1.self_attn.g_proj.biases": "model.safetensors",
|
| 42 |
+
"model.layers.1.self_attn.g_proj.scales": "model.safetensors",
|
| 43 |
+
"model.layers.1.self_attn.g_proj.weight": "model.safetensors",
|
| 44 |
+
"model.layers.1.self_attn.out_proj.biases": "model.safetensors",
|
| 45 |
+
"model.layers.1.self_attn.out_proj.scales": "model.safetensors",
|
| 46 |
+
"model.layers.1.self_attn.out_proj.weight": "model.safetensors",
|
| 47 |
+
"model.layers.1.self_attn.q_k_v_proj.biases": "model.safetensors",
|
| 48 |
+
"model.layers.1.self_attn.q_k_v_proj.scales": "model.safetensors",
|
| 49 |
+
"model.layers.1.self_attn.q_k_v_proj.weight": "model.safetensors",
|
| 50 |
+
"model.layers.10.input_layernorm.weight": "model.safetensors",
|
| 51 |
+
"model.layers.10.mlp.down_proj.biases": "model.safetensors",
|
| 52 |
+
"model.layers.10.mlp.down_proj.scales": "model.safetensors",
|
| 53 |
+
"model.layers.10.mlp.down_proj.weight": "model.safetensors",
|
| 54 |
+
"model.layers.10.mlp.gate_proj.biases": "model.safetensors",
|
| 55 |
+
"model.layers.10.mlp.gate_proj.scales": "model.safetensors",
|
| 56 |
+
"model.layers.10.mlp.gate_proj.weight": "model.safetensors",
|
| 57 |
+
"model.layers.10.mlp.up_proj.biases": "model.safetensors",
|
| 58 |
+
"model.layers.10.mlp.up_proj.scales": "model.safetensors",
|
| 59 |
+
"model.layers.10.mlp.up_proj.weight": "model.safetensors",
|
| 60 |
+
"model.layers.10.post_attention_layernorm.weight": "model.safetensors",
|
| 61 |
+
"model.layers.10.self_attn.g_proj.biases": "model.safetensors",
|
| 62 |
+
"model.layers.10.self_attn.g_proj.scales": "model.safetensors",
|
| 63 |
+
"model.layers.10.self_attn.g_proj.weight": "model.safetensors",
|
| 64 |
+
"model.layers.10.self_attn.out_proj.biases": "model.safetensors",
|
| 65 |
+
"model.layers.10.self_attn.out_proj.scales": "model.safetensors",
|
| 66 |
+
"model.layers.10.self_attn.out_proj.weight": "model.safetensors",
|
| 67 |
+
"model.layers.10.self_attn.q_k_v_proj.biases": "model.safetensors",
|
| 68 |
+
"model.layers.10.self_attn.q_k_v_proj.scales": "model.safetensors",
|
| 69 |
+
"model.layers.10.self_attn.q_k_v_proj.weight": "model.safetensors",
|
| 70 |
+
"model.layers.11.input_layernorm.weight": "model.safetensors",
|
| 71 |
+
"model.layers.11.mlp.down_proj.biases": "model.safetensors",
|
| 72 |
+
"model.layers.11.mlp.down_proj.scales": "model.safetensors",
|
| 73 |
+
"model.layers.11.mlp.down_proj.weight": "model.safetensors",
|
| 74 |
+
"model.layers.11.mlp.gate_proj.biases": "model.safetensors",
|
| 75 |
+
"model.layers.11.mlp.gate_proj.scales": "model.safetensors",
|
| 76 |
+
"model.layers.11.mlp.gate_proj.weight": "model.safetensors",
|
| 77 |
+
"model.layers.11.mlp.up_proj.biases": "model.safetensors",
|
| 78 |
+
"model.layers.11.mlp.up_proj.scales": "model.safetensors",
|
| 79 |
+
"model.layers.11.mlp.up_proj.weight": "model.safetensors",
|
| 80 |
+
"model.layers.11.post_attention_layernorm.weight": "model.safetensors",
|
| 81 |
+
"model.layers.11.self_attn.g_proj.biases": "model.safetensors",
|
| 82 |
+
"model.layers.11.self_attn.g_proj.scales": "model.safetensors",
|
| 83 |
+
"model.layers.11.self_attn.g_proj.weight": "model.safetensors",
|
| 84 |
+
"model.layers.11.self_attn.out_proj.biases": "model.safetensors",
|
| 85 |
+
"model.layers.11.self_attn.out_proj.scales": "model.safetensors",
|
| 86 |
+
"model.layers.11.self_attn.out_proj.weight": "model.safetensors",
|
| 87 |
+
"model.layers.11.self_attn.q_k_v_proj.biases": "model.safetensors",
|
| 88 |
+
"model.layers.11.self_attn.q_k_v_proj.scales": "model.safetensors",
|
| 89 |
+
"model.layers.11.self_attn.q_k_v_proj.weight": "model.safetensors",
|
| 90 |
+
"model.layers.12.input_layernorm.weight": "model.safetensors",
|
| 91 |
+
"model.layers.12.mlp.down_proj.biases": "model.safetensors",
|
| 92 |
+
"model.layers.12.mlp.down_proj.scales": "model.safetensors",
|
| 93 |
+
"model.layers.12.mlp.down_proj.weight": "model.safetensors",
|
| 94 |
+
"model.layers.12.mlp.gate_proj.biases": "model.safetensors",
|
| 95 |
+
"model.layers.12.mlp.gate_proj.scales": "model.safetensors",
|
| 96 |
+
"model.layers.12.mlp.gate_proj.weight": "model.safetensors",
|
| 97 |
+
"model.layers.12.mlp.up_proj.biases": "model.safetensors",
|
| 98 |
+
"model.layers.12.mlp.up_proj.scales": "model.safetensors",
|
| 99 |
+
"model.layers.12.mlp.up_proj.weight": "model.safetensors",
|
| 100 |
+
"model.layers.12.post_attention_layernorm.weight": "model.safetensors",
|
| 101 |
+
"model.layers.12.self_attn.g_proj.biases": "model.safetensors",
|
| 102 |
+
"model.layers.12.self_attn.g_proj.scales": "model.safetensors",
|
| 103 |
+
"model.layers.12.self_attn.g_proj.weight": "model.safetensors",
|
| 104 |
+
"model.layers.12.self_attn.out_proj.biases": "model.safetensors",
|
| 105 |
+
"model.layers.12.self_attn.out_proj.scales": "model.safetensors",
|
| 106 |
+
"model.layers.12.self_attn.out_proj.weight": "model.safetensors",
|
| 107 |
+
"model.layers.12.self_attn.q_k_v_proj.biases": "model.safetensors",
|
| 108 |
+
"model.layers.12.self_attn.q_k_v_proj.scales": "model.safetensors",
|
| 109 |
+
"model.layers.12.self_attn.q_k_v_proj.weight": "model.safetensors",
|
| 110 |
+
"model.layers.13.input_layernorm.weight": "model.safetensors",
|
| 111 |
+
"model.layers.13.mlp.down_proj.biases": "model.safetensors",
|
| 112 |
+
"model.layers.13.mlp.down_proj.scales": "model.safetensors",
|
| 113 |
+
"model.layers.13.mlp.down_proj.weight": "model.safetensors",
|
| 114 |
+
"model.layers.13.mlp.gate_proj.biases": "model.safetensors",
|
| 115 |
+
"model.layers.13.mlp.gate_proj.scales": "model.safetensors",
|
| 116 |
+
"model.layers.13.mlp.gate_proj.weight": "model.safetensors",
|
| 117 |
+
"model.layers.13.mlp.up_proj.biases": "model.safetensors",
|
| 118 |
+
"model.layers.13.mlp.up_proj.scales": "model.safetensors",
|
| 119 |
+
"model.layers.13.mlp.up_proj.weight": "model.safetensors",
|
| 120 |
+
"model.layers.13.post_attention_layernorm.weight": "model.safetensors",
|
| 121 |
+
"model.layers.13.self_attn.g_proj.biases": "model.safetensors",
|
| 122 |
+
"model.layers.13.self_attn.g_proj.scales": "model.safetensors",
|
| 123 |
+
"model.layers.13.self_attn.g_proj.weight": "model.safetensors",
|
| 124 |
+
"model.layers.13.self_attn.out_proj.biases": "model.safetensors",
|
| 125 |
+
"model.layers.13.self_attn.out_proj.scales": "model.safetensors",
|
| 126 |
+
"model.layers.13.self_attn.out_proj.weight": "model.safetensors",
|
| 127 |
+
"model.layers.13.self_attn.q_k_v_proj.biases": "model.safetensors",
|
| 128 |
+
"model.layers.13.self_attn.q_k_v_proj.scales": "model.safetensors",
|
| 129 |
+
"model.layers.13.self_attn.q_k_v_proj.weight": "model.safetensors",
|
| 130 |
+
"model.layers.14.input_layernorm.weight": "model.safetensors",
|
| 131 |
+
"model.layers.14.mlp.down_proj.biases": "model.safetensors",
|
| 132 |
+
"model.layers.14.mlp.down_proj.scales": "model.safetensors",
|
| 133 |
+
"model.layers.14.mlp.down_proj.weight": "model.safetensors",
|
| 134 |
+
"model.layers.14.mlp.gate_proj.biases": "model.safetensors",
|
| 135 |
+
"model.layers.14.mlp.gate_proj.scales": "model.safetensors",
|
| 136 |
+
"model.layers.14.mlp.gate_proj.weight": "model.safetensors",
|
| 137 |
+
"model.layers.14.mlp.up_proj.biases": "model.safetensors",
|
| 138 |
+
"model.layers.14.mlp.up_proj.scales": "model.safetensors",
|
| 139 |
+
"model.layers.14.mlp.up_proj.weight": "model.safetensors",
|
| 140 |
+
"model.layers.14.post_attention_layernorm.weight": "model.safetensors",
|
| 141 |
+
"model.layers.14.self_attn.g_proj.biases": "model.safetensors",
|
| 142 |
+
"model.layers.14.self_attn.g_proj.scales": "model.safetensors",
|
| 143 |
+
"model.layers.14.self_attn.g_proj.weight": "model.safetensors",
|
| 144 |
+
"model.layers.14.self_attn.out_proj.biases": "model.safetensors",
|
| 145 |
+
"model.layers.14.self_attn.out_proj.scales": "model.safetensors",
|
| 146 |
+
"model.layers.14.self_attn.out_proj.weight": "model.safetensors",
|
| 147 |
+
"model.layers.14.self_attn.q_k_v_proj.biases": "model.safetensors",
|
| 148 |
+
"model.layers.14.self_attn.q_k_v_proj.scales": "model.safetensors",
|
| 149 |
+
"model.layers.14.self_attn.q_k_v_proj.weight": "model.safetensors",
|
| 150 |
+
"model.layers.15.input_layernorm.weight": "model.safetensors",
|
| 151 |
+
"model.layers.15.mlp.down_proj.biases": "model.safetensors",
|
| 152 |
+
"model.layers.15.mlp.down_proj.scales": "model.safetensors",
|
| 153 |
+
"model.layers.15.mlp.down_proj.weight": "model.safetensors",
|
| 154 |
+
"model.layers.15.mlp.gate_proj.biases": "model.safetensors",
|
| 155 |
+
"model.layers.15.mlp.gate_proj.scales": "model.safetensors",
|
| 156 |
+
"model.layers.15.mlp.gate_proj.weight": "model.safetensors",
|
| 157 |
+
"model.layers.15.mlp.up_proj.biases": "model.safetensors",
|
| 158 |
+
"model.layers.15.mlp.up_proj.scales": "model.safetensors",
|
| 159 |
+
"model.layers.15.mlp.up_proj.weight": "model.safetensors",
|
| 160 |
+
"model.layers.15.post_attention_layernorm.weight": "model.safetensors",
|
| 161 |
+
"model.layers.15.self_attn.g_proj.biases": "model.safetensors",
|
| 162 |
+
"model.layers.15.self_attn.g_proj.scales": "model.safetensors",
|
| 163 |
+
"model.layers.15.self_attn.g_proj.weight": "model.safetensors",
|
| 164 |
+
"model.layers.15.self_attn.out_proj.biases": "model.safetensors",
|
| 165 |
+
"model.layers.15.self_attn.out_proj.scales": "model.safetensors",
|
| 166 |
+
"model.layers.15.self_attn.out_proj.weight": "model.safetensors",
|
| 167 |
+
"model.layers.15.self_attn.q_k_v_proj.biases": "model.safetensors",
|
| 168 |
+
"model.layers.15.self_attn.q_k_v_proj.scales": "model.safetensors",
|
| 169 |
+
"model.layers.15.self_attn.q_k_v_proj.weight": "model.safetensors",
|
| 170 |
+
"model.layers.16.input_layernorm.weight": "model.safetensors",
|
| 171 |
+
"model.layers.16.mlp.down_proj.biases": "model.safetensors",
|
| 172 |
+
"model.layers.16.mlp.down_proj.scales": "model.safetensors",
|
| 173 |
+
"model.layers.16.mlp.down_proj.weight": "model.safetensors",
|
| 174 |
+
"model.layers.16.mlp.gate_proj.biases": "model.safetensors",
|
| 175 |
+
"model.layers.16.mlp.gate_proj.scales": "model.safetensors",
|
| 176 |
+
"model.layers.16.mlp.gate_proj.weight": "model.safetensors",
|
| 177 |
+
"model.layers.16.mlp.up_proj.biases": "model.safetensors",
|
| 178 |
+
"model.layers.16.mlp.up_proj.scales": "model.safetensors",
|
| 179 |
+
"model.layers.16.mlp.up_proj.weight": "model.safetensors",
|
| 180 |
+
"model.layers.16.post_attention_layernorm.weight": "model.safetensors",
|
| 181 |
+
"model.layers.16.self_attn.g_proj.biases": "model.safetensors",
|
| 182 |
+
"model.layers.16.self_attn.g_proj.scales": "model.safetensors",
|
| 183 |
+
"model.layers.16.self_attn.g_proj.weight": "model.safetensors",
|
| 184 |
+
"model.layers.16.self_attn.out_proj.biases": "model.safetensors",
|
| 185 |
+
"model.layers.16.self_attn.out_proj.scales": "model.safetensors",
|
| 186 |
+
"model.layers.16.self_attn.out_proj.weight": "model.safetensors",
|
| 187 |
+
"model.layers.16.self_attn.q_k_v_proj.biases": "model.safetensors",
|
| 188 |
+
"model.layers.16.self_attn.q_k_v_proj.scales": "model.safetensors",
|
| 189 |
+
"model.layers.16.self_attn.q_k_v_proj.weight": "model.safetensors",
|
| 190 |
+
"model.layers.17.input_layernorm.weight": "model.safetensors",
|
| 191 |
+
"model.layers.17.mlp.down_proj.biases": "model.safetensors",
|
| 192 |
+
"model.layers.17.mlp.down_proj.scales": "model.safetensors",
|
| 193 |
+
"model.layers.17.mlp.down_proj.weight": "model.safetensors",
|
| 194 |
+
"model.layers.17.mlp.gate_proj.biases": "model.safetensors",
|
| 195 |
+
"model.layers.17.mlp.gate_proj.scales": "model.safetensors",
|
| 196 |
+
"model.layers.17.mlp.gate_proj.weight": "model.safetensors",
|
| 197 |
+
"model.layers.17.mlp.up_proj.biases": "model.safetensors",
|
| 198 |
+
"model.layers.17.mlp.up_proj.scales": "model.safetensors",
|
| 199 |
+
"model.layers.17.mlp.up_proj.weight": "model.safetensors",
|
| 200 |
+
"model.layers.17.post_attention_layernorm.weight": "model.safetensors",
|
| 201 |
+
"model.layers.17.self_attn.g_proj.biases": "model.safetensors",
|
| 202 |
+
"model.layers.17.self_attn.g_proj.scales": "model.safetensors",
|
| 203 |
+
"model.layers.17.self_attn.g_proj.weight": "model.safetensors",
|
| 204 |
+
"model.layers.17.self_attn.out_proj.biases": "model.safetensors",
|
| 205 |
+
"model.layers.17.self_attn.out_proj.scales": "model.safetensors",
|
| 206 |
+
"model.layers.17.self_attn.out_proj.weight": "model.safetensors",
|
| 207 |
+
"model.layers.17.self_attn.q_k_v_proj.biases": "model.safetensors",
|
| 208 |
+
"model.layers.17.self_attn.q_k_v_proj.scales": "model.safetensors",
|
| 209 |
+
"model.layers.17.self_attn.q_k_v_proj.weight": "model.safetensors",
|
| 210 |
+
"model.layers.18.input_layernorm.weight": "model.safetensors",
|
| 211 |
+
"model.layers.18.mlp.down_proj.biases": "model.safetensors",
|
| 212 |
+
"model.layers.18.mlp.down_proj.scales": "model.safetensors",
|
| 213 |
+
"model.layers.18.mlp.down_proj.weight": "model.safetensors",
|
| 214 |
+
"model.layers.18.mlp.gate_proj.biases": "model.safetensors",
|
| 215 |
+
"model.layers.18.mlp.gate_proj.scales": "model.safetensors",
|
| 216 |
+
"model.layers.18.mlp.gate_proj.weight": "model.safetensors",
|
| 217 |
+
"model.layers.18.mlp.up_proj.biases": "model.safetensors",
|
| 218 |
+
"model.layers.18.mlp.up_proj.scales": "model.safetensors",
|
| 219 |
+
"model.layers.18.mlp.up_proj.weight": "model.safetensors",
|
| 220 |
+
"model.layers.18.post_attention_layernorm.weight": "model.safetensors",
|
| 221 |
+
"model.layers.18.self_attn.g_proj.biases": "model.safetensors",
|
| 222 |
+
"model.layers.18.self_attn.g_proj.scales": "model.safetensors",
|
| 223 |
+
"model.layers.18.self_attn.g_proj.weight": "model.safetensors",
|
| 224 |
+
"model.layers.18.self_attn.out_proj.biases": "model.safetensors",
|
| 225 |
+
"model.layers.18.self_attn.out_proj.scales": "model.safetensors",
|
| 226 |
+
"model.layers.18.self_attn.out_proj.weight": "model.safetensors",
|
| 227 |
+
"model.layers.18.self_attn.q_k_v_proj.biases": "model.safetensors",
|
| 228 |
+
"model.layers.18.self_attn.q_k_v_proj.scales": "model.safetensors",
|
| 229 |
+
"model.layers.18.self_attn.q_k_v_proj.weight": "model.safetensors",
|
| 230 |
+
"model.layers.19.input_layernorm.weight": "model.safetensors",
|
| 231 |
+
"model.layers.19.mlp.down_proj.biases": "model.safetensors",
|
| 232 |
+
"model.layers.19.mlp.down_proj.scales": "model.safetensors",
|
| 233 |
+
"model.layers.19.mlp.down_proj.weight": "model.safetensors",
|
| 234 |
+
"model.layers.19.mlp.gate_proj.biases": "model.safetensors",
|
| 235 |
+
"model.layers.19.mlp.gate_proj.scales": "model.safetensors",
|
| 236 |
+
"model.layers.19.mlp.gate_proj.weight": "model.safetensors",
|
| 237 |
+
"model.layers.19.mlp.up_proj.biases": "model.safetensors",
|
| 238 |
+
"model.layers.19.mlp.up_proj.scales": "model.safetensors",
|
| 239 |
+
"model.layers.19.mlp.up_proj.weight": "model.safetensors",
|
| 240 |
+
"model.layers.19.post_attention_layernorm.weight": "model.safetensors",
|
| 241 |
+
"model.layers.19.self_attn.g_proj.biases": "model.safetensors",
|
| 242 |
+
"model.layers.19.self_attn.g_proj.scales": "model.safetensors",
|
| 243 |
+
"model.layers.19.self_attn.g_proj.weight": "model.safetensors",
|
| 244 |
+
"model.layers.19.self_attn.out_proj.biases": "model.safetensors",
|
| 245 |
+
"model.layers.19.self_attn.out_proj.scales": "model.safetensors",
|
| 246 |
+
"model.layers.19.self_attn.out_proj.weight": "model.safetensors",
|
| 247 |
+
"model.layers.19.self_attn.q_k_v_proj.biases": "model.safetensors",
|
| 248 |
+
"model.layers.19.self_attn.q_k_v_proj.scales": "model.safetensors",
|
| 249 |
+
"model.layers.19.self_attn.q_k_v_proj.weight": "model.safetensors",
|
| 250 |
+
"model.layers.2.input_layernorm.weight": "model.safetensors",
|
| 251 |
+
"model.layers.2.mlp.down_proj.biases": "model.safetensors",
|
| 252 |
+
"model.layers.2.mlp.down_proj.scales": "model.safetensors",
|
| 253 |
+
"model.layers.2.mlp.down_proj.weight": "model.safetensors",
|
| 254 |
+
"model.layers.2.mlp.gate_proj.biases": "model.safetensors",
|
| 255 |
+
"model.layers.2.mlp.gate_proj.scales": "model.safetensors",
|
| 256 |
+
"model.layers.2.mlp.gate_proj.weight": "model.safetensors",
|
| 257 |
+
"model.layers.2.mlp.up_proj.biases": "model.safetensors",
|
| 258 |
+
"model.layers.2.mlp.up_proj.scales": "model.safetensors",
|
| 259 |
+
"model.layers.2.mlp.up_proj.weight": "model.safetensors",
|
| 260 |
+
"model.layers.2.post_attention_layernorm.weight": "model.safetensors",
|
| 261 |
+
"model.layers.2.self_attn.g_proj.biases": "model.safetensors",
|
| 262 |
+
"model.layers.2.self_attn.g_proj.scales": "model.safetensors",
|
| 263 |
+
"model.layers.2.self_attn.g_proj.weight": "model.safetensors",
|
| 264 |
+
"model.layers.2.self_attn.out_proj.biases": "model.safetensors",
|
| 265 |
+
"model.layers.2.self_attn.out_proj.scales": "model.safetensors",
|
| 266 |
+
"model.layers.2.self_attn.out_proj.weight": "model.safetensors",
|
| 267 |
+
"model.layers.2.self_attn.q_k_v_proj.biases": "model.safetensors",
|
| 268 |
+
"model.layers.2.self_attn.q_k_v_proj.scales": "model.safetensors",
|
| 269 |
+
"model.layers.2.self_attn.q_k_v_proj.weight": "model.safetensors",
|
| 270 |
+
"model.layers.20.input_layernorm.weight": "model.safetensors",
|
| 271 |
+
"model.layers.20.mlp.down_proj.biases": "model.safetensors",
|
| 272 |
+
"model.layers.20.mlp.down_proj.scales": "model.safetensors",
|
| 273 |
+
"model.layers.20.mlp.down_proj.weight": "model.safetensors",
|
| 274 |
+
"model.layers.20.mlp.gate_proj.biases": "model.safetensors",
|
| 275 |
+
"model.layers.20.mlp.gate_proj.scales": "model.safetensors",
|
| 276 |
+
"model.layers.20.mlp.gate_proj.weight": "model.safetensors",
|
| 277 |
+
"model.layers.20.mlp.up_proj.biases": "model.safetensors",
|
| 278 |
+
"model.layers.20.mlp.up_proj.scales": "model.safetensors",
|
| 279 |
+
"model.layers.20.mlp.up_proj.weight": "model.safetensors",
|
| 280 |
+
"model.layers.20.post_attention_layernorm.weight": "model.safetensors",
|
| 281 |
+
"model.layers.20.self_attn.g_proj.biases": "model.safetensors",
|
| 282 |
+
"model.layers.20.self_attn.g_proj.scales": "model.safetensors",
|
| 283 |
+
"model.layers.20.self_attn.g_proj.weight": "model.safetensors",
|
| 284 |
+
"model.layers.20.self_attn.out_proj.biases": "model.safetensors",
|
| 285 |
+
"model.layers.20.self_attn.out_proj.scales": "model.safetensors",
|
| 286 |
+
"model.layers.20.self_attn.out_proj.weight": "model.safetensors",
|
| 287 |
+
"model.layers.20.self_attn.q_k_v_proj.biases": "model.safetensors",
|
| 288 |
+
"model.layers.20.self_attn.q_k_v_proj.scales": "model.safetensors",
|
| 289 |
+
"model.layers.20.self_attn.q_k_v_proj.weight": "model.safetensors",
|
| 290 |
+
"model.layers.21.input_layernorm.weight": "model.safetensors",
|
| 291 |
+
"model.layers.21.mlp.down_proj.biases": "model.safetensors",
|
| 292 |
+
"model.layers.21.mlp.down_proj.scales": "model.safetensors",
|
| 293 |
+
"model.layers.21.mlp.down_proj.weight": "model.safetensors",
|
| 294 |
+
"model.layers.21.mlp.gate_proj.biases": "model.safetensors",
|
| 295 |
+
"model.layers.21.mlp.gate_proj.scales": "model.safetensors",
|
| 296 |
+
"model.layers.21.mlp.gate_proj.weight": "model.safetensors",
|
| 297 |
+
"model.layers.21.mlp.up_proj.biases": "model.safetensors",
|
| 298 |
+
"model.layers.21.mlp.up_proj.scales": "model.safetensors",
|
| 299 |
+
"model.layers.21.mlp.up_proj.weight": "model.safetensors",
|
| 300 |
+
"model.layers.21.post_attention_layernorm.weight": "model.safetensors",
|
| 301 |
+
"model.layers.21.self_attn.g_proj.biases": "model.safetensors",
|
| 302 |
+
"model.layers.21.self_attn.g_proj.scales": "model.safetensors",
|
| 303 |
+
"model.layers.21.self_attn.g_proj.weight": "model.safetensors",
|
| 304 |
+
"model.layers.21.self_attn.out_proj.biases": "model.safetensors",
|
| 305 |
+
"model.layers.21.self_attn.out_proj.scales": "model.safetensors",
|
| 306 |
+
"model.layers.21.self_attn.out_proj.weight": "model.safetensors",
|
| 307 |
+
"model.layers.21.self_attn.q_k_v_proj.biases": "model.safetensors",
|
| 308 |
+
"model.layers.21.self_attn.q_k_v_proj.scales": "model.safetensors",
|
| 309 |
+
"model.layers.21.self_attn.q_k_v_proj.weight": "model.safetensors",
|
| 310 |
+
"model.layers.22.input_layernorm.weight": "model.safetensors",
|
| 311 |
+
"model.layers.22.mlp.down_proj.biases": "model.safetensors",
|
| 312 |
+
"model.layers.22.mlp.down_proj.scales": "model.safetensors",
|
| 313 |
+
"model.layers.22.mlp.down_proj.weight": "model.safetensors",
|
| 314 |
+
"model.layers.22.mlp.gate_proj.biases": "model.safetensors",
|
| 315 |
+
"model.layers.22.mlp.gate_proj.scales": "model.safetensors",
|
| 316 |
+
"model.layers.22.mlp.gate_proj.weight": "model.safetensors",
|
| 317 |
+
"model.layers.22.mlp.up_proj.biases": "model.safetensors",
|
| 318 |
+
"model.layers.22.mlp.up_proj.scales": "model.safetensors",
|
| 319 |
+
"model.layers.22.mlp.up_proj.weight": "model.safetensors",
|
| 320 |
+
"model.layers.22.post_attention_layernorm.weight": "model.safetensors",
|
| 321 |
+
"model.layers.22.self_attn.g_proj.biases": "model.safetensors",
|
| 322 |
+
"model.layers.22.self_attn.g_proj.scales": "model.safetensors",
|
| 323 |
+
"model.layers.22.self_attn.g_proj.weight": "model.safetensors",
|
| 324 |
+
"model.layers.22.self_attn.out_proj.biases": "model.safetensors",
|
| 325 |
+
"model.layers.22.self_attn.out_proj.scales": "model.safetensors",
|
| 326 |
+
"model.layers.22.self_attn.out_proj.weight": "model.safetensors",
|
| 327 |
+
"model.layers.22.self_attn.q_k_v_proj.biases": "model.safetensors",
|
| 328 |
+
"model.layers.22.self_attn.q_k_v_proj.scales": "model.safetensors",
|
| 329 |
+
"model.layers.22.self_attn.q_k_v_proj.weight": "model.safetensors",
|
| 330 |
+
"model.layers.23.input_layernorm.weight": "model.safetensors",
|
| 331 |
+
"model.layers.23.mlp.down_proj.biases": "model.safetensors",
|
| 332 |
+
"model.layers.23.mlp.down_proj.scales": "model.safetensors",
|
| 333 |
+
"model.layers.23.mlp.down_proj.weight": "model.safetensors",
|
| 334 |
+
"model.layers.23.mlp.gate_proj.biases": "model.safetensors",
|
| 335 |
+
"model.layers.23.mlp.gate_proj.scales": "model.safetensors",
|
| 336 |
+
"model.layers.23.mlp.gate_proj.weight": "model.safetensors",
|
| 337 |
+
"model.layers.23.mlp.up_proj.biases": "model.safetensors",
|
| 338 |
+
"model.layers.23.mlp.up_proj.scales": "model.safetensors",
|
| 339 |
+
"model.layers.23.mlp.up_proj.weight": "model.safetensors",
|
| 340 |
+
"model.layers.23.post_attention_layernorm.weight": "model.safetensors",
|
| 341 |
+
"model.layers.23.self_attn.g_proj.biases": "model.safetensors",
|
| 342 |
+
"model.layers.23.self_attn.g_proj.scales": "model.safetensors",
|
| 343 |
+
"model.layers.23.self_attn.g_proj.weight": "model.safetensors",
|
| 344 |
+
"model.layers.23.self_attn.out_proj.biases": "model.safetensors",
|
| 345 |
+
"model.layers.23.self_attn.out_proj.scales": "model.safetensors",
|
| 346 |
+
"model.layers.23.self_attn.out_proj.weight": "model.safetensors",
|
| 347 |
+
"model.layers.23.self_attn.q_k_v_proj.biases": "model.safetensors",
|
| 348 |
+
"model.layers.23.self_attn.q_k_v_proj.scales": "model.safetensors",
|
| 349 |
+
"model.layers.23.self_attn.q_k_v_proj.weight": "model.safetensors",
|
| 350 |
+
"model.layers.24.input_layernorm.weight": "model.safetensors",
|
| 351 |
+
"model.layers.24.mlp.down_proj.biases": "model.safetensors",
|
| 352 |
+
"model.layers.24.mlp.down_proj.scales": "model.safetensors",
|
| 353 |
+
"model.layers.24.mlp.down_proj.weight": "model.safetensors",
|
| 354 |
+
"model.layers.24.mlp.gate_proj.biases": "model.safetensors",
|
| 355 |
+
"model.layers.24.mlp.gate_proj.scales": "model.safetensors",
|
| 356 |
+
"model.layers.24.mlp.gate_proj.weight": "model.safetensors",
|
| 357 |
+
"model.layers.24.mlp.up_proj.biases": "model.safetensors",
|
| 358 |
+
"model.layers.24.mlp.up_proj.scales": "model.safetensors",
|
| 359 |
+
"model.layers.24.mlp.up_proj.weight": "model.safetensors",
|
| 360 |
+
"model.layers.24.post_attention_layernorm.weight": "model.safetensors",
|
| 361 |
+
"model.layers.24.self_attn.g_proj.biases": "model.safetensors",
|
| 362 |
+
"model.layers.24.self_attn.g_proj.scales": "model.safetensors",
|
| 363 |
+
"model.layers.24.self_attn.g_proj.weight": "model.safetensors",
|
| 364 |
+
"model.layers.24.self_attn.out_proj.biases": "model.safetensors",
|
| 365 |
+
"model.layers.24.self_attn.out_proj.scales": "model.safetensors",
|
| 366 |
+
"model.layers.24.self_attn.out_proj.weight": "model.safetensors",
|
| 367 |
+
"model.layers.24.self_attn.q_k_v_proj.biases": "model.safetensors",
|
| 368 |
+
"model.layers.24.self_attn.q_k_v_proj.scales": "model.safetensors",
|
| 369 |
+
"model.layers.24.self_attn.q_k_v_proj.weight": "model.safetensors",
|
| 370 |
+
"model.layers.25.input_layernorm.weight": "model.safetensors",
|
| 371 |
+
"model.layers.25.mlp.down_proj.biases": "model.safetensors",
|
| 372 |
+
"model.layers.25.mlp.down_proj.scales": "model.safetensors",
|
| 373 |
+
"model.layers.25.mlp.down_proj.weight": "model.safetensors",
|
| 374 |
+
"model.layers.25.mlp.gate_proj.biases": "model.safetensors",
|
| 375 |
+
"model.layers.25.mlp.gate_proj.scales": "model.safetensors",
|
| 376 |
+
"model.layers.25.mlp.gate_proj.weight": "model.safetensors",
|
| 377 |
+
"model.layers.25.mlp.up_proj.biases": "model.safetensors",
|
| 378 |
+
"model.layers.25.mlp.up_proj.scales": "model.safetensors",
|
| 379 |
+
"model.layers.25.mlp.up_proj.weight": "model.safetensors",
|
| 380 |
+
"model.layers.25.post_attention_layernorm.weight": "model.safetensors",
|
| 381 |
+
"model.layers.25.self_attn.g_proj.biases": "model.safetensors",
|
| 382 |
+
"model.layers.25.self_attn.g_proj.scales": "model.safetensors",
|
| 383 |
+
"model.layers.25.self_attn.g_proj.weight": "model.safetensors",
|
| 384 |
+
"model.layers.25.self_attn.out_proj.biases": "model.safetensors",
|
| 385 |
+
"model.layers.25.self_attn.out_proj.scales": "model.safetensors",
|
| 386 |
+
"model.layers.25.self_attn.out_proj.weight": "model.safetensors",
|
| 387 |
+
"model.layers.25.self_attn.q_k_v_proj.biases": "model.safetensors",
|
| 388 |
+
"model.layers.25.self_attn.q_k_v_proj.scales": "model.safetensors",
|
| 389 |
+
"model.layers.25.self_attn.q_k_v_proj.weight": "model.safetensors",
|
| 390 |
+
"model.layers.26.input_layernorm.weight": "model.safetensors",
|
| 391 |
+
"model.layers.26.mlp.down_proj.biases": "model.safetensors",
|
| 392 |
+
"model.layers.26.mlp.down_proj.scales": "model.safetensors",
|
| 393 |
+
"model.layers.26.mlp.down_proj.weight": "model.safetensors",
|
| 394 |
+
"model.layers.26.mlp.gate_proj.biases": "model.safetensors",
|
| 395 |
+
"model.layers.26.mlp.gate_proj.scales": "model.safetensors",
|
| 396 |
+
"model.layers.26.mlp.gate_proj.weight": "model.safetensors",
|
| 397 |
+
"model.layers.26.mlp.up_proj.biases": "model.safetensors",
|
| 398 |
+
"model.layers.26.mlp.up_proj.scales": "model.safetensors",
|
| 399 |
+
"model.layers.26.mlp.up_proj.weight": "model.safetensors",
|
| 400 |
+
"model.layers.26.post_attention_layernorm.weight": "model.safetensors",
|
| 401 |
+
"model.layers.26.self_attn.g_proj.biases": "model.safetensors",
|
| 402 |
+
"model.layers.26.self_attn.g_proj.scales": "model.safetensors",
|
| 403 |
+
"model.layers.26.self_attn.g_proj.weight": "model.safetensors",
|
| 404 |
+
"model.layers.26.self_attn.out_proj.biases": "model.safetensors",
|
| 405 |
+
"model.layers.26.self_attn.out_proj.scales": "model.safetensors",
|
| 406 |
+
"model.layers.26.self_attn.out_proj.weight": "model.safetensors",
|
| 407 |
+
"model.layers.26.self_attn.q_k_v_proj.biases": "model.safetensors",
|
| 408 |
+
"model.layers.26.self_attn.q_k_v_proj.scales": "model.safetensors",
|
| 409 |
+
"model.layers.26.self_attn.q_k_v_proj.weight": "model.safetensors",
|
| 410 |
+
"model.layers.27.input_layernorm.weight": "model.safetensors",
|
| 411 |
+
"model.layers.27.mlp.down_proj.biases": "model.safetensors",
|
| 412 |
+
"model.layers.27.mlp.down_proj.scales": "model.safetensors",
|
| 413 |
+
"model.layers.27.mlp.down_proj.weight": "model.safetensors",
|
| 414 |
+
"model.layers.27.mlp.gate_proj.biases": "model.safetensors",
|
| 415 |
+
"model.layers.27.mlp.gate_proj.scales": "model.safetensors",
|
| 416 |
+
"model.layers.27.mlp.gate_proj.weight": "model.safetensors",
|
| 417 |
+
"model.layers.27.mlp.up_proj.biases": "model.safetensors",
|
| 418 |
+
"model.layers.27.mlp.up_proj.scales": "model.safetensors",
|
| 419 |
+
"model.layers.27.mlp.up_proj.weight": "model.safetensors",
|
| 420 |
+
"model.layers.27.post_attention_layernorm.weight": "model.safetensors",
|
| 421 |
+
"model.layers.27.self_attn.g_proj.biases": "model.safetensors",
|
| 422 |
+
"model.layers.27.self_attn.g_proj.scales": "model.safetensors",
|
| 423 |
+
"model.layers.27.self_attn.g_proj.weight": "model.safetensors",
|
| 424 |
+
"model.layers.27.self_attn.out_proj.biases": "model.safetensors",
|
| 425 |
+
"model.layers.27.self_attn.out_proj.scales": "model.safetensors",
|
| 426 |
+
"model.layers.27.self_attn.out_proj.weight": "model.safetensors",
|
| 427 |
+
"model.layers.27.self_attn.q_k_v_proj.biases": "model.safetensors",
|
| 428 |
+
"model.layers.27.self_attn.q_k_v_proj.scales": "model.safetensors",
|
| 429 |
+
"model.layers.27.self_attn.q_k_v_proj.weight": "model.safetensors",
|
| 430 |
+
"model.layers.28.input_layernorm.weight": "model.safetensors",
|
| 431 |
+
"model.layers.28.mlp.down_proj.biases": "model.safetensors",
|
| 432 |
+
"model.layers.28.mlp.down_proj.scales": "model.safetensors",
|
| 433 |
+
"model.layers.28.mlp.down_proj.weight": "model.safetensors",
|
| 434 |
+
"model.layers.28.mlp.gate_proj.biases": "model.safetensors",
|
| 435 |
+
"model.layers.28.mlp.gate_proj.scales": "model.safetensors",
|
| 436 |
+
"model.layers.28.mlp.gate_proj.weight": "model.safetensors",
|
| 437 |
+
"model.layers.28.mlp.up_proj.biases": "model.safetensors",
|
| 438 |
+
"model.layers.28.mlp.up_proj.scales": "model.safetensors",
|
| 439 |
+
"model.layers.28.mlp.up_proj.weight": "model.safetensors",
|
| 440 |
+
"model.layers.28.post_attention_layernorm.weight": "model.safetensors",
|
| 441 |
+
"model.layers.28.self_attn.g_proj.biases": "model.safetensors",
|
| 442 |
+
"model.layers.28.self_attn.g_proj.scales": "model.safetensors",
|
| 443 |
+
"model.layers.28.self_attn.g_proj.weight": "model.safetensors",
|
| 444 |
+
"model.layers.28.self_attn.out_proj.biases": "model.safetensors",
|
| 445 |
+
"model.layers.28.self_attn.out_proj.scales": "model.safetensors",
|
| 446 |
+
"model.layers.28.self_attn.out_proj.weight": "model.safetensors",
|
| 447 |
+
"model.layers.28.self_attn.q_k_v_proj.biases": "model.safetensors",
|
| 448 |
+
"model.layers.28.self_attn.q_k_v_proj.scales": "model.safetensors",
|
| 449 |
+
"model.layers.28.self_attn.q_k_v_proj.weight": "model.safetensors",
|
| 450 |
+
"model.layers.29.input_layernorm.weight": "model.safetensors",
|
| 451 |
+
"model.layers.29.mlp.down_proj.biases": "model.safetensors",
|
| 452 |
+
"model.layers.29.mlp.down_proj.scales": "model.safetensors",
|
| 453 |
+
"model.layers.29.mlp.down_proj.weight": "model.safetensors",
|
| 454 |
+
"model.layers.29.mlp.gate_proj.biases": "model.safetensors",
|
| 455 |
+
"model.layers.29.mlp.gate_proj.scales": "model.safetensors",
|
| 456 |
+
"model.layers.29.mlp.gate_proj.weight": "model.safetensors",
|
| 457 |
+
"model.layers.29.mlp.up_proj.biases": "model.safetensors",
|
| 458 |
+
"model.layers.29.mlp.up_proj.scales": "model.safetensors",
|
| 459 |
+
"model.layers.29.mlp.up_proj.weight": "model.safetensors",
|
| 460 |
+
"model.layers.29.post_attention_layernorm.weight": "model.safetensors",
|
| 461 |
+
"model.layers.29.self_attn.g_proj.biases": "model.safetensors",
|
| 462 |
+
"model.layers.29.self_attn.g_proj.scales": "model.safetensors",
|
| 463 |
+
"model.layers.29.self_attn.g_proj.weight": "model.safetensors",
|
| 464 |
+
"model.layers.29.self_attn.out_proj.biases": "model.safetensors",
|
| 465 |
+
"model.layers.29.self_attn.out_proj.scales": "model.safetensors",
|
| 466 |
+
"model.layers.29.self_attn.out_proj.weight": "model.safetensors",
|
| 467 |
+
"model.layers.29.self_attn.q_k_v_proj.biases": "model.safetensors",
|
| 468 |
+
"model.layers.29.self_attn.q_k_v_proj.scales": "model.safetensors",
|
| 469 |
+
"model.layers.29.self_attn.q_k_v_proj.weight": "model.safetensors",
|
| 470 |
+
"model.layers.3.input_layernorm.weight": "model.safetensors",
|
| 471 |
+
"model.layers.3.mlp.down_proj.biases": "model.safetensors",
|
| 472 |
+
"model.layers.3.mlp.down_proj.scales": "model.safetensors",
|
| 473 |
+
"model.layers.3.mlp.down_proj.weight": "model.safetensors",
|
| 474 |
+
"model.layers.3.mlp.gate_proj.biases": "model.safetensors",
|
| 475 |
+
"model.layers.3.mlp.gate_proj.scales": "model.safetensors",
|
| 476 |
+
"model.layers.3.mlp.gate_proj.weight": "model.safetensors",
|
| 477 |
+
"model.layers.3.mlp.up_proj.biases": "model.safetensors",
|
| 478 |
+
"model.layers.3.mlp.up_proj.scales": "model.safetensors",
|
| 479 |
+
"model.layers.3.mlp.up_proj.weight": "model.safetensors",
|
| 480 |
+
"model.layers.3.post_attention_layernorm.weight": "model.safetensors",
|
| 481 |
+
"model.layers.3.self_attn.g_proj.biases": "model.safetensors",
|
| 482 |
+
"model.layers.3.self_attn.g_proj.scales": "model.safetensors",
|
| 483 |
+
"model.layers.3.self_attn.g_proj.weight": "model.safetensors",
|
| 484 |
+
"model.layers.3.self_attn.out_proj.biases": "model.safetensors",
|
| 485 |
+
"model.layers.3.self_attn.out_proj.scales": "model.safetensors",
|
| 486 |
+
"model.layers.3.self_attn.out_proj.weight": "model.safetensors",
|
| 487 |
+
"model.layers.3.self_attn.q_k_v_proj.biases": "model.safetensors",
|
| 488 |
+
"model.layers.3.self_attn.q_k_v_proj.scales": "model.safetensors",
|
| 489 |
+
"model.layers.3.self_attn.q_k_v_proj.weight": "model.safetensors",
|
| 490 |
+
"model.layers.30.input_layernorm.weight": "model.safetensors",
|
| 491 |
+
"model.layers.30.mlp.down_proj.biases": "model.safetensors",
|
| 492 |
+
"model.layers.30.mlp.down_proj.scales": "model.safetensors",
|
| 493 |
+
"model.layers.30.mlp.down_proj.weight": "model.safetensors",
|
| 494 |
+
"model.layers.30.mlp.gate_proj.biases": "model.safetensors",
|
| 495 |
+
"model.layers.30.mlp.gate_proj.scales": "model.safetensors",
|
| 496 |
+
"model.layers.30.mlp.gate_proj.weight": "model.safetensors",
|
| 497 |
+
"model.layers.30.mlp.up_proj.biases": "model.safetensors",
|
| 498 |
+
"model.layers.30.mlp.up_proj.scales": "model.safetensors",
|
| 499 |
+
"model.layers.30.mlp.up_proj.weight": "model.safetensors",
|
| 500 |
+
"model.layers.30.post_attention_layernorm.weight": "model.safetensors",
|
| 501 |
+
"model.layers.30.self_attn.g_proj.biases": "model.safetensors",
|
| 502 |
+
"model.layers.30.self_attn.g_proj.scales": "model.safetensors",
|
| 503 |
+
"model.layers.30.self_attn.g_proj.weight": "model.safetensors",
|
| 504 |
+
"model.layers.30.self_attn.out_proj.biases": "model.safetensors",
|
| 505 |
+
"model.layers.30.self_attn.out_proj.scales": "model.safetensors",
|
| 506 |
+
"model.layers.30.self_attn.out_proj.weight": "model.safetensors",
|
| 507 |
+
"model.layers.30.self_attn.q_k_v_proj.biases": "model.safetensors",
|
| 508 |
+
"model.layers.30.self_attn.q_k_v_proj.scales": "model.safetensors",
|
| 509 |
+
"model.layers.30.self_attn.q_k_v_proj.weight": "model.safetensors",
|
| 510 |
+
"model.layers.31.input_layernorm.weight": "model.safetensors",
|
| 511 |
+
"model.layers.31.mlp.down_proj.biases": "model.safetensors",
|
| 512 |
+
"model.layers.31.mlp.down_proj.scales": "model.safetensors",
|
| 513 |
+
"model.layers.31.mlp.down_proj.weight": "model.safetensors",
|
| 514 |
+
"model.layers.31.mlp.gate_proj.biases": "model.safetensors",
|
| 515 |
+
"model.layers.31.mlp.gate_proj.scales": "model.safetensors",
|
| 516 |
+
"model.layers.31.mlp.gate_proj.weight": "model.safetensors",
|
| 517 |
+
"model.layers.31.mlp.up_proj.biases": "model.safetensors",
|
| 518 |
+
"model.layers.31.mlp.up_proj.scales": "model.safetensors",
|
| 519 |
+
"model.layers.31.mlp.up_proj.weight": "model.safetensors",
|
| 520 |
+
"model.layers.31.post_attention_layernorm.weight": "model.safetensors",
|
| 521 |
+
"model.layers.31.self_attn.g_proj.biases": "model.safetensors",
|
| 522 |
+
"model.layers.31.self_attn.g_proj.scales": "model.safetensors",
|
| 523 |
+
"model.layers.31.self_attn.g_proj.weight": "model.safetensors",
|
| 524 |
+
"model.layers.31.self_attn.out_proj.biases": "model.safetensors",
|
| 525 |
+
"model.layers.31.self_attn.out_proj.scales": "model.safetensors",
|
| 526 |
+
"model.layers.31.self_attn.out_proj.weight": "model.safetensors",
|
| 527 |
+
"model.layers.31.self_attn.q_k_v_proj.biases": "model.safetensors",
|
| 528 |
+
"model.layers.31.self_attn.q_k_v_proj.scales": "model.safetensors",
|
| 529 |
+
"model.layers.31.self_attn.q_k_v_proj.weight": "model.safetensors",
|
| 530 |
+
"model.layers.32.input_layernorm.weight": "model.safetensors",
|
| 531 |
+
"model.layers.32.mlp.down_proj.biases": "model.safetensors",
|
| 532 |
+
"model.layers.32.mlp.down_proj.scales": "model.safetensors",
|
| 533 |
+
"model.layers.32.mlp.down_proj.weight": "model.safetensors",
|
| 534 |
+
"model.layers.32.mlp.gate_proj.biases": "model.safetensors",
|
| 535 |
+
"model.layers.32.mlp.gate_proj.scales": "model.safetensors",
|
| 536 |
+
"model.layers.32.mlp.gate_proj.weight": "model.safetensors",
|
| 537 |
+
"model.layers.32.mlp.up_proj.biases": "model.safetensors",
|
| 538 |
+
"model.layers.32.mlp.up_proj.scales": "model.safetensors",
|
| 539 |
+
"model.layers.32.mlp.up_proj.weight": "model.safetensors",
|
| 540 |
+
"model.layers.32.post_attention_layernorm.weight": "model.safetensors",
|
| 541 |
+
"model.layers.32.self_attn.g_proj.biases": "model.safetensors",
|
| 542 |
+
"model.layers.32.self_attn.g_proj.scales": "model.safetensors",
|
| 543 |
+
"model.layers.32.self_attn.g_proj.weight": "model.safetensors",
|
| 544 |
+
"model.layers.32.self_attn.out_proj.biases": "model.safetensors",
|
| 545 |
+
"model.layers.32.self_attn.out_proj.scales": "model.safetensors",
|
| 546 |
+
"model.layers.32.self_attn.out_proj.weight": "model.safetensors",
|
| 547 |
+
"model.layers.32.self_attn.q_k_v_proj.biases": "model.safetensors",
|
| 548 |
+
"model.layers.32.self_attn.q_k_v_proj.scales": "model.safetensors",
|
| 549 |
+
"model.layers.32.self_attn.q_k_v_proj.weight": "model.safetensors",
|
| 550 |
+
"model.layers.33.input_layernorm.weight": "model.safetensors",
|
| 551 |
+
"model.layers.33.mlp.down_proj.biases": "model.safetensors",
|
| 552 |
+
"model.layers.33.mlp.down_proj.scales": "model.safetensors",
|
| 553 |
+
"model.layers.33.mlp.down_proj.weight": "model.safetensors",
|
| 554 |
+
"model.layers.33.mlp.gate_proj.biases": "model.safetensors",
|
| 555 |
+
"model.layers.33.mlp.gate_proj.scales": "model.safetensors",
|
| 556 |
+
"model.layers.33.mlp.gate_proj.weight": "model.safetensors",
|
| 557 |
+
"model.layers.33.mlp.up_proj.biases": "model.safetensors",
|
| 558 |
+
"model.layers.33.mlp.up_proj.scales": "model.safetensors",
|
| 559 |
+
"model.layers.33.mlp.up_proj.weight": "model.safetensors",
|
| 560 |
+
"model.layers.33.post_attention_layernorm.weight": "model.safetensors",
|
| 561 |
+
"model.layers.33.self_attn.g_proj.biases": "model.safetensors",
|
| 562 |
+
"model.layers.33.self_attn.g_proj.scales": "model.safetensors",
|
| 563 |
+
"model.layers.33.self_attn.g_proj.weight": "model.safetensors",
|
| 564 |
+
"model.layers.33.self_attn.out_proj.biases": "model.safetensors",
|
| 565 |
+
"model.layers.33.self_attn.out_proj.scales": "model.safetensors",
|
| 566 |
+
"model.layers.33.self_attn.out_proj.weight": "model.safetensors",
|
| 567 |
+
"model.layers.33.self_attn.q_k_v_proj.biases": "model.safetensors",
|
| 568 |
+
"model.layers.33.self_attn.q_k_v_proj.scales": "model.safetensors",
|
| 569 |
+
"model.layers.33.self_attn.q_k_v_proj.weight": "model.safetensors",
|
| 570 |
+
"model.layers.34.input_layernorm.weight": "model.safetensors",
|
| 571 |
+
"model.layers.34.mlp.down_proj.biases": "model.safetensors",
|
| 572 |
+
"model.layers.34.mlp.down_proj.scales": "model.safetensors",
|
| 573 |
+
"model.layers.34.mlp.down_proj.weight": "model.safetensors",
|
| 574 |
+
"model.layers.34.mlp.gate_proj.biases": "model.safetensors",
|
| 575 |
+
"model.layers.34.mlp.gate_proj.scales": "model.safetensors",
|
| 576 |
+
"model.layers.34.mlp.gate_proj.weight": "model.safetensors",
|
| 577 |
+
"model.layers.34.mlp.up_proj.biases": "model.safetensors",
|
| 578 |
+
"model.layers.34.mlp.up_proj.scales": "model.safetensors",
|
| 579 |
+
"model.layers.34.mlp.up_proj.weight": "model.safetensors",
|
| 580 |
+
"model.layers.34.post_attention_layernorm.weight": "model.safetensors",
|
| 581 |
+
"model.layers.34.self_attn.g_proj.biases": "model.safetensors",
|
| 582 |
+
"model.layers.34.self_attn.g_proj.scales": "model.safetensors",
|
| 583 |
+
"model.layers.34.self_attn.g_proj.weight": "model.safetensors",
|
| 584 |
+
"model.layers.34.self_attn.out_proj.biases": "model.safetensors",
|
| 585 |
+
"model.layers.34.self_attn.out_proj.scales": "model.safetensors",
|
| 586 |
+
"model.layers.34.self_attn.out_proj.weight": "model.safetensors",
|
| 587 |
+
"model.layers.34.self_attn.q_k_v_proj.biases": "model.safetensors",
|
| 588 |
+
"model.layers.34.self_attn.q_k_v_proj.scales": "model.safetensors",
|
| 589 |
+
"model.layers.34.self_attn.q_k_v_proj.weight": "model.safetensors",
|
| 590 |
+
"model.layers.35.input_layernorm.weight": "model.safetensors",
|
| 591 |
+
"model.layers.35.mlp.down_proj.biases": "model.safetensors",
|
| 592 |
+
"model.layers.35.mlp.down_proj.scales": "model.safetensors",
|
| 593 |
+
"model.layers.35.mlp.down_proj.weight": "model.safetensors",
|
| 594 |
+
"model.layers.35.mlp.gate_proj.biases": "model.safetensors",
|
| 595 |
+
"model.layers.35.mlp.gate_proj.scales": "model.safetensors",
|
| 596 |
+
"model.layers.35.mlp.gate_proj.weight": "model.safetensors",
|
| 597 |
+
"model.layers.35.mlp.up_proj.biases": "model.safetensors",
|
| 598 |
+
"model.layers.35.mlp.up_proj.scales": "model.safetensors",
|
| 599 |
+
"model.layers.35.mlp.up_proj.weight": "model.safetensors",
|
| 600 |
+
"model.layers.35.post_attention_layernorm.weight": "model.safetensors",
|
| 601 |
+
"model.layers.35.self_attn.g_proj.biases": "model.safetensors",
|
| 602 |
+
"model.layers.35.self_attn.g_proj.scales": "model.safetensors",
|
| 603 |
+
"model.layers.35.self_attn.g_proj.weight": "model.safetensors",
|
| 604 |
+
"model.layers.35.self_attn.out_proj.biases": "model.safetensors",
|
| 605 |
+
"model.layers.35.self_attn.out_proj.scales": "model.safetensors",
|
| 606 |
+
"model.layers.35.self_attn.out_proj.weight": "model.safetensors",
|
| 607 |
+
"model.layers.35.self_attn.q_k_v_proj.biases": "model.safetensors",
|
| 608 |
+
"model.layers.35.self_attn.q_k_v_proj.scales": "model.safetensors",
|
| 609 |
+
"model.layers.35.self_attn.q_k_v_proj.weight": "model.safetensors",
|
| 610 |
+
"model.layers.4.input_layernorm.weight": "model.safetensors",
|
| 611 |
+
"model.layers.4.mlp.down_proj.biases": "model.safetensors",
|
| 612 |
+
"model.layers.4.mlp.down_proj.scales": "model.safetensors",
|
| 613 |
+
"model.layers.4.mlp.down_proj.weight": "model.safetensors",
|
| 614 |
+
"model.layers.4.mlp.gate_proj.biases": "model.safetensors",
|
| 615 |
+
"model.layers.4.mlp.gate_proj.scales": "model.safetensors",
|
| 616 |
+
"model.layers.4.mlp.gate_proj.weight": "model.safetensors",
|
| 617 |
+
"model.layers.4.mlp.up_proj.biases": "model.safetensors",
|
| 618 |
+
"model.layers.4.mlp.up_proj.scales": "model.safetensors",
|
| 619 |
+
"model.layers.4.mlp.up_proj.weight": "model.safetensors",
|
| 620 |
+
"model.layers.4.post_attention_layernorm.weight": "model.safetensors",
|
| 621 |
+
"model.layers.4.self_attn.g_proj.biases": "model.safetensors",
|
| 622 |
+
"model.layers.4.self_attn.g_proj.scales": "model.safetensors",
|
| 623 |
+
"model.layers.4.self_attn.g_proj.weight": "model.safetensors",
|
| 624 |
+
"model.layers.4.self_attn.out_proj.biases": "model.safetensors",
|
| 625 |
+
"model.layers.4.self_attn.out_proj.scales": "model.safetensors",
|
| 626 |
+
"model.layers.4.self_attn.out_proj.weight": "model.safetensors",
|
| 627 |
+
"model.layers.4.self_attn.q_k_v_proj.biases": "model.safetensors",
|
| 628 |
+
"model.layers.4.self_attn.q_k_v_proj.scales": "model.safetensors",
|
| 629 |
+
"model.layers.4.self_attn.q_k_v_proj.weight": "model.safetensors",
|
| 630 |
+
"model.layers.5.input_layernorm.weight": "model.safetensors",
|
| 631 |
+
"model.layers.5.mlp.down_proj.biases": "model.safetensors",
|
| 632 |
+
"model.layers.5.mlp.down_proj.scales": "model.safetensors",
|
| 633 |
+
"model.layers.5.mlp.down_proj.weight": "model.safetensors",
|
| 634 |
+
"model.layers.5.mlp.gate_proj.biases": "model.safetensors",
|
| 635 |
+
"model.layers.5.mlp.gate_proj.scales": "model.safetensors",
|
| 636 |
+
"model.layers.5.mlp.gate_proj.weight": "model.safetensors",
|
| 637 |
+
"model.layers.5.mlp.up_proj.biases": "model.safetensors",
|
| 638 |
+
"model.layers.5.mlp.up_proj.scales": "model.safetensors",
|
| 639 |
+
"model.layers.5.mlp.up_proj.weight": "model.safetensors",
|
| 640 |
+
"model.layers.5.post_attention_layernorm.weight": "model.safetensors",
|
| 641 |
+
"model.layers.5.self_attn.g_proj.biases": "model.safetensors",
|
| 642 |
+
"model.layers.5.self_attn.g_proj.scales": "model.safetensors",
|
| 643 |
+
"model.layers.5.self_attn.g_proj.weight": "model.safetensors",
|
| 644 |
+
"model.layers.5.self_attn.out_proj.biases": "model.safetensors",
|
| 645 |
+
"model.layers.5.self_attn.out_proj.scales": "model.safetensors",
|
| 646 |
+
"model.layers.5.self_attn.out_proj.weight": "model.safetensors",
|
| 647 |
+
"model.layers.5.self_attn.q_k_v_proj.biases": "model.safetensors",
|
| 648 |
+
"model.layers.5.self_attn.q_k_v_proj.scales": "model.safetensors",
|
| 649 |
+
"model.layers.5.self_attn.q_k_v_proj.weight": "model.safetensors",
|
| 650 |
+
"model.layers.6.input_layernorm.weight": "model.safetensors",
|
| 651 |
+
"model.layers.6.mlp.down_proj.biases": "model.safetensors",
|
| 652 |
+
"model.layers.6.mlp.down_proj.scales": "model.safetensors",
|
| 653 |
+
"model.layers.6.mlp.down_proj.weight": "model.safetensors",
|
| 654 |
+
"model.layers.6.mlp.gate_proj.biases": "model.safetensors",
|
| 655 |
+
"model.layers.6.mlp.gate_proj.scales": "model.safetensors",
|
| 656 |
+
"model.layers.6.mlp.gate_proj.weight": "model.safetensors",
|
| 657 |
+
"model.layers.6.mlp.up_proj.biases": "model.safetensors",
|
| 658 |
+
"model.layers.6.mlp.up_proj.scales": "model.safetensors",
|
| 659 |
+
"model.layers.6.mlp.up_proj.weight": "model.safetensors",
|
| 660 |
+
"model.layers.6.post_attention_layernorm.weight": "model.safetensors",
|
| 661 |
+
"model.layers.6.self_attn.g_proj.biases": "model.safetensors",
|
| 662 |
+
"model.layers.6.self_attn.g_proj.scales": "model.safetensors",
|
| 663 |
+
"model.layers.6.self_attn.g_proj.weight": "model.safetensors",
|
| 664 |
+
"model.layers.6.self_attn.out_proj.biases": "model.safetensors",
|
| 665 |
+
"model.layers.6.self_attn.out_proj.scales": "model.safetensors",
|
| 666 |
+
"model.layers.6.self_attn.out_proj.weight": "model.safetensors",
|
| 667 |
+
"model.layers.6.self_attn.q_k_v_proj.biases": "model.safetensors",
|
| 668 |
+
"model.layers.6.self_attn.q_k_v_proj.scales": "model.safetensors",
|
| 669 |
+
"model.layers.6.self_attn.q_k_v_proj.weight": "model.safetensors",
|
| 670 |
+
"model.layers.7.input_layernorm.weight": "model.safetensors",
|
| 671 |
+
"model.layers.7.mlp.down_proj.biases": "model.safetensors",
|
| 672 |
+
"model.layers.7.mlp.down_proj.scales": "model.safetensors",
|
| 673 |
+
"model.layers.7.mlp.down_proj.weight": "model.safetensors",
|
| 674 |
+
"model.layers.7.mlp.gate_proj.biases": "model.safetensors",
|
| 675 |
+
"model.layers.7.mlp.gate_proj.scales": "model.safetensors",
|
| 676 |
+
"model.layers.7.mlp.gate_proj.weight": "model.safetensors",
|
| 677 |
+
"model.layers.7.mlp.up_proj.biases": "model.safetensors",
|
| 678 |
+
"model.layers.7.mlp.up_proj.scales": "model.safetensors",
|
| 679 |
+
"model.layers.7.mlp.up_proj.weight": "model.safetensors",
|
| 680 |
+
"model.layers.7.post_attention_layernorm.weight": "model.safetensors",
|
| 681 |
+
"model.layers.7.self_attn.g_proj.biases": "model.safetensors",
|
| 682 |
+
"model.layers.7.self_attn.g_proj.scales": "model.safetensors",
|
| 683 |
+
"model.layers.7.self_attn.g_proj.weight": "model.safetensors",
|
| 684 |
+
"model.layers.7.self_attn.out_proj.biases": "model.safetensors",
|
| 685 |
+
"model.layers.7.self_attn.out_proj.scales": "model.safetensors",
|
| 686 |
+
"model.layers.7.self_attn.out_proj.weight": "model.safetensors",
|
| 687 |
+
"model.layers.7.self_attn.q_k_v_proj.biases": "model.safetensors",
|
| 688 |
+
"model.layers.7.self_attn.q_k_v_proj.scales": "model.safetensors",
|
| 689 |
+
"model.layers.7.self_attn.q_k_v_proj.weight": "model.safetensors",
|
| 690 |
+
"model.layers.8.input_layernorm.weight": "model.safetensors",
|
| 691 |
+
"model.layers.8.mlp.down_proj.biases": "model.safetensors",
|
| 692 |
+
"model.layers.8.mlp.down_proj.scales": "model.safetensors",
|
| 693 |
+
"model.layers.8.mlp.down_proj.weight": "model.safetensors",
|
| 694 |
+
"model.layers.8.mlp.gate_proj.biases": "model.safetensors",
|
| 695 |
+
"model.layers.8.mlp.gate_proj.scales": "model.safetensors",
|
| 696 |
+
"model.layers.8.mlp.gate_proj.weight": "model.safetensors",
|
| 697 |
+
"model.layers.8.mlp.up_proj.biases": "model.safetensors",
|
| 698 |
+
"model.layers.8.mlp.up_proj.scales": "model.safetensors",
|
| 699 |
+
"model.layers.8.mlp.up_proj.weight": "model.safetensors",
|
| 700 |
+
"model.layers.8.post_attention_layernorm.weight": "model.safetensors",
|
| 701 |
+
"model.layers.8.self_attn.g_proj.biases": "model.safetensors",
|
| 702 |
+
"model.layers.8.self_attn.g_proj.scales": "model.safetensors",
|
| 703 |
+
"model.layers.8.self_attn.g_proj.weight": "model.safetensors",
|
| 704 |
+
"model.layers.8.self_attn.out_proj.biases": "model.safetensors",
|
| 705 |
+
"model.layers.8.self_attn.out_proj.scales": "model.safetensors",
|
| 706 |
+
"model.layers.8.self_attn.out_proj.weight": "model.safetensors",
|
| 707 |
+
"model.layers.8.self_attn.q_k_v_proj.biases": "model.safetensors",
|
| 708 |
+
"model.layers.8.self_attn.q_k_v_proj.scales": "model.safetensors",
|
| 709 |
+
"model.layers.8.self_attn.q_k_v_proj.weight": "model.safetensors",
|
| 710 |
+
"model.layers.9.input_layernorm.weight": "model.safetensors",
|
| 711 |
+
"model.layers.9.mlp.down_proj.biases": "model.safetensors",
|
| 712 |
+
"model.layers.9.mlp.down_proj.scales": "model.safetensors",
|
| 713 |
+
"model.layers.9.mlp.down_proj.weight": "model.safetensors",
|
| 714 |
+
"model.layers.9.mlp.gate_proj.biases": "model.safetensors",
|
| 715 |
+
"model.layers.9.mlp.gate_proj.scales": "model.safetensors",
|
| 716 |
+
"model.layers.9.mlp.gate_proj.weight": "model.safetensors",
|
| 717 |
+
"model.layers.9.mlp.up_proj.biases": "model.safetensors",
|
| 718 |
+
"model.layers.9.mlp.up_proj.scales": "model.safetensors",
|
| 719 |
+
"model.layers.9.mlp.up_proj.weight": "model.safetensors",
|
| 720 |
+
"model.layers.9.post_attention_layernorm.weight": "model.safetensors",
|
| 721 |
+
"model.layers.9.self_attn.g_proj.biases": "model.safetensors",
|
| 722 |
+
"model.layers.9.self_attn.g_proj.scales": "model.safetensors",
|
| 723 |
+
"model.layers.9.self_attn.g_proj.weight": "model.safetensors",
|
| 724 |
+
"model.layers.9.self_attn.out_proj.biases": "model.safetensors",
|
| 725 |
+
"model.layers.9.self_attn.out_proj.scales": "model.safetensors",
|
| 726 |
+
"model.layers.9.self_attn.out_proj.weight": "model.safetensors",
|
| 727 |
+
"model.layers.9.self_attn.q_k_v_proj.biases": "model.safetensors",
|
| 728 |
+
"model.layers.9.self_attn.q_k_v_proj.scales": "model.safetensors",
|
| 729 |
+
"model.layers.9.self_attn.q_k_v_proj.weight": "model.safetensors",
|
| 730 |
+
"model.norm.weight": "model.safetensors"
|
| 731 |
+
}
|
| 732 |
+
}
|
modeling_spark.py
ADDED
|
@@ -0,0 +1,483 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import math
|
| 2 |
+
|
| 3 |
+
import torch
|
| 4 |
+
import torch.nn.functional as F
|
| 5 |
+
from torch import nn
|
| 6 |
+
from transformers.activations import ACT2FN
|
| 7 |
+
from transformers.cache_utils import Cache, DynamicCache
|
| 8 |
+
from transformers.generation import GenerationMixin
|
| 9 |
+
from transformers.masking_utils import create_causal_mask, create_sliding_window_causal_mask
|
| 10 |
+
from transformers.modeling_outputs import (
|
| 11 |
+
BaseModelOutputWithPast,
|
| 12 |
+
CausalLMOutputWithPast,
|
| 13 |
+
)
|
| 14 |
+
from transformers.modeling_utils import PreTrainedModel
|
| 15 |
+
from transformers.processing_utils import Unpack
|
| 16 |
+
from transformers.pytorch_utils import ALL_LAYERNORM_LAYERS
|
| 17 |
+
from transformers.utils import TransformersKwargs, can_return_tuple, logging
|
| 18 |
+
|
| 19 |
+
from .configuration_spark import Spark2_5Config
|
| 20 |
+
|
| 21 |
+
logger = logging.get_logger(__name__)
|
| 22 |
+
|
| 23 |
+
_CONFIG_FOR_DOC = "Spark2_5Config"
|
| 24 |
+
|
| 25 |
+
def rotate_half(x):
|
| 26 |
+
x1 = x[..., : x.shape[-1] // 2]
|
| 27 |
+
x2 = x[..., x.shape[-1] // 2 :]
|
| 28 |
+
return torch.cat((-x2, x1), dim=-1)
|
| 29 |
+
|
| 30 |
+
|
| 31 |
+
def compute_rope_cos_sin(positions, head_dim, rope_theta, partial_rotary_factor=1.0, device="cpu"):
|
| 32 |
+
rope_head_dim = int(head_dim * partial_rotary_factor)
|
| 33 |
+
inv_freq = 1.0 / (rope_theta ** (torch.arange(0, rope_head_dim, 2, dtype=torch.int64).to(device="cpu", dtype=torch.float) / rope_head_dim))
|
| 34 |
+
inv_freq = inv_freq.to(device)
|
| 35 |
+
t = positions.to(device=device, dtype=torch.float32)
|
| 36 |
+
freqs = torch.outer(t, inv_freq)
|
| 37 |
+
freqs = torch.cat([freqs, freqs], dim=-1)
|
| 38 |
+
cos = freqs.cos()
|
| 39 |
+
sin = freqs.sin()
|
| 40 |
+
return cos, sin
|
| 41 |
+
|
| 42 |
+
|
| 43 |
+
def apply_rotary_pos_emb(x, cos, sin):
|
| 44 |
+
rope_head_dim = cos.shape[-1]
|
| 45 |
+
x_f32 = x.float()
|
| 46 |
+
if x_f32.shape[-1] > rope_head_dim:
|
| 47 |
+
x_rot = x_f32[..., :rope_head_dim]
|
| 48 |
+
x_pass = x_f32[..., rope_head_dim:]
|
| 49 |
+
c = cos.unsqueeze(0).unsqueeze(0)
|
| 50 |
+
s = sin.unsqueeze(0).unsqueeze(0)
|
| 51 |
+
x_rot = x_rot * c + rotate_half(x_rot) * s
|
| 52 |
+
result = torch.cat([x_rot, x_pass], dim=-1)
|
| 53 |
+
else:
|
| 54 |
+
c = cos.unsqueeze(0).unsqueeze(0)
|
| 55 |
+
s = sin.unsqueeze(0).unsqueeze(0)
|
| 56 |
+
result = x_f32 * c + rotate_half(x_f32) * s
|
| 57 |
+
return result.to(x.dtype)
|
| 58 |
+
|
| 59 |
+
|
| 60 |
+
def repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor:
|
| 61 |
+
batch, num_key_value_heads, slen, head_dim = hidden_states.shape
|
| 62 |
+
if n_rep == 1:
|
| 63 |
+
return hidden_states
|
| 64 |
+
hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)
|
| 65 |
+
return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)
|
| 66 |
+
|
| 67 |
+
|
| 68 |
+
def eager_attention_forward(
|
| 69 |
+
module: nn.Module,
|
| 70 |
+
query: torch.Tensor,
|
| 71 |
+
key: torch.Tensor,
|
| 72 |
+
value: torch.Tensor,
|
| 73 |
+
attention_mask: torch.Tensor | None = None,
|
| 74 |
+
scaling: float | None = None,
|
| 75 |
+
dropout: float = 0.0,
|
| 76 |
+
**kwargs: Unpack[TransformersKwargs],
|
| 77 |
+
):
|
| 78 |
+
key = repeat_kv(key, module.num_key_value_groups)
|
| 79 |
+
value = repeat_kv(value, module.num_key_value_groups)
|
| 80 |
+
|
| 81 |
+
if scaling is None:
|
| 82 |
+
scaling = 1.0 / math.sqrt(query.shape[-1])
|
| 83 |
+
|
| 84 |
+
attn_weights = torch.matmul(query, key.transpose(2, 3)) * scaling
|
| 85 |
+
if attention_mask is not None:
|
| 86 |
+
causal_mask = attention_mask[:, :, :, : key.shape[-2]]
|
| 87 |
+
attn_weights = attn_weights + causal_mask
|
| 88 |
+
|
| 89 |
+
attn_weights = attn_weights - attn_weights.max(dim=-1, keepdim=True).values
|
| 90 |
+
attn_weights = F.softmax(attn_weights, dim=-1, dtype=torch.float32).to(query.dtype)
|
| 91 |
+
attn_weights = nn.functional.dropout(attn_weights, p=dropout, training=module.training)
|
| 92 |
+
attn_output = torch.matmul(attn_weights, value)
|
| 93 |
+
return attn_output, attn_weights
|
| 94 |
+
|
| 95 |
+
|
| 96 |
+
class Spark2_5RMSNorm(nn.Module):
|
| 97 |
+
def __init__(self, hidden_size, eps=1e-6):
|
| 98 |
+
super().__init__()
|
| 99 |
+
self.weight = nn.Parameter(torch.ones(hidden_size))
|
| 100 |
+
self.variance_epsilon = eps
|
| 101 |
+
|
| 102 |
+
def forward(self, hidden_states):
|
| 103 |
+
input_dtype = hidden_states.dtype
|
| 104 |
+
hidden_states = hidden_states.to(torch.float32)
|
| 105 |
+
variance = hidden_states.pow(2).mean(-1, keepdim=True)
|
| 106 |
+
hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)
|
| 107 |
+
return (self.weight.float() * hidden_states).to(input_dtype)
|
| 108 |
+
|
| 109 |
+
def extra_repr(self):
|
| 110 |
+
return f"{tuple(self.weight.shape)}, eps={self.variance_epsilon}"
|
| 111 |
+
|
| 112 |
+
|
| 113 |
+
ALL_LAYERNORM_LAYERS.append(Spark2_5RMSNorm)
|
| 114 |
+
|
| 115 |
+
|
| 116 |
+
class Spark2_5MLP(nn.Module):
|
| 117 |
+
def __init__(self, config):
|
| 118 |
+
super().__init__()
|
| 119 |
+
self.config = config
|
| 120 |
+
self.hidden_size = config.hidden_size
|
| 121 |
+
self.intermediate_size = config.intermediate_size
|
| 122 |
+
self.gate_proj = nn.Linear(self.hidden_size, self.intermediate_size, bias=config.mlp_bias)
|
| 123 |
+
self.up_proj = nn.Linear(self.hidden_size, self.intermediate_size, bias=config.mlp_bias)
|
| 124 |
+
self.down_proj = nn.Linear(self.intermediate_size, self.hidden_size, bias=config.mlp_bias)
|
| 125 |
+
|
| 126 |
+
if config.hidden_act != "gelu":
|
| 127 |
+
raise ValueError(f"只支持hidden_act='gelu'，当前传入:{config.hidden_act}")
|
| 128 |
+
|
| 129 |
+
self.act_fn = ACT2FN[config.hidden_act]
|
| 130 |
+
|
| 131 |
+
def forward(self, x):
|
| 132 |
+
return self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))
|
| 133 |
+
|
| 134 |
+
|
| 135 |
+
class Spark2_5Attention(nn.Module):
|
| 136 |
+
def __init__(self, config: Spark2_5Config, layer_idx: int | None = None):
|
| 137 |
+
super().__init__()
|
| 138 |
+
self.config = config
|
| 139 |
+
self.layer_idx = layer_idx
|
| 140 |
+
self.attention_dropout = config.attention_dropout
|
| 141 |
+
self.hidden_size = config.hidden_size
|
| 142 |
+
self.num_heads = config.num_attention_heads
|
| 143 |
+
self.head_dim = config.head_dim
|
| 144 |
+
self.num_key_value_heads = config.num_key_value_heads
|
| 145 |
+
self.num_key_value_groups = self.num_heads // self.num_key_value_heads
|
| 146 |
+
self.scaling = 1.0 / math.sqrt(self.head_dim)
|
| 147 |
+
self.headwise_attn_output_gate = config.headwise_attn_output_gate
|
| 148 |
+
self.gate_attn_act_mode = config.gate_attn_act_mode
|
| 149 |
+
self.q_dim = self.num_heads * self.head_dim
|
| 150 |
+
self.kv_dim = self.num_key_value_heads * self.head_dim
|
| 151 |
+
|
| 152 |
+
qkv_out_dim = self.q_dim + 2 * self.kv_dim
|
| 153 |
+
self.q_k_v_proj = nn.Linear(self.hidden_size, qkv_out_dim, bias=config.attention_bias)
|
| 154 |
+
self.g_proj = nn.Linear(self.hidden_size, self.num_heads, bias=config.attention_bias) if self.headwise_attn_output_gate else None
|
| 155 |
+
self.out_proj = nn.Linear(self.num_heads * self.head_dim, self.hidden_size, bias=config.attention_bias)
|
| 156 |
+
self.sliding_window = None
|
| 157 |
+
|
| 158 |
+
def forward(
|
| 159 |
+
self,
|
| 160 |
+
hidden_states: torch.Tensor,
|
| 161 |
+
position_embeddings: tuple[torch.Tensor, torch.Tensor],
|
| 162 |
+
attention_mask: torch.Tensor | None = None,
|
| 163 |
+
past_key_values: Cache | None = None,
|
| 164 |
+
cache_position: torch.LongTensor | None = None,
|
| 165 |
+
**kwargs: Unpack[TransformersKwargs],
|
| 166 |
+
) -> tuple[torch.Tensor, torch.Tensor]:
|
| 167 |
+
input_shape = hidden_states.shape[:-1]
|
| 168 |
+
bsz, seq_len = input_shape
|
| 169 |
+
|
| 170 |
+
qkv = self.q_k_v_proj(hidden_states)
|
| 171 |
+
q = qkv[..., :self.q_dim]
|
| 172 |
+
k = qkv[..., self.q_dim:self.q_dim + self.kv_dim]
|
| 173 |
+
v = qkv[..., self.q_dim + self.kv_dim:]
|
| 174 |
+
gate_score = self.g_proj(hidden_states) if self.g_proj is not None else None
|
| 175 |
+
|
| 176 |
+
q = q.view(bsz, seq_len, self.num_heads, self.head_dim).transpose(1, 2)
|
| 177 |
+
k = k.view(bsz, seq_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
|
| 178 |
+
v = v.view(bsz, seq_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
|
| 179 |
+
if gate_score is not None:
|
| 180 |
+
gate_score = gate_score.view(bsz, seq_len, self.num_heads, 1).transpose(1, 2)
|
| 181 |
+
|
| 182 |
+
cos, sin = position_embeddings
|
| 183 |
+
q = apply_rotary_pos_emb(q, cos, sin)
|
| 184 |
+
k = apply_rotary_pos_emb(k, cos, sin)
|
| 185 |
+
|
| 186 |
+
|
| 187 |
+
if past_key_values is not None:
|
| 188 |
+
cache_kwargs = {"sin": sin, "cos": cos, "cache_position": cache_position}
|
| 189 |
+
k, v = past_key_values.update(k, v, self.layer_idx, cache_kwargs)
|
| 190 |
+
|
| 191 |
+
attn_output, attn_weights = eager_attention_forward(
|
| 192 |
+
self, q, k, v,
|
| 193 |
+
attention_mask=attention_mask,
|
| 194 |
+
scaling=self.scaling,
|
| 195 |
+
dropout=self.attention_dropout if self.training else 0.0,
|
| 196 |
+
)
|
| 197 |
+
|
| 198 |
+
if gate_score is not None:
|
| 199 |
+
if self.gate_attn_act_mode == "sigmoid":
|
| 200 |
+
gate = torch.sigmoid(gate_score.float())
|
| 201 |
+
elif self.gate_attn_act_mode == "silu":
|
| 202 |
+
gate = F.silu(gate_score.float())
|
| 203 |
+
else:
|
| 204 |
+
raise ValueError(f"Unsupported gate_attn_act_mode: {self.gate_attn_act_mode}")
|
| 205 |
+
gate = gate.to(attn_output.dtype)
|
| 206 |
+
attn_output = attn_output * gate
|
| 207 |
+
|
| 208 |
+
attn_output = attn_output.transpose(1, 2).contiguous().view(bsz, seq_len, -1)
|
| 209 |
+
attn_output = self.out_proj(attn_output)
|
| 210 |
+
|
| 211 |
+
return attn_output, attn_weights
|
| 212 |
+
|
| 213 |
+
|
| 214 |
+
class Spark2_5DecoderLayer(nn.Module):
|
| 215 |
+
def __init__(self, config: Spark2_5Config, layer_idx: int):
|
| 216 |
+
super().__init__()
|
| 217 |
+
self.hidden_size = config.hidden_size
|
| 218 |
+
|
| 219 |
+
self.self_attn = Spark2_5Attention(config=config, layer_idx=layer_idx)
|
| 220 |
+
self.mlp = Spark2_5MLP(config)
|
| 221 |
+
self.input_layernorm = Spark2_5RMSNorm(config.hidden_size, eps=config.rms_norm_eps)
|
| 222 |
+
self.post_attention_layernorm = Spark2_5RMSNorm(config.hidden_size, eps=config.rms_norm_eps)
|
| 223 |
+
|
| 224 |
+
self.layer_type = config.layer_types[layer_idx] if layer_idx < len(config.layer_types) else "full_attention"
|
| 225 |
+
if self.layer_type == "sliding_attention" and config.sliding_window is not None:
|
| 226 |
+
self.self_attn.sliding_window = config.sliding_window
|
| 227 |
+
else:
|
| 228 |
+
self.self_attn.sliding_window = None
|
| 229 |
+
self.self_attn.partial_rotary_factor = config.get_partial_rotary_factor(self.layer_type)
|
| 230 |
+
|
| 231 |
+
def forward(
|
| 232 |
+
self,
|
| 233 |
+
hidden_states: torch.Tensor,
|
| 234 |
+
position_embeddings: tuple[torch.Tensor, torch.Tensor],
|
| 235 |
+
attention_mask: torch.Tensor | None = None,
|
| 236 |
+
past_key_values: Cache | None = None,
|
| 237 |
+
cache_position: torch.LongTensor | None = None,
|
| 238 |
+
position_ids: torch.LongTensor | None = None,
|
| 239 |
+
**kwargs: Unpack[TransformersKwargs]
|
| 240 |
+
) -> torch.Tensor:
|
| 241 |
+
|
| 242 |
+
residual = hidden_states
|
| 243 |
+
hidden_states = self.input_layernorm(hidden_states)
|
| 244 |
+
hidden_states = hidden_states.to(self.mlp.gate_proj.weight.dtype)
|
| 245 |
+
|
| 246 |
+
hidden_states, _ = self.self_attn(
|
| 247 |
+
hidden_states=hidden_states,
|
| 248 |
+
position_embeddings=position_embeddings,
|
| 249 |
+
attention_mask=attention_mask,
|
| 250 |
+
past_key_values=past_key_values,
|
| 251 |
+
cache_position=cache_position,
|
| 252 |
+
position_ids=position_ids,
|
| 253 |
+
)
|
| 254 |
+
hidden_states = residual + hidden_states
|
| 255 |
+
|
| 256 |
+
residual = hidden_states
|
| 257 |
+
hidden_states = self.post_attention_layernorm(hidden_states)
|
| 258 |
+
hidden_states = hidden_states.to(self.mlp.gate_proj.weight.dtype)
|
| 259 |
+
|
| 260 |
+
|
| 261 |
+
hidden_states = self.mlp(hidden_states)
|
| 262 |
+
hidden_states = residual + hidden_states
|
| 263 |
+
|
| 264 |
+
return hidden_states
|
| 265 |
+
|
| 266 |
+
|
| 267 |
+
class Spark2_5PreTrainedModel(PreTrainedModel):
|
| 268 |
+
config_class = Spark2_5Config
|
| 269 |
+
base_model_prefix = "model"
|
| 270 |
+
supports_gradient_checkpointing = True
|
| 271 |
+
_no_split_modules = ["Spark2_5DecoderLayer"] # noqa: RUF012
|
| 272 |
+
_skip_keys_device_placement = ["past_key_values"] # noqa: RUF012
|
| 273 |
+
|
| 274 |
+
def _init_weights(self, module):
|
| 275 |
+
std = self.config.initializer_range
|
| 276 |
+
if isinstance(module, nn.Linear):
|
| 277 |
+
module.weight.data.normal_(mean=0.0, std=std)
|
| 278 |
+
if module.bias is not None:
|
| 279 |
+
module.bias.data.zero_()
|
| 280 |
+
elif isinstance(module, nn.Embedding):
|
| 281 |
+
module.weight.data.normal_(mean=0.0, std=std)
|
| 282 |
+
if module.padding_idx is not None:
|
| 283 |
+
module.weight.data[module.padding_idx].zero_()
|
| 284 |
+
|
| 285 |
+
|
| 286 |
+
class Spark2_5Model(Spark2_5PreTrainedModel):
|
| 287 |
+
def __init__(self, config: Spark2_5Config):
|
| 288 |
+
super().__init__(config)
|
| 289 |
+
self.padding_idx = config.pad_token_id
|
| 290 |
+
self.vocab_size = config.vocab_size
|
| 291 |
+
|
| 292 |
+
self.embedding = nn.Embedding(config.vocab_size, config.hidden_size, self.padding_idx)
|
| 293 |
+
self.layers = nn.ModuleList(
|
| 294 |
+
[Spark2_5DecoderLayer(config, layer_idx) for layer_idx in range(config.num_hidden_layers)]
|
| 295 |
+
)
|
| 296 |
+
self.norm = Spark2_5RMSNorm(config.hidden_size, eps=config.rms_norm_eps)
|
| 297 |
+
self.gradient_checkpointing = False
|
| 298 |
+
self.has_sliding_layers = "sliding_attention" in config.layer_types
|
| 299 |
+
|
| 300 |
+
self.post_init()
|
| 301 |
+
|
| 302 |
+
def get_input_embeddings(self):
|
| 303 |
+
return self.embedding
|
| 304 |
+
|
| 305 |
+
def set_input_embeddings(self, value):
|
| 306 |
+
self.embedding = value
|
| 307 |
+
def forward(
|
| 308 |
+
self,
|
| 309 |
+
input_ids: torch.LongTensor = None,
|
| 310 |
+
attention_mask: torch.Tensor | None = None,
|
| 311 |
+
position_ids: torch.LongTensor | None = None,
|
| 312 |
+
past_key_values: Cache | list[torch.FloatTensor] | None = None,
|
| 313 |
+
inputs_embeds: torch.FloatTensor | None = None,
|
| 314 |
+
use_cache: bool | None = None,
|
| 315 |
+
cache_position: torch.LongTensor | None = None,
|
| 316 |
+
token_type_ids: torch.LongTensor | None = None,
|
| 317 |
+
**kwargs: Unpack[TransformersKwargs],
|
| 318 |
+
) -> BaseModelOutputWithPast:
|
| 319 |
+
use_cache = use_cache if use_cache is not None else self.config.use_cache
|
| 320 |
+
if (input_ids is None) ^ (inputs_embeds is not None):
|
| 321 |
+
raise ValueError(
|
| 322 |
+
"You cannot specify both input_ids and inputs_embeds at the same time, and must specify either one"
|
| 323 |
+
)
|
| 324 |
+
|
| 325 |
+
if self.gradient_checkpointing and self.training and use_cache:
|
| 326 |
+
logger.warning_once(
|
| 327 |
+
"`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`."
|
| 328 |
+
)
|
| 329 |
+
use_cache = False
|
| 330 |
+
|
| 331 |
+
if inputs_embeds is None:
|
| 332 |
+
inputs_embeds = self.embedding(input_ids)
|
| 333 |
+
|
| 334 |
+
if use_cache and past_key_values is None:
|
| 335 |
+
past_key_values = DynamicCache(config=self.config)
|
| 336 |
+
|
| 337 |
+
if cache_position is None:
|
| 338 |
+
past_seen_tokens = past_key_values.get_seq_length() if past_key_values is not None else 0
|
| 339 |
+
cache_position = torch.arange(
|
| 340 |
+
past_seen_tokens, past_seen_tokens + inputs_embeds.shape[1], device=inputs_embeds.device
|
| 341 |
+
)
|
| 342 |
+
|
| 343 |
+
if position_ids is None:
|
| 344 |
+
position_ids = cache_position.unsqueeze(0)
|
| 345 |
+
|
| 346 |
+
if not isinstance(attention_mask, dict):
|
| 347 |
+
mask_kwargs = {
|
| 348 |
+
"config": self.config,
|
| 349 |
+
"input_embeds": inputs_embeds,
|
| 350 |
+
"attention_mask": attention_mask,
|
| 351 |
+
"cache_position": cache_position,
|
| 352 |
+
"past_key_values": past_key_values,
|
| 353 |
+
"position_ids": position_ids,
|
| 354 |
+
}
|
| 355 |
+
causal_mask_mapping = {
|
| 356 |
+
"full_attention": create_causal_mask(**mask_kwargs),
|
| 357 |
+
}
|
| 358 |
+
if self.has_sliding_layers:
|
| 359 |
+
causal_mask_mapping["sliding_attention"] = create_sliding_window_causal_mask(**mask_kwargs)
|
| 360 |
+
else:
|
| 361 |
+
causal_mask_mapping = attention_mask
|
| 362 |
+
|
| 363 |
+
hidden_states = inputs_embeds.float()
|
| 364 |
+
|
| 365 |
+
device = hidden_states.device
|
| 366 |
+
dtype = self.embedding.weight.dtype
|
| 367 |
+
|
| 368 |
+
head_dim = self.config.head_dim
|
| 369 |
+
rope_cache = {}
|
| 370 |
+
for lt in set(self.config.layer_types):
|
| 371 |
+
rope_theta = self.config.get_rope_theta(lt)
|
| 372 |
+
prf = self.config.get_partial_rotary_factor(lt)
|
| 373 |
+
cos, sin = compute_rope_cos_sin(cache_position, head_dim, rope_theta, partial_rotary_factor=prf, device=device)
|
| 374 |
+
rope_cache[lt] = (cos, sin)
|
| 375 |
+
|
| 376 |
+
for decoder_layer in self.layers:
|
| 377 |
+
layer_type = decoder_layer.layer_type
|
| 378 |
+
position_embeddings = rope_cache.get(layer_type, rope_cache.get("full_attention"))
|
| 379 |
+
layer_attention_mask = causal_mask_mapping.get(layer_type, causal_mask_mapping.get("full_attention"))
|
| 380 |
+
|
| 381 |
+
if self.gradient_checkpointing and self.training:
|
| 382 |
+
layer_outputs = self._gradient_checkpointing_func(
|
| 383 |
+
decoder_layer.__call__,
|
| 384 |
+
hidden_states,
|
| 385 |
+
position_embeddings,
|
| 386 |
+
layer_attention_mask,
|
| 387 |
+
)
|
| 388 |
+
hidden_states = layer_outputs[0] if isinstance(layer_outputs, tuple) else layer_outputs
|
| 389 |
+
else:
|
| 390 |
+
hidden_states = decoder_layer(
|
| 391 |
+
hidden_states,
|
| 392 |
+
position_embeddings=position_embeddings,
|
| 393 |
+
attention_mask=layer_attention_mask,
|
| 394 |
+
past_key_values=past_key_values,
|
| 395 |
+
cache_position=cache_position,
|
| 396 |
+
position_ids=position_ids,
|
| 397 |
+
)
|
| 398 |
+
|
| 399 |
+
hidden_states = self.norm(hidden_states)
|
| 400 |
+
hidden_states = hidden_states.to(dtype)
|
| 401 |
+
|
| 402 |
+
return BaseModelOutputWithPast(
|
| 403 |
+
last_hidden_state=hidden_states,
|
| 404 |
+
past_key_values=past_key_values if use_cache else None,
|
| 405 |
+
)
|
| 406 |
+
|
| 407 |
+
class Spark2_5ForCausalLM(Spark2_5PreTrainedModel, GenerationMixin):
|
| 408 |
+
_tied_weights_keys = ["lm_head.weight"] # noqa: RUF012
|
| 409 |
+
|
| 410 |
+
def __init__(self, config):
|
| 411 |
+
super().__init__(config)
|
| 412 |
+
self.model = Spark2_5Model(config)
|
| 413 |
+
self.vocab_size = config.vocab_size
|
| 414 |
+
self.lm_head = nn.Linear(config.hidden_size, config.vocab_size, bias=False)
|
| 415 |
+
self.post_init()
|
| 416 |
+
|
| 417 |
+
def get_input_embeddings(self):
|
| 418 |
+
return self.model.embedding
|
| 419 |
+
|
| 420 |
+
def set_input_embeddings(self, value):
|
| 421 |
+
self.model.embedding = value
|
| 422 |
+
|
| 423 |
+
def get_output_embeddings(self):
|
| 424 |
+
return self.lm_head
|
| 425 |
+
|
| 426 |
+
def set_output_embeddings(self, new_embeddings):
|
| 427 |
+
self.lm_head = new_embeddings
|
| 428 |
+
|
| 429 |
+
def set_decoder(self, decoder):
|
| 430 |
+
self.model = decoder
|
| 431 |
+
|
| 432 |
+
def get_decoder(self):
|
| 433 |
+
return self.model
|
| 434 |
+
|
| 435 |
+
@can_return_tuple
|
| 436 |
+
def forward(
|
| 437 |
+
self,
|
| 438 |
+
input_ids: torch.LongTensor = None,
|
| 439 |
+
attention_mask: torch.Tensor | None = None,
|
| 440 |
+
position_ids: torch.LongTensor | None = None,
|
| 441 |
+
past_key_values: Cache | list[torch.FloatTensor] | None = None,
|
| 442 |
+
inputs_embeds: torch.FloatTensor | None = None,
|
| 443 |
+
labels: torch.LongTensor | None = None,
|
| 444 |
+
use_cache: bool | None = None,
|
| 445 |
+
cache_position: torch.LongTensor | None = None,
|
| 446 |
+
logits_to_keep: int = 0,
|
| 447 |
+
token_type_ids: torch.LongTensor | None = None,
|
| 448 |
+
**kwargs: Unpack[TransformersKwargs],
|
| 449 |
+
) -> CausalLMOutputWithPast:
|
| 450 |
+
outputs: BaseModelOutputWithPast = self.model(
|
| 451 |
+
input_ids=input_ids,
|
| 452 |
+
attention_mask=attention_mask,
|
| 453 |
+
position_ids=position_ids,
|
| 454 |
+
past_key_values=past_key_values,
|
| 455 |
+
inputs_embeds=inputs_embeds,
|
| 456 |
+
use_cache=use_cache,
|
| 457 |
+
cache_position=cache_position,
|
| 458 |
+
)
|
| 459 |
+
|
| 460 |
+
hidden_states = outputs.last_hidden_state
|
| 461 |
+
slice_indices = slice(-logits_to_keep, None) if isinstance(logits_to_keep, int) else logits_to_keep
|
| 462 |
+
hidden_states = hidden_states[:, slice_indices, :]
|
| 463 |
+
|
| 464 |
+
if self.config.tie_word_embeddings:
|
| 465 |
+
embed_weight = self.model.embedding.weight
|
| 466 |
+
logits = F.linear(hidden_states, embed_weight)
|
| 467 |
+
else:
|
| 468 |
+
logits = self.lm_head(hidden_states)
|
| 469 |
+
|
| 470 |
+
loss = None
|
| 471 |
+
if labels is not None:
|
| 472 |
+
loss = self.loss_function(logits=logits, labels=labels, vocab_size=self.config.vocab_size, **kwargs)
|
| 473 |
+
|
| 474 |
+
return CausalLMOutputWithPast(
|
| 475 |
+
loss=loss,
|
| 476 |
+
logits=logits,
|
| 477 |
+
past_key_values=outputs.past_key_values,
|
| 478 |
+
hidden_states=outputs.hidden_states,
|
| 479 |
+
attentions=outputs.attentions,
|
| 480 |
+
)
|
| 481 |
+
|
| 482 |
+
|
| 483 |
+
__all__ = ["Spark2_5Config", "Spark2_5ForCausalLM", "Spark2_5Model"]
|
tokenizer.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
tokenizer_config.json
ADDED
|
@@ -0,0 +1,14 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"backend": "tokenizers",
|
| 3 |
+
"bos_token": "<｜start▁of▁sentence｜>",
|
| 4 |
+
"clean_up_tokenization_spaces": false,
|
| 5 |
+
"eos_token": "<｜end▁of▁sentence｜>",
|
| 6 |
+
"is_local": true,
|
| 7 |
+
"legacy": true,
|
| 8 |
+
"local_files_only": false,
|
| 9 |
+
"model_max_length": 131072,
|
| 10 |
+
"pad_token": "<｜▁pad▁｜>",
|
| 11 |
+
"sp_model_kwargs": {},
|
| 12 |
+
"tokenizer_class": "TokenizersBackend",
|
| 13 |
+
"unk_token": "<unk>"
|
| 14 |
+
}
|