ZarubinA commited on
Commit
0cb52ef
·
0 Parent(s):

Initial VariantAssist Gemma 4 31B GGUF release

Browse files
.gitattributes ADDED
@@ -0,0 +1,42 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
36
+ assets/accuracy_vs_reasoning_tokens.png filter=lfs diff=lfs merge=lfs -text
37
+ assets/confusion_matrix_variantassist_ud_q8.png filter=lfs diff=lfs merge=lfs -text
38
+ gemma-4-31B-it.BF16-mmproj.gguf filter=lfs diff=lfs merge=lfs -text
39
+ VA-gemma-4-31B-it.BF16-mmproj.gguf filter=lfs diff=lfs merge=lfs -text
40
+ VA-Gemma4-31B-Q4_K_M.gguf filter=lfs diff=lfs merge=lfs -text
41
+ VA-Gemma4-31B-Q8_0.gguf filter=lfs diff=lfs merge=lfs -text
42
+ VA-Gemma4-31B-UD-Q4_K_M.gguf filter=lfs diff=lfs merge=lfs -text
LICENSE ADDED
@@ -0,0 +1,172 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ Apache License
2
+ Version 2.0, January 2004
3
+ https://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
+
39
+ "Derivative Works" shall mean any work, whether in Source or Object
40
+ form, that is based on or derived from the Work and for which the
41
+ editorial revisions, annotations, elaborations, or other modifications
42
+ represent, as a whole, an original work of authorship. For the purposes
43
+ of this License, Derivative Works shall not include works that remain
44
+ separable from, or merely link (or bind by name) to the interfaces of,
45
+ the Work and Derivative Works thereof.
46
+
47
+ "Contribution" shall mean any work of authorship, including
48
+ the original version of the Work and any modifications or additions
49
+ to that Work or Derivative Works thereof, that is intentionally
50
+ submitted to Licensor for inclusion in the Work by the copyright owner
51
+ or by an individual or Legal Entity authorized to submit on behalf of
52
+ the copyright owner. For the purposes of this definition, "submitted"
53
+ means any form of electronic, verbal, or written communication sent
54
+ to the Licensor or its representatives, including but not limited to
55
+ communication on electronic mailing lists, source code control systems,
56
+ and issue tracking systems that are managed by, or on behalf of, the
57
+ Licensor for the purpose of discussing and improving the Work, but
58
+ excluding communication that is conspicuously marked or otherwise
59
+ designated in writing by the copyright owner as "Not a Contribution."
60
+
61
+ "Contributor" shall mean Licensor and any individual or Legal Entity
62
+ on behalf of whom a Contribution has been received by Licensor and
63
+ subsequently incorporated within the Work.
64
+
65
+ 2. Grant of Copyright License. Subject to the terms and conditions of
66
+ this License, each Contributor hereby grants to You a perpetual,
67
+ worldwide, non-exclusive, no-charge, royalty-free, irrevocable
68
+ copyright license to reproduce, prepare Derivative Works of,
69
+ publicly display, publicly perform, sublicense, and distribute the
70
+ Work and such Derivative Works in Source or Object form.
71
+
72
+ 3. Grant of Patent License. Subject to the terms and conditions of
73
+ this License, each Contributor hereby grants to You a perpetual,
74
+ worldwide, non-exclusive, no-charge, royalty-free, irrevocable
75
+ patent license to make, have made, use, offer to sell, sell, import,
76
+ and otherwise transfer the Work, where such license applies only to
77
+ those patent claims licensable by such Contributor that are necessarily
78
+ infringed by their Contribution alone or by combination of their
79
+ Contribution with the Work to which such Contribution was submitted.
80
+ If You institute patent litigation against any entity alleging that
81
+ the Work or a Contribution incorporated within the Work constitutes
82
+ direct or contributory patent infringement, then any patent licenses
83
+ granted to You under this License for that Work shall terminate as of
84
+ the date such litigation is filed.
85
+
86
+ 4. Redistribution. You may reproduce and distribute copies of the
87
+ Work or Derivative Works thereof in any medium, with or without
88
+ modifications, and in Source or Object form, provided that You
89
+ meet the following conditions:
90
+
91
+ (a) You must give any other recipients of the Work or
92
+ Derivative Works a copy of this License; and
93
+
94
+ (b) You must cause any modified files to carry prominent notices
95
+ stating that You changed the files; and
96
+
97
+ (c) You must retain, in the Source form of any Derivative Works
98
+ that You distribute, all copyright, patent, trademark, and
99
+ attribution notices from the Source form of the Work,
100
+ excluding those notices that do not pertain to any part of
101
+ the Derivative Works; and
102
+
103
+ (d) If the Work includes a "NOTICE" text file as part of its
104
+ distribution, then any Derivative Works that You distribute must
105
+ include a readable copy of the attribution notices contained
106
+ within such NOTICE file, excluding those notices that do not
107
+ pertain to any part of the Derivative Works, in at least one
108
+ of the following places: within a NOTICE text file distributed
109
+ as part of the Derivative Works; within the Source form or
110
+ documentation, if provided along with the Derivative Works; or,
111
+ within a display generated by the Derivative Works, if and
112
+ wherever such third-party notices normally appear. The contents
113
+ of the NOTICE file are for informational purposes only and
114
+ do not modify the License. You may add Your own attribution
115
+ notices within Derivative Works that You distribute, alongside
116
+ or as an addendum to the NOTICE text from the Work, provided
117
+ that such additional attribution notices cannot be construed
118
+ as modifying the License.
119
+
120
+ You may add Your own copyright statement to Your modifications and
121
+ may provide additional or different license terms and conditions
122
+ for use, reproduction, or distribution of Your modifications, or
123
+ for any such Derivative Works as a whole, provided Your use,
124
+ reproduction, and distribution of the Work otherwise complies with
125
+ the conditions stated in this License.
126
+
127
+ 5. Submission of Contributions. Unless You explicitly state otherwise,
128
+ any Contribution intentionally submitted for inclusion in the Work
129
+ by You to the Licensor shall be under the terms and conditions of
130
+ this License, without any additional terms or conditions.
131
+ Notwithstanding the above, nothing herein shall supersede or modify
132
+ the terms of any separate license agreement you may have executed
133
+ with Licensor regarding such Contributions.
134
+
135
+ 6. Trademarks. This License does not grant permission to use the trade
136
+ names, trademarks, service marks, or product names of the Licensor,
137
+ except as required for reasonable and customary use in describing the
138
+ origin of the Work and reproducing the content of the NOTICE file.
139
+
140
+ 7. Disclaimer of Warranty. Unless required by applicable law or
141
+ agreed to in writing, Licensor provides the Work on an "AS IS" BASIS,
142
+ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
143
+ implied, including, without limitation, any warranties or conditions
144
+ of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
145
+ PARTICULAR PURPOSE. You are solely responsible for determining the
146
+ appropriateness of using or redistributing the Work and assume any
147
+ risks associated with Your exercise of permissions under this License.
148
+
149
+ 8. Limitation of Liability. In no event and under no legal theory,
150
+ whether in tort (including negligence), contract, or otherwise,
151
+ unless required by applicable law (such as deliberate and grossly
152
+ negligent acts) or agreed to in writing, shall any Contributor be
153
+ liable to You for damages, including any direct, indirect, special,
154
+ incidental, or consequential damages of any character arising as a
155
+ result of this License or out of the use or inability to use the
156
+ Work (including but not limited to damages for loss of goodwill,
157
+ work stoppage, computer failure or malfunction, or any and all
158
+ other commercial damages or losses), even if such Contributor
159
+ has been advised of the possibility of such damages.
160
+
161
+ 9. Accepting Warranty or Additional Liability. While redistributing
162
+ the Work or Derivative Works thereof, You may choose to offer,
163
+ and charge a fee for, acceptance of support, warranty, indemnity,
164
+ or other liability obligations and/or rights consistent with this
165
+ License. However, in accepting such obligations, You may act only
166
+ on Your own behalf and on Your sole responsibility, not on behalf
167
+ of any other Contributor, and only if You agree to indemnify,
168
+ defend, and hold each Contributor harmless for any liability
169
+ incurred by, or claims asserted against, such Contributor by reason
170
+ of your accepting any such warranty or additional liability.
171
+
172
+ END OF TERMS AND CONDITIONS
NOTICE.md ADDED
@@ -0,0 +1,20 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # Notice
2
+
3
+ This repository contains GGUF model files produced from a merged VariantAssist LoRA adapter and Gemma 4 31B IT base model.
4
+
5
+ Upstream base model:
6
+
7
+ ```text
8
+ google/gemma-4-31B-it
9
+ ```
10
+
11
+ VariantAssist LoRA adapter:
12
+
13
+ ```text
14
+ LocusForge/VariantAssist-Gemma4-31B-LoRA
15
+ ```
16
+
17
+ The GGUF files are derivative quantized artifacts for local inference with llama.cpp-compatible runtimes.
18
+ Use of these files remains subject to the applicable upstream model license and terms.
19
+
20
+ VariantAssist is intended to support structured expert review and is not a diagnostic device.
README.md ADDED
@@ -0,0 +1,208 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ base_model: LocusForge/VariantAssist-Gemma4-31B-LoRA
4
+ base_model_relation: quantized
5
+ library_name: gguf
6
+ pipeline_tag: text-generation
7
+ tags:
8
+ - gguf
9
+ - llama.cpp
10
+ - llama-server
11
+ - unsloth
12
+ - unsloth-quantized
13
+ - llama-cpp-quantized
14
+ - gemma
15
+ - gemma-4
16
+ - clinical-genetics
17
+ - variant-interpretation
18
+ - acmg-amp
19
+ - local-first
20
+ - json
21
+ - variantassist
22
+ language:
23
+ - en
24
+ ---
25
+
26
+ <p align="center">
27
+ <img src="assets/variantassist-logo.svg" alt="VariantAssist" width="720">
28
+ </p>
29
+
30
+ <p align="center">
31
+ <a href="https://variantassist.com/">variantassist.com</a>
32
+ ·
33
+ <a href="https://github.com/LocusForge/VariantAssist">GitHub</a>
34
+ ·
35
+ <a href="https://huggingface.co/LocusForge/VariantAssist-Gemma4-31B-GGUF/blob/main/LICENSE">License</a>
36
+ </p>
37
+
38
+ <div style="background-color:#fff6df;border-left:5px solid #d4900a;padding:14px 16px;margin:18px 0;border-radius:6px;">
39
+ <strong>Compatibility note:</strong> these VariantAssist-tuned GGUF models are currently intended only for <strong>Level-1 Annotation</strong>.
40
+ For other VariantAssist workflow stages, use the original Q8 model rather than these tuned quantizations.
41
+ </div>
42
+
43
+ # VariantAssist Gemma 4 31B GGUF
44
+
45
+ VariantAssist Gemma 4 31B GGUF is the local-inference release of the VariantAssist Gemma 4 31B LoRA model.
46
+ The files in this repository are produced by merging the [VariantAssist LoRA adapter](https://huggingface.co/LocusForge/VariantAssist-Gemma4-31B-LoRA) with Gemma 4 31B IT and converting/quantizing the merged model for llama.cpp-compatible runtimes.
47
+
48
+ VariantAssist is designed to support structured clinical genetic variant review.
49
+ It is not a diagnostic device and must not replace a clinician, medical geneticist, laboratory director, or ACMG/AMP-trained reviewer.
50
+
51
+ ## Evaluation Protocol
52
+
53
+ All model scores below are evaluated after the VariantAssist 3-to-5 consensus procedure.
54
+ For each variant, the model is first run three times.
55
+ If all three runs return the same pathogenicity level, that level is accepted.
56
+ If any run differs, two additional runs are performed; a result is accepted only if one pathogenicity level appears at least three times across the five runs.
57
+ If no level reaches that threshold, the result is marked as no consensus and may be rerun.
58
+
59
+ No dissensus/no-consensus cases occurred in this benchmark.
60
+ In practical use, no-consensus cases have been observed at roughly 1 in 5000 variants.
61
+
62
+ ## Available GGUF Files
63
+
64
+ | File | Size | Match | Quant | Role |
65
+ |---|---:|---:|---:|---|
66
+ | `VA-Gemma4-31B-UD-Q8_0.gguf` | 31 GB | 86 | UDQ | Best current benchmark result |
67
+ | `VA-Gemma4-31B-Q4_K_M.gguf` | 18 GB | 85 | LQ | Practical default |
68
+ | `VA-Gemma4-31B-Q8_0.gguf` | 31 GB | 83 | LQ | Classic Q8 variant |
69
+ | `VA-Gemma4-31B-UD-Q4_K_M.gguf` | 18 GB | 82 | UDQ | Smaller UDQ variant |
70
+ | `VA-Gemma4-31B-F16.gguf` | 58 GB | 81 | F16 | Reference GGUF |
71
+ | `VA-Gemma4-31B-BF16-00002-of-00002.gguf` | 11 GB | - | BF16 | BF16 export shard |
72
+ | `VA-Gemma4-31B-BF16-mmproj.gguf` | 1.2 GB | - | MMProj | Not needed for text-only runs |
73
+
74
+ <p style="font-size:0.9em;color:#4f5b66;">
75
+ <strong>UDQ</strong> = Unsloth dynamic quantization.
76
+ <strong>LQ</strong> = classic llama.cpp quantization.
77
+ The Unsloth quantized variants were selected/validated on examples with the correct VariantAssist Level-1 input/output structure.
78
+ </p>
79
+
80
+ ## Benchmark Results
81
+
82
+ The ATP7B benchmark contains 100 Wilson disease variants with consensus labels from five independent expert annotations.
83
+ The primary ground truth is strict majority consensus.
84
+
85
+ <p align="center">
86
+ <img src="assets/accuracy_vs_reasoning_tokens.png" alt="ATP7B benchmark accuracy versus reasoning-token budget">
87
+ </p>
88
+
89
+ Reasoning budget is usually an important quality driver for classic quantized models.
90
+ In this benchmark, the VariantAssist-tuned quantized runs improve accuracy while also reducing the reasoning-token budget compared with the original quantized baseline.
91
+
92
+ Current highlighted result:
93
+
94
+ - `VariantAssist UD-Q8`: 86/100 exact matches on the ATP7B benchmark.
95
+ - No strong errors in the selected released-model comparison.
96
+ - Expert-consensus reference: 15 average expert disagreements, equivalent to 85/100 agreement.
97
+
98
+ <p align="center">
99
+ <img src="assets/confusion_matrix_variantassist_ud_q8.png" alt="VariantAssist UD-Q8 ATP7B confusion matrix" width="720">
100
+ </p>
101
+
102
+ ## Prompts, Schema, And Reproducibility
103
+
104
+ Use the public prompt archive for reproducible evaluation:
105
+
106
+ - [ATP7B prompt archive](https://github.com/LocusForge/VariantAssist-supplement/tree/main/benchmark/atp7b-prompts)
107
+ - [System instruction](https://github.com/LocusForge/VariantAssist-supplement/blob/main/benchmark/atp7b-prompts/system_instruction.txt)
108
+ - [Response schema](https://github.com/LocusForge/VariantAssist-supplement/blob/main/benchmark/atp7b-prompts/response_schema.json)
109
+ - [Annotation rules](https://github.com/LocusForge/VariantAssist-supplement/blob/main/benchmark/atp7b-prompts/annotation_rules.json)
110
+ - [ATP7B benchmark ground truth](https://github.com/LocusForge/VariantAssist-supplement/tree/main/benchmark)
111
+
112
+ That archive contains the system prompt, schema, annotation rules, and per-variant prompts used for benchmark-style evaluation.
113
+
114
+ ## Runtime
115
+
116
+ Recommended runtime is `llama-server` from a recent llama.cpp build with Gemma 4 reasoning support.
117
+
118
+ Recommended server command:
119
+
120
+ ```bash
121
+ llama-server \
122
+ -m /path/to/VA-Gemma4-31B-Q4_K_M.gguf \
123
+ --no-mmproj \
124
+ --jinja \
125
+ -ngl auto \
126
+ -c 32768 \
127
+ -fa on \
128
+ --swa-full \
129
+ -np 1 \
130
+ --cache-prompt \
131
+ --cache-reuse 256 \
132
+ --slot-prompt-similarity 0.10 \
133
+ --ctx-checkpoints 1 \
134
+ --checkpoint-every-n-tokens 4096 \
135
+ --cache-ram 2048 \
136
+ --kv-unified \
137
+ --cache-type-k f16 \
138
+ --cache-type-v f16 \
139
+ -b 2048 \
140
+ -ub 512 \
141
+ --no-cont-batching \
142
+ --perf \
143
+ --metrics \
144
+ --host 127.0.0.1 \
145
+ --port 8091 \
146
+ --reasoning on \
147
+ --reasoning-budget 8192 \
148
+ -t 24 \
149
+ -tb 24
150
+ ```
151
+
152
+ Small-machine optimization:
153
+
154
+ ```bash
155
+ -c 8192 --reasoning-budget 4096
156
+ ```
157
+
158
+ What to change:
159
+
160
+ - `-m`: select the GGUF file.
161
+ - `--host` / `--port`: set your serving endpoint.
162
+ - `-t` / `-tb`: match your CPU thread budget.
163
+ - `-c` and `--reasoning-budget`: reduce on smaller machines if needed.
164
+
165
+ What to keep for VariantAssist Level-1 runs:
166
+
167
+ - `--reasoning on`: benchmarked runs use reasoning mode.
168
+ - `--jinja`: uses the Gemma chat template.
169
+ - `--no-mmproj`: this release is text-only.
170
+ - `--cache-type-k f16 --cache-type-v f16`: keeps KV cache quality stable.
171
+ - `--no-cont-batching`: keeps single-review behavior predictable.
172
+
173
+ Reasoning should remain enabled for VariantAssist-style review.
174
+ In our workflow, no-reasoning runs could generate shorter single responses, but were less reliable in the completed 3-to-5 consensus process and could require reruns.
175
+
176
+ ## Intended Use
177
+
178
+ Use this release for:
179
+
180
+ - local-first VariantAssist review workflows;
181
+ - structured evidence synthesis for expert review;
182
+ - JSON-oriented draft outputs;
183
+ - reproducible local benchmarking with the public ATP7B prompt archive.
184
+
185
+ ## Out Of Scope
186
+
187
+ Do not use this model for:
188
+
189
+ - autonomous diagnosis;
190
+ - direct patient-facing medical advice;
191
+ - final ACMG/AMP classification without expert review;
192
+ - clinical interpretation outside the supplied evidence context;
193
+ - high-stakes clinical workflows without local validation.
194
+
195
+ ## Training Data
196
+
197
+ The full fine-tuning corpus is not distributed with this release because it may include clinical-context and literature-derived materials requiring separate privacy and licensing review.
198
+ Public benchmark data, prompt templates, response schema, and de-identified examples are provided separately to support reproducible evaluation.
199
+
200
+ ## Links
201
+
202
+ - Website: https://variantassist.com/
203
+ - LoRA adapter: https://huggingface.co/LocusForge/VariantAssist-Gemma4-31B-LoRA
204
+ - Main project: https://github.com/LocusForge/VariantAssist
205
+ - Benchmark and prompts: https://github.com/LocusForge/VariantAssist-supplement/tree/main/benchmark
206
+ - Upstream base model: https://huggingface.co/google/gemma-4-31B-it
207
+ - License: [Apache License 2.0](https://huggingface.co/LocusForge/VariantAssist-Gemma4-31B-GGUF/blob/main/LICENSE)
208
+ - Notice: [NOTICE.md](https://huggingface.co/LocusForge/VariantAssist-Gemma4-31B-GGUF/blob/main/NOTICE.md)
VA-Gemma4-31B-Q4_K_M.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d2b50619b28d1eab6bca8456243f85ca42cd3fe45200b586982d7eb89b186e8c
3
+ size 18687061696
VA-Gemma4-31B-Q8_0.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c3310fd4a14fe7e23cf6126e4d3eac516a79e3fc4d7775a42525ab3d9739b03f
3
+ size 32635674304
VA-Gemma4-31B-UD-Q4_K_M.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5ac032f63e3d6a3dbf13f98e565d3f32dec7b52de5ad58d740e1f4a723325d6d
3
+ size 18687061984
VA-gemma-4-31B-it.BF16-mmproj.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c6ab2f76a7543fd8495b913be015380db583ecf470abcdf093bdaba261fd2d91
3
+ size 1200726016
assets/accuracy_vs_reasoning_tokens.png ADDED

Git LFS Details

  • SHA256: f33097903b9e1ec52d93eedb5577e3dcb6b6f4812969b824478cf0c0c94f9650
  • Pointer size: 131 Bytes
  • Size of remote file: 216 kB
assets/confusion_matrix_variantassist_ud_q8.png ADDED

Git LFS Details

  • SHA256: e459edb9b9a8e81e5d38863febee31f3adb1e0f57db07231b6f1e1f5f877e307
  • Pointer size: 131 Bytes
  • Size of remote file: 103 kB
assets/variantassist-logo.svg ADDED