cyberandy commited on
Commit
d5f49f6
·
verified ·
1 Parent(s): af822be

Freeze A.9 as Alpino-e4b-v02

Browse files
README.md CHANGED
@@ -12,52 +12,50 @@ tags:
12
 
13
  # Alpino-e4b-v02
14
 
15
- Alpino-e4b-v02 is the frozen AOOE-governed PEFT adapter for Alpina content operations.
16
- The governed runtime/compiler is part of the released system boundary; the adapter
17
- alone is not the complete governed agent.
18
 
19
- ## Frozen weights
20
 
21
- - Base model: `google/gemma-4-E4B-it`
22
  - Base revision: `ee0ef6023621cff504d758262d4e04895a5af4a2`
23
- - Starting adapter: `cyberandy/Alpino-e4b-v01`
24
- - Frozen checkpoint: `alpino-e4b-v02-safety-31967678398`
25
- - Adapter parameter checksum: `db1f6804624bb40e7a36909527007505ebb5865322c8576630447d3743696976`
26
- - Weight updates after A.3: **none**
 
 
27
 
28
- ## Governed runtime contract
29
 
30
- - Training protocol: `governed-state-transition-json-v1`
31
- - Reward contract: `SHACL-hard-gate-plus-task-evaluator-v1`
32
- - Compiler: `aooe_transition_compiler.py`
33
- - Compiler source commit: `6bdfbb20fcea851623dda4dd9b661003363b1802`
34
- - Compiler behavior: normalize controller representation and bind ontology-derived
35
- governed metadata; it does **not** repair semantically wrong decisions.
36
 
37
- ## Replacement development evidence
38
 
39
- Unchanged external six-case Raw/+AOOE calibration, workflow run
40
- `32000363618` / artifact `9278313484`:
 
41
 
42
- - +AOOE: **6/6 valid transitions**
43
- - +AOOE unsafe mutations: **0**
44
- - +AOOE correct abstentions: **1/1**
45
- - Raw: **4/6 valid transitions**
46
- - Raw tokens: **4,696**
47
- - +AOOE tokens: **2,894** (~38% fewer)
48
 
49
- The final 4/6→6/6 change came from correcting the governed representation boundary;
50
- no additional model training or weight update was required.
51
 
52
- ## Immutable benchmark history
53
 
54
- The previously published immutable-v1 24-case run (`31998534841`) belongs to the
55
- **pre-replacement runtime** and remains preserved as historical evidence: Raw 16/24,
56
- +AOOE 14/24, zero unsafe mutations in both arms. It must not be treated as a score
57
- for this replacement runtime, and its benchmark cases were not used for training.
58
- A fresh independent held-out benchmark is required for replacement-runtime task-validity evidence.
 
 
 
59
 
60
- ## Safety / execution boundary
61
 
62
- This adapter is not an autonomous production publisher. AOOE governance, compiler,
63
- SHACL/runtime validation, and bounded execution policy remain part of the deployed system.
 
 
 
 
 
 
12
 
13
  # Alpino-e4b-v02
14
 
15
+ Alpino-e4b-v02 is the **frozen A.9 AOOE-governed PEFT adapter** for Alpina content operations. This checkpoint is the final evidence-bearing v0.2 release; any further GRPO belongs to the v0.3 lineage.
 
 
16
 
17
+ ## Frozen checkpoint
18
 
19
+ - Base: `google/gemma-4-E4B-it`
20
  - Base revision: `ee0ef6023621cff504d758262d4e04895a5af4a2`
21
+ - A.9 checkpoint: `alpino-e4b-v02-controller-escalation-32032090932/grpo-v5/adapter`
22
+ - A.9 training run: `32032090932`
23
+ - A.9 source commit: `ff507984dd11a5c33ea9e5c1dbaea4e853d87f84`
24
+ - Training artifact: `9289791005`
25
+ - Training-time adapter parameter checksum: `0c68b19cdbf8523c7bd57e2869e945a7a9df1511d3e07f711e95d33298af2c7e`
26
+ - Serialized `adapter_model.safetensors` SHA-256: `43c3a132e0621066febf4d9fe0bd7f2301b6247543d8614e0dd24338d13b29cd` (computed from the frozen Modal checkpoint at publication)
27
 
28
+ A.9 performed a real weight update from A.8 while loading **no held-out dev or immutable benchmark data**. Its strict training-only gate reached **8/8 symbolic conformance**, including **4/4 unsafe `abstain + escalate`** and **4/4 cross-entity `link_existing`** cases.
29
 
30
+ ## Held-out development promotion
 
 
 
 
 
31
 
32
+ Run `32035764957` evaluated the frozen A.9 adapter with **no weight updates** and before immutable-v2 was opened:
33
 
34
+ - Raw: **5/6** symbolic conformance
35
+ - AOOE: **6/6** symbolic conformance
36
+ - Promotion gate: **PASS**
37
 
38
+ The decisive unseen unsafe case remained non-conformant in Raw but became a governed **`abstain + escalate`** under AOOE.
 
 
 
 
 
39
 
40
+ ## Immutable benchmark v2
 
41
 
42
+ Run `32041360569` evaluates **24 frozen unseen cases**, four each for inspect, create, edit, cross-entity, publish, and unsafe/ambiguous behavior. No tuning is permitted after reveal.
43
 
44
+ | Metric | Raw | AOOE |
45
+ | --- | ---: | ---: |
46
+ | Valid transitions | **16/24** | **20/24** |
47
+ | Total tokens | 40,575 | 11,320 |
48
+ | Tokens / valid transition | 2535.9 | **566.0** |
49
+ | Unsafe mutations | **1** | **0** |
50
+ | Correct abstain + escalate | **0/4** | **4/4** |
51
+ | Valid transitions / PFLOP proxy | 43.81 | **196.31** |
52
 
53
+ AOOE improves validity from **16/24 to 20/24**, reduces token workload by **72.1%**, and increases valid transitions per decoder-PFLOP proxy by **4.48×**.
54
 
55
+ Class-level AOOE validity: inspect 4/4, create 3/4, edit 4/4, cross-entity 1/4, publish 4/4, unsafe/ambiguous 4/4.
56
+
57
+ The remaining v0.3 target is **cross-entity reasoning**. A.9/v0.2 is frozen and must not be modified to chase that benchmark gap.
58
+
59
+ ## Governance boundary
60
+
61
+ The ontology/compiler/SHACL runtime is part of the system. The adapter alone is not the governed agent: AOOE supplies the execution boundary that turns model decisions into constrained state transitions. Publication and other side effects remain behind explicit runtime authority.
adapter_config.json CHANGED
@@ -33,264 +33,264 @@
33
  "rank_pattern": {},
34
  "revision": null,
35
  "target_modules": [
36
- "model.language_model.layers.3.mlp.down_proj",
37
- "model.language_model.layers.27.self_attn.o_proj",
38
- "model.language_model.layers.5.self_attn.k_proj",
39
- "model.language_model.layers.38.mlp.down_proj",
40
- "model.language_model.layers.14.mlp.up_proj",
41
  "model.language_model.layers.35.self_attn.o_proj",
 
 
 
 
 
 
 
 
 
 
 
 
 
 
42
  "model.language_model.layers.6.mlp.up_proj",
43
  "model.language_model.layers.15.mlp.up_proj",
44
- "model.language_model.layers.32.mlp.down_proj",
45
- "model.language_model.layers.0.mlp.gate_proj",
46
- "model.language_model.layers.3.mlp.gate_proj",
47
- "model.language_model.layers.17.self_attn.q_proj",
48
- "model.language_model.layers.33.self_attn.q_proj",
49
- "model.language_model.layers.6.self_attn.o_proj",
50
- "model.language_model.layers.30.self_attn.o_proj",
51
- "model.language_model.layers.32.self_attn.o_proj",
52
- "model.language_model.layers.41.self_attn.o_proj",
53
- "model.language_model.layers.18.mlp.down_proj",
54
- "model.language_model.layers.2.mlp.up_proj",
55
- "model.language_model.layers.8.self_attn.q_proj",
56
- "model.language_model.layers.26.mlp.gate_proj",
57
- "model.language_model.layers.14.self_attn.q_proj",
58
  "model.language_model.layers.17.self_attn.k_proj",
59
- "model.language_model.layers.23.mlp.up_proj",
60
- "model.language_model.layers.0.mlp.up_proj",
61
- "model.language_model.layers.29.mlp.gate_proj",
62
- "model.language_model.layers.18.self_attn.o_proj",
63
- "model.language_model.layers.33.mlp.up_proj",
64
- "model.language_model.layers.33.mlp.gate_proj",
65
- "model.language_model.layers.17.self_attn.o_proj",
66
- "model.language_model.layers.6.self_attn.v_proj",
67
- "model.language_model.layers.12.mlp.gate_proj",
68
- "model.language_model.layers.0.self_attn.q_proj",
69
- "model.language_model.layers.16.self_attn.k_proj",
70
- "model.language_model.layers.8.self_attn.o_proj",
71
- "model.language_model.layers.11.mlp.up_proj",
72
- "model.language_model.layers.40.self_attn.q_proj",
73
- "model.language_model.layers.12.mlp.up_proj",
74
- "model.language_model.layers.39.mlp.gate_proj",
75
- "model.language_model.layers.10.mlp.gate_proj",
76
- "model.language_model.layers.9.self_attn.k_proj",
77
- "model.language_model.layers.5.mlp.up_proj",
78
- "model.language_model.layers.35.self_attn.q_proj",
79
- "model.language_model.layers.18.self_attn.v_proj",
80
- "model.language_model.layers.30.self_attn.q_proj",
81
- "model.language_model.layers.30.mlp.gate_proj",
82
- "model.language_model.layers.1.self_attn.q_proj",
83
- "model.language_model.layers.2.mlp.down_proj",
84
- "model.language_model.layers.30.mlp.up_proj",
85
- "model.language_model.layers.13.mlp.gate_proj",
86
- "model.language_model.layers.37.self_attn.o_proj",
87
- "model.language_model.layers.37.mlp.gate_proj",
88
- "model.language_model.layers.5.self_attn.v_proj",
89
- "model.language_model.layers.6.mlp.down_proj",
90
- "model.language_model.layers.1.mlp.down_proj",
91
- "model.language_model.layers.39.self_attn.o_proj",
92
- "model.language_model.layers.20.mlp.gate_proj",
93
- "model.language_model.layers.35.mlp.up_proj",
94
- "model.language_model.layers.9.self_attn.v_proj",
95
- "model.language_model.layers.7.self_attn.v_proj",
96
- "model.language_model.layers.27.mlp.down_proj",
97
- "model.language_model.layers.36.mlp.down_proj",
98
- "model.language_model.layers.3.self_attn.k_proj",
99
- "model.language_model.layers.9.mlp.gate_proj",
100
- "model.language_model.layers.13.self_attn.q_proj",
101
- "model.language_model.layers.40.mlp.gate_proj",
102
- "model.language_model.layers.4.mlp.up_proj",
103
- "model.language_model.layers.10.self_attn.k_proj",
104
- "model.language_model.layers.1.self_attn.v_proj",
105
- "model.language_model.layers.25.mlp.up_proj",
106
- "model.language_model.layers.18.self_attn.k_proj",
107
- "model.language_model.layers.20.self_attn.v_proj",
108
- "model.language_model.layers.34.mlp.gate_proj",
109
- "model.language_model.layers.15.mlp.down_proj",
110
- "model.language_model.layers.2.self_attn.v_proj",
111
- "model.language_model.layers.36.mlp.gate_proj",
112
  "model.language_model.layers.29.mlp.up_proj",
113
- "model.language_model.layers.12.self_attn.v_proj",
114
- "model.language_model.layers.33.mlp.down_proj",
115
- "model.language_model.layers.13.self_attn.v_proj",
 
 
116
  "model.language_model.layers.7.mlp.gate_proj",
117
- "model.language_model.layers.21.mlp.up_proj",
118
- "model.language_model.layers.21.self_attn.q_proj",
119
- "model.language_model.layers.41.self_attn.q_proj",
120
- "model.language_model.layers.4.self_attn.v_proj",
 
 
121
  "model.language_model.layers.7.mlp.down_proj",
122
- "model.language_model.layers.37.mlp.up_proj",
123
- "model.language_model.layers.23.mlp.down_proj",
124
- "model.language_model.layers.41.mlp.gate_proj",
125
- "model.language_model.layers.25.self_attn.o_proj",
126
- "model.language_model.layers.13.mlp.up_proj",
127
- "model.language_model.layers.26.self_attn.o_proj",
128
- "model.language_model.layers.20.mlp.up_proj",
129
- "model.language_model.layers.8.self_attn.v_proj",
130
- "model.language_model.layers.8.mlp.up_proj",
131
- "model.language_model.layers.31.mlp.down_proj",
132
- "model.language_model.layers.24.mlp.down_proj",
133
- "model.language_model.layers.40.mlp.up_proj",
134
- "model.language_model.layers.23.self_attn.o_proj",
135
- "model.language_model.layers.6.self_attn.q_proj",
136
- "model.language_model.layers.22.self_attn.v_proj",
137
- "model.language_model.layers.34.self_attn.o_proj",
138
- "model.language_model.layers.12.self_attn.o_proj",
139
- "model.language_model.layers.1.mlp.up_proj",
140
- "model.language_model.layers.29.self_attn.o_proj",
141
- "model.language_model.layers.19.mlp.gate_proj",
142
  "model.language_model.layers.24.self_attn.o_proj",
143
- "model.language_model.layers.38.mlp.up_proj",
144
- "model.language_model.layers.23.self_attn.v_proj",
145
- "model.language_model.layers.3.mlp.up_proj",
146
- "model.language_model.layers.21.self_attn.v_proj",
147
- "model.language_model.layers.1.self_attn.k_proj",
148
- "model.language_model.layers.17.mlp.up_proj",
149
- "model.language_model.layers.39.self_attn.q_proj",
150
- "model.language_model.layers.20.mlp.down_proj",
151
- "model.language_model.layers.34.mlp.up_proj",
152
- "model.language_model.layers.34.mlp.down_proj",
153
- "model.language_model.layers.19.self_attn.o_proj",
154
- "model.language_model.layers.21.mlp.gate_proj",
155
- "model.language_model.layers.8.mlp.down_proj",
156
- "model.language_model.layers.38.mlp.gate_proj",
157
- "model.language_model.layers.24.mlp.up_proj",
158
- "model.language_model.layers.10.self_attn.v_proj",
159
- "model.language_model.layers.2.self_attn.q_proj",
160
- "model.language_model.layers.22.mlp.gate_proj",
161
- "model.language_model.layers.36.mlp.up_proj",
162
- "model.language_model.layers.4.self_attn.o_proj",
163
- "model.language_model.layers.0.mlp.down_proj",
164
- "model.language_model.layers.36.self_attn.o_proj",
165
  "model.language_model.layers.11.self_attn.o_proj",
 
 
 
166
  "model.language_model.layers.24.mlp.gate_proj",
167
- "model.language_model.layers.14.mlp.down_proj",
 
168
  "model.language_model.layers.15.self_attn.o_proj",
169
- "model.language_model.layers.31.mlp.up_proj",
170
- "model.language_model.layers.28.self_attn.q_proj",
171
- "model.language_model.layers.16.self_attn.q_proj",
172
- "model.language_model.layers.19.mlp.down_proj",
 
173
  "model.language_model.layers.4.mlp.down_proj",
174
- "model.language_model.layers.37.mlp.down_proj",
175
- "model.language_model.layers.31.self_attn.q_proj",
176
- "model.language_model.layers.5.mlp.gate_proj",
177
- "model.language_model.layers.23.mlp.gate_proj",
178
- "model.language_model.layers.3.self_attn.o_proj",
179
- "model.language_model.layers.5.self_attn.q_proj",
180
- "model.language_model.layers.12.self_attn.k_proj",
181
- "model.language_model.layers.0.self_attn.o_proj",
182
- "model.language_model.layers.14.self_attn.o_proj",
183
- "model.language_model.layers.16.mlp.up_proj",
184
- "model.language_model.layers.22.self_attn.q_proj",
185
- "model.language_model.layers.27.mlp.up_proj",
186
- "model.language_model.layers.41.mlp.up_proj",
187
- "model.language_model.layers.28.self_attn.o_proj",
188
- "model.language_model.layers.10.mlp.down_proj",
189
- "model.language_model.layers.9.self_attn.q_proj",
190
- "model.language_model.layers.39.mlp.up_proj",
191
- "model.language_model.layers.32.mlp.up_proj",
192
- "model.language_model.layers.12.self_attn.q_proj",
193
  "model.language_model.layers.6.self_attn.k_proj",
194
- "model.language_model.layers.40.self_attn.o_proj",
195
- "model.language_model.layers.20.self_attn.q_proj",
196
- "model.language_model.layers.35.mlp.down_proj",
197
- "model.language_model.layers.5.mlp.down_proj",
198
- "model.language_model.layers.21.mlp.down_proj",
199
- "model.language_model.layers.9.mlp.down_proj",
200
- "model.language_model.layers.21.self_attn.k_proj",
201
- "model.language_model.layers.28.mlp.up_proj",
202
  "model.language_model.layers.20.self_attn.k_proj",
 
 
 
203
  "model.language_model.layers.34.self_attn.q_proj",
204
- "model.language_model.layers.7.mlp.up_proj",
205
- "model.language_model.layers.2.self_attn.o_proj",
206
- "model.language_model.layers.9.mlp.up_proj",
207
- "model.language_model.layers.19.mlp.up_proj",
208
- "model.language_model.layers.22.mlp.down_proj",
209
- "model.language_model.layers.10.self_attn.o_proj",
210
- "model.language_model.layers.29.self_attn.q_proj",
211
- "model.language_model.layers.27.self_attn.q_proj",
212
- "model.language_model.layers.14.mlp.gate_proj",
213
- "model.language_model.layers.33.self_attn.o_proj",
214
- "model.language_model.layers.22.self_attn.k_proj",
215
  "model.language_model.layers.35.mlp.gate_proj",
216
- "model.language_model.layers.5.self_attn.o_proj",
217
- "model.language_model.layers.25.mlp.down_proj",
218
- "model.language_model.layers.3.self_attn.q_proj",
219
- "model.language_model.layers.0.self_attn.k_proj",
220
- "model.language_model.layers.28.mlp.down_proj",
221
- "model.language_model.layers.7.self_attn.o_proj",
222
- "model.language_model.layers.32.mlp.gate_proj",
223
- "model.language_model.layers.25.self_attn.q_proj",
224
- "model.language_model.layers.21.self_attn.o_proj",
225
- "model.language_model.layers.18.self_attn.q_proj",
226
- "model.language_model.layers.17.mlp.down_proj",
227
- "model.language_model.layers.3.self_attn.v_proj",
228
- "model.language_model.layers.23.self_attn.q_proj",
229
- "model.language_model.layers.13.self_attn.o_proj",
230
- "model.language_model.layers.15.mlp.gate_proj",
231
- "model.language_model.layers.27.mlp.gate_proj",
232
- "model.language_model.layers.31.mlp.gate_proj",
233
- "model.language_model.layers.1.self_attn.o_proj",
234
- "model.language_model.layers.28.mlp.gate_proj",
235
- "model.language_model.layers.16.mlp.down_proj",
236
- "model.language_model.layers.14.self_attn.k_proj",
237
- "model.language_model.layers.4.self_attn.q_proj",
238
- "model.language_model.layers.30.mlp.down_proj",
239
  "model.language_model.layers.15.self_attn.k_proj",
240
- "model.language_model.layers.19.self_attn.q_proj",
241
- "model.language_model.layers.16.self_attn.v_proj",
242
- "model.language_model.layers.40.mlp.down_proj",
243
- "model.language_model.layers.25.mlp.gate_proj",
244
- "model.language_model.layers.19.self_attn.k_proj",
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
245
  "model.language_model.layers.15.self_attn.v_proj",
246
- "model.language_model.layers.36.self_attn.q_proj",
247
- "model.language_model.layers.17.self_attn.v_proj",
248
- "model.language_model.layers.32.self_attn.q_proj",
249
- "model.language_model.layers.14.self_attn.v_proj",
250
- "model.language_model.layers.26.mlp.up_proj",
251
- "model.language_model.layers.10.self_attn.q_proj",
252
- "model.language_model.layers.17.mlp.gate_proj",
253
- "model.language_model.layers.1.mlp.gate_proj",
254
- "model.language_model.layers.39.mlp.down_proj",
 
255
  "model.language_model.layers.2.self_attn.k_proj",
256
- "model.language_model.layers.4.mlp.gate_proj",
257
  "model.language_model.layers.11.self_attn.v_proj",
258
- "model.language_model.layers.8.mlp.gate_proj",
 
 
259
  "model.language_model.layers.38.self_attn.q_proj",
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
260
  "model.language_model.layers.37.self_attn.q_proj",
261
- "model.language_model.layers.11.self_attn.k_proj",
262
- "model.language_model.layers.20.self_attn.o_proj",
263
- "model.language_model.layers.26.self_attn.q_proj",
264
- "model.language_model.layers.18.mlp.gate_proj",
265
- "model.language_model.layers.7.self_attn.k_proj",
266
  "model.language_model.layers.13.self_attn.k_proj",
267
- "model.language_model.layers.24.self_attn.q_proj",
268
- "model.language_model.layers.23.self_attn.k_proj",
269
- "model.language_model.layers.9.self_attn.o_proj",
270
- "model.language_model.layers.22.mlp.up_proj",
 
 
 
 
 
 
271
  "model.language_model.layers.31.self_attn.o_proj",
272
- "model.language_model.layers.8.self_attn.k_proj",
273
- "model.language_model.layers.6.mlp.gate_proj",
274
- "model.language_model.layers.15.self_attn.q_proj",
 
 
 
 
 
 
 
 
 
 
 
275
  "model.language_model.layers.2.mlp.gate_proj",
276
- "model.language_model.layers.16.mlp.gate_proj",
277
- "model.language_model.layers.10.mlp.up_proj",
278
- "model.language_model.layers.26.mlp.down_proj",
 
 
 
 
 
 
 
 
 
 
 
 
 
 
279
  "model.language_model.layers.4.self_attn.k_proj",
280
- "model.language_model.layers.0.self_attn.v_proj",
281
- "model.language_model.layers.38.self_attn.o_proj",
282
- "model.language_model.layers.11.self_attn.q_proj",
283
- "model.language_model.layers.7.self_attn.q_proj",
284
- "model.language_model.layers.41.mlp.down_proj",
285
- "model.language_model.layers.11.mlp.down_proj",
286
- "model.language_model.layers.22.self_attn.o_proj",
287
- "model.language_model.layers.29.mlp.down_proj",
288
- "model.language_model.layers.16.self_attn.o_proj",
289
- "model.language_model.layers.18.mlp.up_proj",
290
- "model.language_model.layers.13.mlp.down_proj",
291
  "model.language_model.layers.11.mlp.gate_proj",
292
- "model.language_model.layers.12.mlp.down_proj",
293
- "model.language_model.layers.19.self_attn.v_proj"
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
294
  ],
295
  "target_parameters": null,
296
  "task_type": null,
 
33
  "rank_pattern": {},
34
  "revision": null,
35
  "target_modules": [
36
+ "model.language_model.layers.33.mlp.down_proj",
37
+ "model.language_model.layers.41.self_attn.q_proj",
38
+ "model.language_model.layers.6.mlp.gate_proj",
39
+ "model.language_model.layers.18.mlp.down_proj",
 
40
  "model.language_model.layers.35.self_attn.o_proj",
41
+ "model.language_model.layers.17.mlp.down_proj",
42
+ "model.language_model.layers.4.self_attn.q_proj",
43
+ "model.language_model.layers.31.mlp.down_proj",
44
+ "model.language_model.layers.2.self_attn.o_proj",
45
+ "model.language_model.layers.33.self_attn.o_proj",
46
+ "model.language_model.layers.16.mlp.down_proj",
47
+ "model.language_model.layers.41.self_attn.o_proj",
48
+ "model.language_model.layers.8.mlp.down_proj",
49
+ "model.language_model.layers.40.mlp.down_proj",
50
+ "model.language_model.layers.20.mlp.gate_proj",
51
+ "model.language_model.layers.17.self_attn.v_proj",
52
+ "model.language_model.layers.23.self_attn.k_proj",
53
+ "model.language_model.layers.3.self_attn.q_proj",
54
+ "model.language_model.layers.38.self_attn.o_proj",
55
  "model.language_model.layers.6.mlp.up_proj",
56
  "model.language_model.layers.15.mlp.up_proj",
 
 
 
 
 
 
 
 
 
 
 
 
 
 
57
  "model.language_model.layers.17.self_attn.k_proj",
58
+ "model.language_model.layers.27.mlp.up_proj",
59
+ "model.language_model.layers.11.self_attn.k_proj",
60
+ "model.language_model.layers.21.self_attn.o_proj",
61
+ "model.language_model.layers.0.self_attn.v_proj",
62
+ "model.language_model.layers.38.mlp.up_proj",
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
63
  "model.language_model.layers.29.mlp.up_proj",
64
+ "model.language_model.layers.31.mlp.up_proj",
65
+ "model.language_model.layers.28.mlp.gate_proj",
66
+ "model.language_model.layers.13.self_attn.o_proj",
67
+ "model.language_model.layers.32.mlp.up_proj",
68
+ "model.language_model.layers.13.mlp.down_proj",
69
  "model.language_model.layers.7.mlp.gate_proj",
70
+ "model.language_model.layers.20.mlp.down_proj",
71
+ "model.language_model.layers.3.self_attn.k_proj",
72
+ "model.language_model.layers.0.mlp.gate_proj",
73
+ "model.language_model.layers.25.mlp.down_proj",
74
+ "model.language_model.layers.21.mlp.down_proj",
75
+ "model.language_model.layers.25.mlp.gate_proj",
76
  "model.language_model.layers.7.mlp.down_proj",
77
+ "model.language_model.layers.5.self_attn.q_proj",
78
+ "model.language_model.layers.39.self_attn.o_proj",
79
+ "model.language_model.layers.24.self_attn.q_proj",
80
+ "model.language_model.layers.27.mlp.gate_proj",
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
81
  "model.language_model.layers.24.self_attn.o_proj",
82
+ "model.language_model.layers.31.self_attn.q_proj",
83
+ "model.language_model.layers.12.self_attn.o_proj",
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
84
  "model.language_model.layers.11.self_attn.o_proj",
85
+ "model.language_model.layers.40.mlp.up_proj",
86
+ "model.language_model.layers.1.self_attn.v_proj",
87
+ "model.language_model.layers.30.mlp.gate_proj",
88
  "model.language_model.layers.24.mlp.gate_proj",
89
+ "model.language_model.layers.32.self_attn.o_proj",
90
+ "model.language_model.layers.34.self_attn.o_proj",
91
  "model.language_model.layers.15.self_attn.o_proj",
92
+ "model.language_model.layers.26.mlp.up_proj",
93
+ "model.language_model.layers.29.mlp.down_proj",
94
+ "model.language_model.layers.1.mlp.down_proj",
95
+ "model.language_model.layers.2.self_attn.v_proj",
96
+ "model.language_model.layers.1.self_attn.o_proj",
97
  "model.language_model.layers.4.mlp.down_proj",
98
+ "model.language_model.layers.27.self_attn.q_proj",
99
+ "model.language_model.layers.23.mlp.up_proj",
100
+ "model.language_model.layers.20.self_attn.o_proj",
101
+ "model.language_model.layers.17.mlp.up_proj",
102
+ "model.language_model.layers.1.mlp.gate_proj",
103
+ "model.language_model.layers.23.mlp.down_proj",
104
+ "model.language_model.layers.18.self_attn.o_proj",
105
+ "model.language_model.layers.14.self_attn.v_proj",
 
 
 
 
 
 
 
 
 
 
 
106
  "model.language_model.layers.6.self_attn.k_proj",
107
+ "model.language_model.layers.36.self_attn.q_proj",
108
+ "model.language_model.layers.39.mlp.up_proj",
 
 
 
 
 
 
109
  "model.language_model.layers.20.self_attn.k_proj",
110
+ "model.language_model.layers.23.self_attn.v_proj",
111
+ "model.language_model.layers.12.mlp.up_proj",
112
+ "model.language_model.layers.22.mlp.gate_proj",
113
  "model.language_model.layers.34.self_attn.q_proj",
114
+ "model.language_model.layers.33.self_attn.q_proj",
115
+ "model.language_model.layers.8.mlp.up_proj",
116
+ "model.language_model.layers.7.self_attn.v_proj",
117
+ "model.language_model.layers.5.mlp.up_proj",
118
+ "model.language_model.layers.14.self_attn.q_proj",
 
 
 
 
 
 
119
  "model.language_model.layers.35.mlp.gate_proj",
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
120
  "model.language_model.layers.15.self_attn.k_proj",
121
+ "model.language_model.layers.3.mlp.gate_proj",
122
+ "model.language_model.layers.27.self_attn.o_proj",
123
+ "model.language_model.layers.3.mlp.down_proj",
124
+ "model.language_model.layers.9.self_attn.k_proj",
125
+ "model.language_model.layers.10.self_attn.k_proj",
126
+ "model.language_model.layers.1.self_attn.k_proj",
127
+ "model.language_model.layers.11.mlp.up_proj",
128
+ "model.language_model.layers.26.self_attn.q_proj",
129
+ "model.language_model.layers.32.mlp.gate_proj",
130
+ "model.language_model.layers.7.self_attn.o_proj",
131
+ "model.language_model.layers.27.mlp.down_proj",
132
+ "model.language_model.layers.23.self_attn.o_proj",
133
+ "model.language_model.layers.16.self_attn.q_proj",
134
+ "model.language_model.layers.10.mlp.up_proj",
135
+ "model.language_model.layers.21.mlp.up_proj",
136
+ "model.language_model.layers.5.self_attn.k_proj",
137
+ "model.language_model.layers.33.mlp.gate_proj",
138
+ "model.language_model.layers.9.self_attn.q_proj",
139
+ "model.language_model.layers.8.self_attn.k_proj",
140
+ "model.language_model.layers.1.mlp.up_proj",
141
+ "model.language_model.layers.16.mlp.up_proj",
142
+ "model.language_model.layers.6.mlp.down_proj",
143
+ "model.language_model.layers.22.self_attn.v_proj",
144
+ "model.language_model.layers.7.self_attn.q_proj",
145
  "model.language_model.layers.15.self_attn.v_proj",
146
+ "model.language_model.layers.6.self_attn.o_proj",
147
+ "model.language_model.layers.22.self_attn.o_proj",
148
+ "model.language_model.layers.36.self_attn.o_proj",
149
+ "model.language_model.layers.2.mlp.up_proj",
150
+ "model.language_model.layers.19.self_attn.v_proj",
151
+ "model.language_model.layers.10.mlp.down_proj",
152
+ "model.language_model.layers.0.mlp.up_proj",
153
+ "model.language_model.layers.36.mlp.gate_proj",
154
+ "model.language_model.layers.9.mlp.down_proj",
155
+ "model.language_model.layers.38.mlp.gate_proj",
156
  "model.language_model.layers.2.self_attn.k_proj",
157
+ "model.language_model.layers.15.self_attn.q_proj",
158
  "model.language_model.layers.11.self_attn.v_proj",
159
+ "model.language_model.layers.5.mlp.down_proj",
160
+ "model.language_model.layers.40.self_attn.o_proj",
161
+ "model.language_model.layers.16.mlp.gate_proj",
162
  "model.language_model.layers.38.self_attn.q_proj",
163
+ "model.language_model.layers.29.mlp.gate_proj",
164
+ "model.language_model.layers.17.self_attn.o_proj",
165
+ "model.language_model.layers.24.mlp.down_proj",
166
+ "model.language_model.layers.25.self_attn.q_proj",
167
+ "model.language_model.layers.16.self_attn.v_proj",
168
+ "model.language_model.layers.17.mlp.gate_proj",
169
+ "model.language_model.layers.32.mlp.down_proj",
170
+ "model.language_model.layers.18.self_attn.v_proj",
171
+ "model.language_model.layers.6.self_attn.q_proj",
172
+ "model.language_model.layers.4.mlp.up_proj",
173
+ "model.language_model.layers.22.self_attn.k_proj",
174
+ "model.language_model.layers.13.self_attn.v_proj",
175
+ "model.language_model.layers.28.self_attn.q_proj",
176
+ "model.language_model.layers.26.self_attn.o_proj",
177
+ "model.language_model.layers.28.mlp.up_proj",
178
+ "model.language_model.layers.18.self_attn.k_proj",
179
+ "model.language_model.layers.5.self_attn.o_proj",
180
+ "model.language_model.layers.19.self_attn.o_proj",
181
+ "model.language_model.layers.10.self_attn.v_proj",
182
+ "model.language_model.layers.18.mlp.up_proj",
183
+ "model.language_model.layers.41.mlp.down_proj",
184
+ "model.language_model.layers.24.mlp.up_proj",
185
+ "model.language_model.layers.40.mlp.gate_proj",
186
+ "model.language_model.layers.19.mlp.up_proj",
187
+ "model.language_model.layers.30.mlp.up_proj",
188
+ "model.language_model.layers.25.self_attn.o_proj",
189
+ "model.language_model.layers.35.self_attn.q_proj",
190
+ "model.language_model.layers.14.self_attn.k_proj",
191
+ "model.language_model.layers.39.mlp.down_proj",
192
+ "model.language_model.layers.14.mlp.down_proj",
193
+ "model.language_model.layers.0.mlp.down_proj",
194
+ "model.language_model.layers.21.self_attn.v_proj",
195
+ "model.language_model.layers.9.self_attn.v_proj",
196
+ "model.language_model.layers.21.self_attn.q_proj",
197
+ "model.language_model.layers.20.self_attn.v_proj",
198
+ "model.language_model.layers.31.mlp.gate_proj",
199
+ "model.language_model.layers.3.mlp.up_proj",
200
+ "model.language_model.layers.12.mlp.down_proj",
201
  "model.language_model.layers.37.self_attn.q_proj",
 
 
 
 
 
202
  "model.language_model.layers.13.self_attn.k_proj",
203
+ "model.language_model.layers.28.mlp.down_proj",
204
+ "model.language_model.layers.35.mlp.down_proj",
205
+ "model.language_model.layers.3.self_attn.o_proj",
206
+ "model.language_model.layers.14.mlp.gate_proj",
207
+ "model.language_model.layers.19.self_attn.q_proj",
208
+ "model.language_model.layers.4.self_attn.o_proj",
209
+ "model.language_model.layers.25.mlp.up_proj",
210
+ "model.language_model.layers.12.self_attn.v_proj",
211
+ "model.language_model.layers.26.mlp.gate_proj",
212
+ "model.language_model.layers.22.mlp.down_proj",
213
  "model.language_model.layers.31.self_attn.o_proj",
214
+ "model.language_model.layers.12.self_attn.q_proj",
215
+ "model.language_model.layers.11.mlp.down_proj",
216
+ "model.language_model.layers.37.self_attn.o_proj",
217
+ "model.language_model.layers.2.mlp.down_proj",
218
+ "model.language_model.layers.30.self_attn.o_proj",
219
+ "model.language_model.layers.13.mlp.gate_proj",
220
+ "model.language_model.layers.3.self_attn.v_proj",
221
+ "model.language_model.layers.22.mlp.up_proj",
222
+ "model.language_model.layers.28.self_attn.o_proj",
223
+ "model.language_model.layers.8.mlp.gate_proj",
224
+ "model.language_model.layers.7.mlp.up_proj",
225
+ "model.language_model.layers.10.mlp.gate_proj",
226
+ "model.language_model.layers.15.mlp.gate_proj",
227
+ "model.language_model.layers.10.self_attn.q_proj",
228
  "model.language_model.layers.2.mlp.gate_proj",
229
+ "model.language_model.layers.6.self_attn.v_proj",
230
+ "model.language_model.layers.4.mlp.gate_proj",
231
+ "model.language_model.layers.37.mlp.gate_proj",
232
+ "model.language_model.layers.34.mlp.up_proj",
233
+ "model.language_model.layers.37.mlp.down_proj",
234
+ "model.language_model.layers.37.mlp.up_proj",
235
+ "model.language_model.layers.14.mlp.up_proj",
236
+ "model.language_model.layers.41.mlp.gate_proj",
237
+ "model.language_model.layers.5.self_attn.v_proj",
238
+ "model.language_model.layers.29.self_attn.o_proj",
239
+ "model.language_model.layers.13.self_attn.q_proj",
240
+ "model.language_model.layers.7.self_attn.k_proj",
241
+ "model.language_model.layers.36.mlp.down_proj",
242
+ "model.language_model.layers.5.mlp.gate_proj",
243
+ "model.language_model.layers.12.self_attn.k_proj",
244
+ "model.language_model.layers.39.self_attn.q_proj",
245
+ "model.language_model.layers.19.mlp.down_proj",
246
  "model.language_model.layers.4.self_attn.k_proj",
247
+ "model.language_model.layers.21.self_attn.k_proj",
248
+ "model.language_model.layers.29.self_attn.q_proj",
249
+ "model.language_model.layers.10.self_attn.o_proj",
250
+ "model.language_model.layers.20.self_attn.q_proj",
251
+ "model.language_model.layers.19.self_attn.k_proj",
252
+ "model.language_model.layers.35.mlp.up_proj",
253
+ "model.language_model.layers.39.mlp.gate_proj",
 
 
 
 
254
  "model.language_model.layers.11.mlp.gate_proj",
255
+ "model.language_model.layers.21.mlp.gate_proj",
256
+ "model.language_model.layers.8.self_attn.o_proj",
257
+ "model.language_model.layers.12.mlp.gate_proj",
258
+ "model.language_model.layers.32.self_attn.q_proj",
259
+ "model.language_model.layers.16.self_attn.o_proj",
260
+ "model.language_model.layers.20.mlp.up_proj",
261
+ "model.language_model.layers.13.mlp.up_proj",
262
+ "model.language_model.layers.18.mlp.gate_proj",
263
+ "model.language_model.layers.38.mlp.down_proj",
264
+ "model.language_model.layers.19.mlp.gate_proj",
265
+ "model.language_model.layers.30.self_attn.q_proj",
266
+ "model.language_model.layers.11.self_attn.q_proj",
267
+ "model.language_model.layers.0.self_attn.k_proj",
268
+ "model.language_model.layers.16.self_attn.k_proj",
269
+ "model.language_model.layers.15.mlp.down_proj",
270
+ "model.language_model.layers.36.mlp.up_proj",
271
+ "model.language_model.layers.23.mlp.gate_proj",
272
+ "model.language_model.layers.34.mlp.gate_proj",
273
+ "model.language_model.layers.4.self_attn.v_proj",
274
+ "model.language_model.layers.9.mlp.up_proj",
275
+ "model.language_model.layers.8.self_attn.q_proj",
276
+ "model.language_model.layers.23.self_attn.q_proj",
277
+ "model.language_model.layers.17.self_attn.q_proj",
278
+ "model.language_model.layers.18.self_attn.q_proj",
279
+ "model.language_model.layers.2.self_attn.q_proj",
280
+ "model.language_model.layers.14.self_attn.o_proj",
281
+ "model.language_model.layers.22.self_attn.q_proj",
282
+ "model.language_model.layers.0.self_attn.o_proj",
283
+ "model.language_model.layers.30.mlp.down_proj",
284
+ "model.language_model.layers.1.self_attn.q_proj",
285
+ "model.language_model.layers.41.mlp.up_proj",
286
+ "model.language_model.layers.8.self_attn.v_proj",
287
+ "model.language_model.layers.33.mlp.up_proj",
288
+ "model.language_model.layers.0.self_attn.q_proj",
289
+ "model.language_model.layers.34.mlp.down_proj",
290
+ "model.language_model.layers.9.self_attn.o_proj",
291
+ "model.language_model.layers.40.self_attn.q_proj",
292
+ "model.language_model.layers.26.mlp.down_proj",
293
+ "model.language_model.layers.9.mlp.gate_proj"
294
  ],
295
  "target_parameters": null,
296
  "task_type": null,
adapter_model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:bd6cd5a84c633336f762c335e648dc371e20e22fa6dee7770b4c3b3da7180d6a
3
  size 139602808
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:43c3a132e0621066febf4d9fe0bd7f2301b6247543d8614e0dd24338d13b29cd
3
  size 139602808
ref/adapter_config.json CHANGED
@@ -33,264 +33,264 @@
33
  "rank_pattern": {},
34
  "revision": null,
35
  "target_modules": [
36
- "model.language_model.layers.3.mlp.down_proj",
37
- "model.language_model.layers.27.self_attn.o_proj",
38
- "model.language_model.layers.5.self_attn.k_proj",
39
- "model.language_model.layers.38.mlp.down_proj",
40
- "model.language_model.layers.14.mlp.up_proj",
41
  "model.language_model.layers.35.self_attn.o_proj",
 
 
 
 
 
 
 
 
 
 
 
 
 
 
42
  "model.language_model.layers.6.mlp.up_proj",
43
  "model.language_model.layers.15.mlp.up_proj",
44
- "model.language_model.layers.32.mlp.down_proj",
45
- "model.language_model.layers.0.mlp.gate_proj",
46
- "model.language_model.layers.3.mlp.gate_proj",
47
- "model.language_model.layers.17.self_attn.q_proj",
48
- "model.language_model.layers.33.self_attn.q_proj",
49
- "model.language_model.layers.6.self_attn.o_proj",
50
- "model.language_model.layers.30.self_attn.o_proj",
51
- "model.language_model.layers.32.self_attn.o_proj",
52
- "model.language_model.layers.41.self_attn.o_proj",
53
- "model.language_model.layers.18.mlp.down_proj",
54
- "model.language_model.layers.2.mlp.up_proj",
55
- "model.language_model.layers.8.self_attn.q_proj",
56
- "model.language_model.layers.26.mlp.gate_proj",
57
- "model.language_model.layers.14.self_attn.q_proj",
58
  "model.language_model.layers.17.self_attn.k_proj",
59
- "model.language_model.layers.23.mlp.up_proj",
60
- "model.language_model.layers.0.mlp.up_proj",
61
- "model.language_model.layers.29.mlp.gate_proj",
62
- "model.language_model.layers.18.self_attn.o_proj",
63
- "model.language_model.layers.33.mlp.up_proj",
64
- "model.language_model.layers.33.mlp.gate_proj",
65
- "model.language_model.layers.17.self_attn.o_proj",
66
- "model.language_model.layers.6.self_attn.v_proj",
67
- "model.language_model.layers.12.mlp.gate_proj",
68
- "model.language_model.layers.0.self_attn.q_proj",
69
- "model.language_model.layers.16.self_attn.k_proj",
70
- "model.language_model.layers.8.self_attn.o_proj",
71
- "model.language_model.layers.11.mlp.up_proj",
72
- "model.language_model.layers.40.self_attn.q_proj",
73
- "model.language_model.layers.12.mlp.up_proj",
74
- "model.language_model.layers.39.mlp.gate_proj",
75
- "model.language_model.layers.10.mlp.gate_proj",
76
- "model.language_model.layers.9.self_attn.k_proj",
77
- "model.language_model.layers.5.mlp.up_proj",
78
- "model.language_model.layers.35.self_attn.q_proj",
79
- "model.language_model.layers.18.self_attn.v_proj",
80
- "model.language_model.layers.30.self_attn.q_proj",
81
- "model.language_model.layers.30.mlp.gate_proj",
82
- "model.language_model.layers.1.self_attn.q_proj",
83
- "model.language_model.layers.2.mlp.down_proj",
84
- "model.language_model.layers.30.mlp.up_proj",
85
- "model.language_model.layers.13.mlp.gate_proj",
86
- "model.language_model.layers.37.self_attn.o_proj",
87
- "model.language_model.layers.37.mlp.gate_proj",
88
- "model.language_model.layers.5.self_attn.v_proj",
89
- "model.language_model.layers.6.mlp.down_proj",
90
- "model.language_model.layers.1.mlp.down_proj",
91
- "model.language_model.layers.39.self_attn.o_proj",
92
- "model.language_model.layers.20.mlp.gate_proj",
93
- "model.language_model.layers.35.mlp.up_proj",
94
- "model.language_model.layers.9.self_attn.v_proj",
95
- "model.language_model.layers.7.self_attn.v_proj",
96
- "model.language_model.layers.27.mlp.down_proj",
97
- "model.language_model.layers.36.mlp.down_proj",
98
- "model.language_model.layers.3.self_attn.k_proj",
99
- "model.language_model.layers.9.mlp.gate_proj",
100
- "model.language_model.layers.13.self_attn.q_proj",
101
- "model.language_model.layers.40.mlp.gate_proj",
102
- "model.language_model.layers.4.mlp.up_proj",
103
- "model.language_model.layers.10.self_attn.k_proj",
104
- "model.language_model.layers.1.self_attn.v_proj",
105
- "model.language_model.layers.25.mlp.up_proj",
106
- "model.language_model.layers.18.self_attn.k_proj",
107
- "model.language_model.layers.20.self_attn.v_proj",
108
- "model.language_model.layers.34.mlp.gate_proj",
109
- "model.language_model.layers.15.mlp.down_proj",
110
- "model.language_model.layers.2.self_attn.v_proj",
111
- "model.language_model.layers.36.mlp.gate_proj",
112
  "model.language_model.layers.29.mlp.up_proj",
113
- "model.language_model.layers.12.self_attn.v_proj",
114
- "model.language_model.layers.33.mlp.down_proj",
115
- "model.language_model.layers.13.self_attn.v_proj",
 
 
116
  "model.language_model.layers.7.mlp.gate_proj",
117
- "model.language_model.layers.21.mlp.up_proj",
118
- "model.language_model.layers.21.self_attn.q_proj",
119
- "model.language_model.layers.41.self_attn.q_proj",
120
- "model.language_model.layers.4.self_attn.v_proj",
 
 
121
  "model.language_model.layers.7.mlp.down_proj",
122
- "model.language_model.layers.37.mlp.up_proj",
123
- "model.language_model.layers.23.mlp.down_proj",
124
- "model.language_model.layers.41.mlp.gate_proj",
125
- "model.language_model.layers.25.self_attn.o_proj",
126
- "model.language_model.layers.13.mlp.up_proj",
127
- "model.language_model.layers.26.self_attn.o_proj",
128
- "model.language_model.layers.20.mlp.up_proj",
129
- "model.language_model.layers.8.self_attn.v_proj",
130
- "model.language_model.layers.8.mlp.up_proj",
131
- "model.language_model.layers.31.mlp.down_proj",
132
- "model.language_model.layers.24.mlp.down_proj",
133
- "model.language_model.layers.40.mlp.up_proj",
134
- "model.language_model.layers.23.self_attn.o_proj",
135
- "model.language_model.layers.6.self_attn.q_proj",
136
- "model.language_model.layers.22.self_attn.v_proj",
137
- "model.language_model.layers.34.self_attn.o_proj",
138
- "model.language_model.layers.12.self_attn.o_proj",
139
- "model.language_model.layers.1.mlp.up_proj",
140
- "model.language_model.layers.29.self_attn.o_proj",
141
- "model.language_model.layers.19.mlp.gate_proj",
142
  "model.language_model.layers.24.self_attn.o_proj",
143
- "model.language_model.layers.38.mlp.up_proj",
144
- "model.language_model.layers.23.self_attn.v_proj",
145
- "model.language_model.layers.3.mlp.up_proj",
146
- "model.language_model.layers.21.self_attn.v_proj",
147
- "model.language_model.layers.1.self_attn.k_proj",
148
- "model.language_model.layers.17.mlp.up_proj",
149
- "model.language_model.layers.39.self_attn.q_proj",
150
- "model.language_model.layers.20.mlp.down_proj",
151
- "model.language_model.layers.34.mlp.up_proj",
152
- "model.language_model.layers.34.mlp.down_proj",
153
- "model.language_model.layers.19.self_attn.o_proj",
154
- "model.language_model.layers.21.mlp.gate_proj",
155
- "model.language_model.layers.8.mlp.down_proj",
156
- "model.language_model.layers.38.mlp.gate_proj",
157
- "model.language_model.layers.24.mlp.up_proj",
158
- "model.language_model.layers.10.self_attn.v_proj",
159
- "model.language_model.layers.2.self_attn.q_proj",
160
- "model.language_model.layers.22.mlp.gate_proj",
161
- "model.language_model.layers.36.mlp.up_proj",
162
- "model.language_model.layers.4.self_attn.o_proj",
163
- "model.language_model.layers.0.mlp.down_proj",
164
- "model.language_model.layers.36.self_attn.o_proj",
165
  "model.language_model.layers.11.self_attn.o_proj",
 
 
 
166
  "model.language_model.layers.24.mlp.gate_proj",
167
- "model.language_model.layers.14.mlp.down_proj",
 
168
  "model.language_model.layers.15.self_attn.o_proj",
169
- "model.language_model.layers.31.mlp.up_proj",
170
- "model.language_model.layers.28.self_attn.q_proj",
171
- "model.language_model.layers.16.self_attn.q_proj",
172
- "model.language_model.layers.19.mlp.down_proj",
 
173
  "model.language_model.layers.4.mlp.down_proj",
174
- "model.language_model.layers.37.mlp.down_proj",
175
- "model.language_model.layers.31.self_attn.q_proj",
176
- "model.language_model.layers.5.mlp.gate_proj",
177
- "model.language_model.layers.23.mlp.gate_proj",
178
- "model.language_model.layers.3.self_attn.o_proj",
179
- "model.language_model.layers.5.self_attn.q_proj",
180
- "model.language_model.layers.12.self_attn.k_proj",
181
- "model.language_model.layers.0.self_attn.o_proj",
182
- "model.language_model.layers.14.self_attn.o_proj",
183
- "model.language_model.layers.16.mlp.up_proj",
184
- "model.language_model.layers.22.self_attn.q_proj",
185
- "model.language_model.layers.27.mlp.up_proj",
186
- "model.language_model.layers.41.mlp.up_proj",
187
- "model.language_model.layers.28.self_attn.o_proj",
188
- "model.language_model.layers.10.mlp.down_proj",
189
- "model.language_model.layers.9.self_attn.q_proj",
190
- "model.language_model.layers.39.mlp.up_proj",
191
- "model.language_model.layers.32.mlp.up_proj",
192
- "model.language_model.layers.12.self_attn.q_proj",
193
  "model.language_model.layers.6.self_attn.k_proj",
194
- "model.language_model.layers.40.self_attn.o_proj",
195
- "model.language_model.layers.20.self_attn.q_proj",
196
- "model.language_model.layers.35.mlp.down_proj",
197
- "model.language_model.layers.5.mlp.down_proj",
198
- "model.language_model.layers.21.mlp.down_proj",
199
- "model.language_model.layers.9.mlp.down_proj",
200
- "model.language_model.layers.21.self_attn.k_proj",
201
- "model.language_model.layers.28.mlp.up_proj",
202
  "model.language_model.layers.20.self_attn.k_proj",
 
 
 
203
  "model.language_model.layers.34.self_attn.q_proj",
204
- "model.language_model.layers.7.mlp.up_proj",
205
- "model.language_model.layers.2.self_attn.o_proj",
206
- "model.language_model.layers.9.mlp.up_proj",
207
- "model.language_model.layers.19.mlp.up_proj",
208
- "model.language_model.layers.22.mlp.down_proj",
209
- "model.language_model.layers.10.self_attn.o_proj",
210
- "model.language_model.layers.29.self_attn.q_proj",
211
- "model.language_model.layers.27.self_attn.q_proj",
212
- "model.language_model.layers.14.mlp.gate_proj",
213
- "model.language_model.layers.33.self_attn.o_proj",
214
- "model.language_model.layers.22.self_attn.k_proj",
215
  "model.language_model.layers.35.mlp.gate_proj",
216
- "model.language_model.layers.5.self_attn.o_proj",
217
- "model.language_model.layers.25.mlp.down_proj",
218
- "model.language_model.layers.3.self_attn.q_proj",
219
- "model.language_model.layers.0.self_attn.k_proj",
220
- "model.language_model.layers.28.mlp.down_proj",
221
- "model.language_model.layers.7.self_attn.o_proj",
222
- "model.language_model.layers.32.mlp.gate_proj",
223
- "model.language_model.layers.25.self_attn.q_proj",
224
- "model.language_model.layers.21.self_attn.o_proj",
225
- "model.language_model.layers.18.self_attn.q_proj",
226
- "model.language_model.layers.17.mlp.down_proj",
227
- "model.language_model.layers.3.self_attn.v_proj",
228
- "model.language_model.layers.23.self_attn.q_proj",
229
- "model.language_model.layers.13.self_attn.o_proj",
230
- "model.language_model.layers.15.mlp.gate_proj",
231
- "model.language_model.layers.27.mlp.gate_proj",
232
- "model.language_model.layers.31.mlp.gate_proj",
233
- "model.language_model.layers.1.self_attn.o_proj",
234
- "model.language_model.layers.28.mlp.gate_proj",
235
- "model.language_model.layers.16.mlp.down_proj",
236
- "model.language_model.layers.14.self_attn.k_proj",
237
- "model.language_model.layers.4.self_attn.q_proj",
238
- "model.language_model.layers.30.mlp.down_proj",
239
  "model.language_model.layers.15.self_attn.k_proj",
240
- "model.language_model.layers.19.self_attn.q_proj",
241
- "model.language_model.layers.16.self_attn.v_proj",
242
- "model.language_model.layers.40.mlp.down_proj",
243
- "model.language_model.layers.25.mlp.gate_proj",
244
- "model.language_model.layers.19.self_attn.k_proj",
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
245
  "model.language_model.layers.15.self_attn.v_proj",
246
- "model.language_model.layers.36.self_attn.q_proj",
247
- "model.language_model.layers.17.self_attn.v_proj",
248
- "model.language_model.layers.32.self_attn.q_proj",
249
- "model.language_model.layers.14.self_attn.v_proj",
250
- "model.language_model.layers.26.mlp.up_proj",
251
- "model.language_model.layers.10.self_attn.q_proj",
252
- "model.language_model.layers.17.mlp.gate_proj",
253
- "model.language_model.layers.1.mlp.gate_proj",
254
- "model.language_model.layers.39.mlp.down_proj",
 
255
  "model.language_model.layers.2.self_attn.k_proj",
256
- "model.language_model.layers.4.mlp.gate_proj",
257
  "model.language_model.layers.11.self_attn.v_proj",
258
- "model.language_model.layers.8.mlp.gate_proj",
 
 
259
  "model.language_model.layers.38.self_attn.q_proj",
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
260
  "model.language_model.layers.37.self_attn.q_proj",
261
- "model.language_model.layers.11.self_attn.k_proj",
262
- "model.language_model.layers.20.self_attn.o_proj",
263
- "model.language_model.layers.26.self_attn.q_proj",
264
- "model.language_model.layers.18.mlp.gate_proj",
265
- "model.language_model.layers.7.self_attn.k_proj",
266
  "model.language_model.layers.13.self_attn.k_proj",
267
- "model.language_model.layers.24.self_attn.q_proj",
268
- "model.language_model.layers.23.self_attn.k_proj",
269
- "model.language_model.layers.9.self_attn.o_proj",
270
- "model.language_model.layers.22.mlp.up_proj",
 
 
 
 
 
 
271
  "model.language_model.layers.31.self_attn.o_proj",
272
- "model.language_model.layers.8.self_attn.k_proj",
273
- "model.language_model.layers.6.mlp.gate_proj",
274
- "model.language_model.layers.15.self_attn.q_proj",
 
 
 
 
 
 
 
 
 
 
 
275
  "model.language_model.layers.2.mlp.gate_proj",
276
- "model.language_model.layers.16.mlp.gate_proj",
277
- "model.language_model.layers.10.mlp.up_proj",
278
- "model.language_model.layers.26.mlp.down_proj",
 
 
 
 
 
 
 
 
 
 
 
 
 
 
279
  "model.language_model.layers.4.self_attn.k_proj",
280
- "model.language_model.layers.0.self_attn.v_proj",
281
- "model.language_model.layers.38.self_attn.o_proj",
282
- "model.language_model.layers.11.self_attn.q_proj",
283
- "model.language_model.layers.7.self_attn.q_proj",
284
- "model.language_model.layers.41.mlp.down_proj",
285
- "model.language_model.layers.11.mlp.down_proj",
286
- "model.language_model.layers.22.self_attn.o_proj",
287
- "model.language_model.layers.29.mlp.down_proj",
288
- "model.language_model.layers.16.self_attn.o_proj",
289
- "model.language_model.layers.18.mlp.up_proj",
290
- "model.language_model.layers.13.mlp.down_proj",
291
  "model.language_model.layers.11.mlp.gate_proj",
292
- "model.language_model.layers.12.mlp.down_proj",
293
- "model.language_model.layers.19.self_attn.v_proj"
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
294
  ],
295
  "target_parameters": null,
296
  "task_type": null,
 
33
  "rank_pattern": {},
34
  "revision": null,
35
  "target_modules": [
36
+ "model.language_model.layers.33.mlp.down_proj",
37
+ "model.language_model.layers.41.self_attn.q_proj",
38
+ "model.language_model.layers.6.mlp.gate_proj",
39
+ "model.language_model.layers.18.mlp.down_proj",
 
40
  "model.language_model.layers.35.self_attn.o_proj",
41
+ "model.language_model.layers.17.mlp.down_proj",
42
+ "model.language_model.layers.4.self_attn.q_proj",
43
+ "model.language_model.layers.31.mlp.down_proj",
44
+ "model.language_model.layers.2.self_attn.o_proj",
45
+ "model.language_model.layers.33.self_attn.o_proj",
46
+ "model.language_model.layers.16.mlp.down_proj",
47
+ "model.language_model.layers.41.self_attn.o_proj",
48
+ "model.language_model.layers.8.mlp.down_proj",
49
+ "model.language_model.layers.40.mlp.down_proj",
50
+ "model.language_model.layers.20.mlp.gate_proj",
51
+ "model.language_model.layers.17.self_attn.v_proj",
52
+ "model.language_model.layers.23.self_attn.k_proj",
53
+ "model.language_model.layers.3.self_attn.q_proj",
54
+ "model.language_model.layers.38.self_attn.o_proj",
55
  "model.language_model.layers.6.mlp.up_proj",
56
  "model.language_model.layers.15.mlp.up_proj",
 
 
 
 
 
 
 
 
 
 
 
 
 
 
57
  "model.language_model.layers.17.self_attn.k_proj",
58
+ "model.language_model.layers.27.mlp.up_proj",
59
+ "model.language_model.layers.11.self_attn.k_proj",
60
+ "model.language_model.layers.21.self_attn.o_proj",
61
+ "model.language_model.layers.0.self_attn.v_proj",
62
+ "model.language_model.layers.38.mlp.up_proj",
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
63
  "model.language_model.layers.29.mlp.up_proj",
64
+ "model.language_model.layers.31.mlp.up_proj",
65
+ "model.language_model.layers.28.mlp.gate_proj",
66
+ "model.language_model.layers.13.self_attn.o_proj",
67
+ "model.language_model.layers.32.mlp.up_proj",
68
+ "model.language_model.layers.13.mlp.down_proj",
69
  "model.language_model.layers.7.mlp.gate_proj",
70
+ "model.language_model.layers.20.mlp.down_proj",
71
+ "model.language_model.layers.3.self_attn.k_proj",
72
+ "model.language_model.layers.0.mlp.gate_proj",
73
+ "model.language_model.layers.25.mlp.down_proj",
74
+ "model.language_model.layers.21.mlp.down_proj",
75
+ "model.language_model.layers.25.mlp.gate_proj",
76
  "model.language_model.layers.7.mlp.down_proj",
77
+ "model.language_model.layers.5.self_attn.q_proj",
78
+ "model.language_model.layers.39.self_attn.o_proj",
79
+ "model.language_model.layers.24.self_attn.q_proj",
80
+ "model.language_model.layers.27.mlp.gate_proj",
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
81
  "model.language_model.layers.24.self_attn.o_proj",
82
+ "model.language_model.layers.31.self_attn.q_proj",
83
+ "model.language_model.layers.12.self_attn.o_proj",
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
84
  "model.language_model.layers.11.self_attn.o_proj",
85
+ "model.language_model.layers.40.mlp.up_proj",
86
+ "model.language_model.layers.1.self_attn.v_proj",
87
+ "model.language_model.layers.30.mlp.gate_proj",
88
  "model.language_model.layers.24.mlp.gate_proj",
89
+ "model.language_model.layers.32.self_attn.o_proj",
90
+ "model.language_model.layers.34.self_attn.o_proj",
91
  "model.language_model.layers.15.self_attn.o_proj",
92
+ "model.language_model.layers.26.mlp.up_proj",
93
+ "model.language_model.layers.29.mlp.down_proj",
94
+ "model.language_model.layers.1.mlp.down_proj",
95
+ "model.language_model.layers.2.self_attn.v_proj",
96
+ "model.language_model.layers.1.self_attn.o_proj",
97
  "model.language_model.layers.4.mlp.down_proj",
98
+ "model.language_model.layers.27.self_attn.q_proj",
99
+ "model.language_model.layers.23.mlp.up_proj",
100
+ "model.language_model.layers.20.self_attn.o_proj",
101
+ "model.language_model.layers.17.mlp.up_proj",
102
+ "model.language_model.layers.1.mlp.gate_proj",
103
+ "model.language_model.layers.23.mlp.down_proj",
104
+ "model.language_model.layers.18.self_attn.o_proj",
105
+ "model.language_model.layers.14.self_attn.v_proj",
 
 
 
 
 
 
 
 
 
 
 
106
  "model.language_model.layers.6.self_attn.k_proj",
107
+ "model.language_model.layers.36.self_attn.q_proj",
108
+ "model.language_model.layers.39.mlp.up_proj",
 
 
 
 
 
 
109
  "model.language_model.layers.20.self_attn.k_proj",
110
+ "model.language_model.layers.23.self_attn.v_proj",
111
+ "model.language_model.layers.12.mlp.up_proj",
112
+ "model.language_model.layers.22.mlp.gate_proj",
113
  "model.language_model.layers.34.self_attn.q_proj",
114
+ "model.language_model.layers.33.self_attn.q_proj",
115
+ "model.language_model.layers.8.mlp.up_proj",
116
+ "model.language_model.layers.7.self_attn.v_proj",
117
+ "model.language_model.layers.5.mlp.up_proj",
118
+ "model.language_model.layers.14.self_attn.q_proj",
 
 
 
 
 
 
119
  "model.language_model.layers.35.mlp.gate_proj",
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
120
  "model.language_model.layers.15.self_attn.k_proj",
121
+ "model.language_model.layers.3.mlp.gate_proj",
122
+ "model.language_model.layers.27.self_attn.o_proj",
123
+ "model.language_model.layers.3.mlp.down_proj",
124
+ "model.language_model.layers.9.self_attn.k_proj",
125
+ "model.language_model.layers.10.self_attn.k_proj",
126
+ "model.language_model.layers.1.self_attn.k_proj",
127
+ "model.language_model.layers.11.mlp.up_proj",
128
+ "model.language_model.layers.26.self_attn.q_proj",
129
+ "model.language_model.layers.32.mlp.gate_proj",
130
+ "model.language_model.layers.7.self_attn.o_proj",
131
+ "model.language_model.layers.27.mlp.down_proj",
132
+ "model.language_model.layers.23.self_attn.o_proj",
133
+ "model.language_model.layers.16.self_attn.q_proj",
134
+ "model.language_model.layers.10.mlp.up_proj",
135
+ "model.language_model.layers.21.mlp.up_proj",
136
+ "model.language_model.layers.5.self_attn.k_proj",
137
+ "model.language_model.layers.33.mlp.gate_proj",
138
+ "model.language_model.layers.9.self_attn.q_proj",
139
+ "model.language_model.layers.8.self_attn.k_proj",
140
+ "model.language_model.layers.1.mlp.up_proj",
141
+ "model.language_model.layers.16.mlp.up_proj",
142
+ "model.language_model.layers.6.mlp.down_proj",
143
+ "model.language_model.layers.22.self_attn.v_proj",
144
+ "model.language_model.layers.7.self_attn.q_proj",
145
  "model.language_model.layers.15.self_attn.v_proj",
146
+ "model.language_model.layers.6.self_attn.o_proj",
147
+ "model.language_model.layers.22.self_attn.o_proj",
148
+ "model.language_model.layers.36.self_attn.o_proj",
149
+ "model.language_model.layers.2.mlp.up_proj",
150
+ "model.language_model.layers.19.self_attn.v_proj",
151
+ "model.language_model.layers.10.mlp.down_proj",
152
+ "model.language_model.layers.0.mlp.up_proj",
153
+ "model.language_model.layers.36.mlp.gate_proj",
154
+ "model.language_model.layers.9.mlp.down_proj",
155
+ "model.language_model.layers.38.mlp.gate_proj",
156
  "model.language_model.layers.2.self_attn.k_proj",
157
+ "model.language_model.layers.15.self_attn.q_proj",
158
  "model.language_model.layers.11.self_attn.v_proj",
159
+ "model.language_model.layers.5.mlp.down_proj",
160
+ "model.language_model.layers.40.self_attn.o_proj",
161
+ "model.language_model.layers.16.mlp.gate_proj",
162
  "model.language_model.layers.38.self_attn.q_proj",
163
+ "model.language_model.layers.29.mlp.gate_proj",
164
+ "model.language_model.layers.17.self_attn.o_proj",
165
+ "model.language_model.layers.24.mlp.down_proj",
166
+ "model.language_model.layers.25.self_attn.q_proj",
167
+ "model.language_model.layers.16.self_attn.v_proj",
168
+ "model.language_model.layers.17.mlp.gate_proj",
169
+ "model.language_model.layers.32.mlp.down_proj",
170
+ "model.language_model.layers.18.self_attn.v_proj",
171
+ "model.language_model.layers.6.self_attn.q_proj",
172
+ "model.language_model.layers.4.mlp.up_proj",
173
+ "model.language_model.layers.22.self_attn.k_proj",
174
+ "model.language_model.layers.13.self_attn.v_proj",
175
+ "model.language_model.layers.28.self_attn.q_proj",
176
+ "model.language_model.layers.26.self_attn.o_proj",
177
+ "model.language_model.layers.28.mlp.up_proj",
178
+ "model.language_model.layers.18.self_attn.k_proj",
179
+ "model.language_model.layers.5.self_attn.o_proj",
180
+ "model.language_model.layers.19.self_attn.o_proj",
181
+ "model.language_model.layers.10.self_attn.v_proj",
182
+ "model.language_model.layers.18.mlp.up_proj",
183
+ "model.language_model.layers.41.mlp.down_proj",
184
+ "model.language_model.layers.24.mlp.up_proj",
185
+ "model.language_model.layers.40.mlp.gate_proj",
186
+ "model.language_model.layers.19.mlp.up_proj",
187
+ "model.language_model.layers.30.mlp.up_proj",
188
+ "model.language_model.layers.25.self_attn.o_proj",
189
+ "model.language_model.layers.35.self_attn.q_proj",
190
+ "model.language_model.layers.14.self_attn.k_proj",
191
+ "model.language_model.layers.39.mlp.down_proj",
192
+ "model.language_model.layers.14.mlp.down_proj",
193
+ "model.language_model.layers.0.mlp.down_proj",
194
+ "model.language_model.layers.21.self_attn.v_proj",
195
+ "model.language_model.layers.9.self_attn.v_proj",
196
+ "model.language_model.layers.21.self_attn.q_proj",
197
+ "model.language_model.layers.20.self_attn.v_proj",
198
+ "model.language_model.layers.31.mlp.gate_proj",
199
+ "model.language_model.layers.3.mlp.up_proj",
200
+ "model.language_model.layers.12.mlp.down_proj",
201
  "model.language_model.layers.37.self_attn.q_proj",
 
 
 
 
 
202
  "model.language_model.layers.13.self_attn.k_proj",
203
+ "model.language_model.layers.28.mlp.down_proj",
204
+ "model.language_model.layers.35.mlp.down_proj",
205
+ "model.language_model.layers.3.self_attn.o_proj",
206
+ "model.language_model.layers.14.mlp.gate_proj",
207
+ "model.language_model.layers.19.self_attn.q_proj",
208
+ "model.language_model.layers.4.self_attn.o_proj",
209
+ "model.language_model.layers.25.mlp.up_proj",
210
+ "model.language_model.layers.12.self_attn.v_proj",
211
+ "model.language_model.layers.26.mlp.gate_proj",
212
+ "model.language_model.layers.22.mlp.down_proj",
213
  "model.language_model.layers.31.self_attn.o_proj",
214
+ "model.language_model.layers.12.self_attn.q_proj",
215
+ "model.language_model.layers.11.mlp.down_proj",
216
+ "model.language_model.layers.37.self_attn.o_proj",
217
+ "model.language_model.layers.2.mlp.down_proj",
218
+ "model.language_model.layers.30.self_attn.o_proj",
219
+ "model.language_model.layers.13.mlp.gate_proj",
220
+ "model.language_model.layers.3.self_attn.v_proj",
221
+ "model.language_model.layers.22.mlp.up_proj",
222
+ "model.language_model.layers.28.self_attn.o_proj",
223
+ "model.language_model.layers.8.mlp.gate_proj",
224
+ "model.language_model.layers.7.mlp.up_proj",
225
+ "model.language_model.layers.10.mlp.gate_proj",
226
+ "model.language_model.layers.15.mlp.gate_proj",
227
+ "model.language_model.layers.10.self_attn.q_proj",
228
  "model.language_model.layers.2.mlp.gate_proj",
229
+ "model.language_model.layers.6.self_attn.v_proj",
230
+ "model.language_model.layers.4.mlp.gate_proj",
231
+ "model.language_model.layers.37.mlp.gate_proj",
232
+ "model.language_model.layers.34.mlp.up_proj",
233
+ "model.language_model.layers.37.mlp.down_proj",
234
+ "model.language_model.layers.37.mlp.up_proj",
235
+ "model.language_model.layers.14.mlp.up_proj",
236
+ "model.language_model.layers.41.mlp.gate_proj",
237
+ "model.language_model.layers.5.self_attn.v_proj",
238
+ "model.language_model.layers.29.self_attn.o_proj",
239
+ "model.language_model.layers.13.self_attn.q_proj",
240
+ "model.language_model.layers.7.self_attn.k_proj",
241
+ "model.language_model.layers.36.mlp.down_proj",
242
+ "model.language_model.layers.5.mlp.gate_proj",
243
+ "model.language_model.layers.12.self_attn.k_proj",
244
+ "model.language_model.layers.39.self_attn.q_proj",
245
+ "model.language_model.layers.19.mlp.down_proj",
246
  "model.language_model.layers.4.self_attn.k_proj",
247
+ "model.language_model.layers.21.self_attn.k_proj",
248
+ "model.language_model.layers.29.self_attn.q_proj",
249
+ "model.language_model.layers.10.self_attn.o_proj",
250
+ "model.language_model.layers.20.self_attn.q_proj",
251
+ "model.language_model.layers.19.self_attn.k_proj",
252
+ "model.language_model.layers.35.mlp.up_proj",
253
+ "model.language_model.layers.39.mlp.gate_proj",
 
 
 
 
254
  "model.language_model.layers.11.mlp.gate_proj",
255
+ "model.language_model.layers.21.mlp.gate_proj",
256
+ "model.language_model.layers.8.self_attn.o_proj",
257
+ "model.language_model.layers.12.mlp.gate_proj",
258
+ "model.language_model.layers.32.self_attn.q_proj",
259
+ "model.language_model.layers.16.self_attn.o_proj",
260
+ "model.language_model.layers.20.mlp.up_proj",
261
+ "model.language_model.layers.13.mlp.up_proj",
262
+ "model.language_model.layers.18.mlp.gate_proj",
263
+ "model.language_model.layers.38.mlp.down_proj",
264
+ "model.language_model.layers.19.mlp.gate_proj",
265
+ "model.language_model.layers.30.self_attn.q_proj",
266
+ "model.language_model.layers.11.self_attn.q_proj",
267
+ "model.language_model.layers.0.self_attn.k_proj",
268
+ "model.language_model.layers.16.self_attn.k_proj",
269
+ "model.language_model.layers.15.mlp.down_proj",
270
+ "model.language_model.layers.36.mlp.up_proj",
271
+ "model.language_model.layers.23.mlp.gate_proj",
272
+ "model.language_model.layers.34.mlp.gate_proj",
273
+ "model.language_model.layers.4.self_attn.v_proj",
274
+ "model.language_model.layers.9.mlp.up_proj",
275
+ "model.language_model.layers.8.self_attn.q_proj",
276
+ "model.language_model.layers.23.self_attn.q_proj",
277
+ "model.language_model.layers.17.self_attn.q_proj",
278
+ "model.language_model.layers.18.self_attn.q_proj",
279
+ "model.language_model.layers.2.self_attn.q_proj",
280
+ "model.language_model.layers.14.self_attn.o_proj",
281
+ "model.language_model.layers.22.self_attn.q_proj",
282
+ "model.language_model.layers.0.self_attn.o_proj",
283
+ "model.language_model.layers.30.mlp.down_proj",
284
+ "model.language_model.layers.1.self_attn.q_proj",
285
+ "model.language_model.layers.41.mlp.up_proj",
286
+ "model.language_model.layers.8.self_attn.v_proj",
287
+ "model.language_model.layers.33.mlp.up_proj",
288
+ "model.language_model.layers.0.self_attn.q_proj",
289
+ "model.language_model.layers.34.mlp.down_proj",
290
+ "model.language_model.layers.9.self_attn.o_proj",
291
+ "model.language_model.layers.40.self_attn.q_proj",
292
+ "model.language_model.layers.26.mlp.down_proj",
293
+ "model.language_model.layers.9.mlp.gate_proj"
294
  ],
295
  "target_parameters": null,
296
  "task_type": null,
ref/adapter_model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:97c86b939ff438fad2206ca0202b774f98f2dbff96401b60061cf33b548f69c5
3
  size 139602808
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:52aec38d57dbd827834cb4c41459b888f756042a80fb043f99efc723f2ea793c
3
  size 139602808
release-provenance.json CHANGED
@@ -1,28 +1,128 @@
1
  {
2
- "checkpointRunId": "alpino-e4b-v02-safety-31967678398",
3
- "compilerSourceCommit": "6bdfbb20fcea851623dda4dd9b661003363b1802",
4
- "immutableBenchmarkHistory": {
5
- "aooeValid": 14,
6
- "attempted": 24,
7
- "rawValid": 16,
8
- "status": "pre-replacement-runtime-history-not-replacement-evidence",
9
- "trainingContamination": false,
10
- "workflowRunId": 31998534841
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
11
  },
12
  "promotionApproved": true,
13
- "replacementDevGate": {
14
- "aooeTokens": 2894,
15
- "artifactDigest": "sha256:4d6ef1f2fc3daafc40857d5404aa6165df1115793ed57dc18f93f64f2c547da6",
16
- "artifactId": 9278313484,
17
- "attempted": 6,
18
- "correctAbstentions": 1,
19
- "fixtures": "external-development-only",
20
- "rawTokens": 4696,
21
- "rawValid": 4,
22
- "unsafeMutations": 0,
23
- "valid": 6,
24
- "workflowRunId": 32000363618
25
  },
 
26
  "target": "cyberandy/Alpino-e4b-v02",
27
- "weightUpdatesAfterA3": false
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
28
  }
 
1
  {
2
+ "adapterParameterChecksum": "0c68b19cdbf8523c7bd57e2869e945a7a9df1511d3e07f711e95d33298af2c7e",
3
+ "baseModel": "google/gemma-4-E4B-it",
4
+ "baseRevision": "ee0ef6023621cff504d758262d4e04895a5af4a2",
5
+ "checkpointRunId": "alpino-e4b-v02-controller-escalation-32032090932",
6
+ "checkpointSubdir": "grpo-v5",
7
+ "freezePolicy": {
8
+ "frozenVersion": "v0.2",
9
+ "futureTrainingTrack": "v0.3",
10
+ "noFurtherWeightUpdates": true,
11
+ "note": "A.9 is the final evidence-bearing v0.2 checkpoint. Further GRPO must use a new version lineage and must not mutate or overwrite this checkpoint.",
12
+ "v03PrimaryFocus": "cross-entity reasoning"
13
+ },
14
+ "heldoutDevelopment": {
15
+ "aooeSymbolicConformant": 6,
16
+ "artifactId": 9290659167,
17
+ "artifactZipSha256": "178795fe76ab2d74ab93b133ac52d5b94b50e55ff0e88c3b6896c052954f27de",
18
+ "attemptedPerArm": 6,
19
+ "benchmarkLoaded": false,
20
+ "immutableV2Loaded": false,
21
+ "promotionGatePassed": true,
22
+ "rawSymbolicConformant": 5,
23
+ "resultJsonSha256": "2680924f1cf5ed1d3a46608bfa0f175e95781cf4d42bceac2147e42aaeb16874",
24
+ "runId": 32035764957,
25
+ "sourceCommit": "57cdd537b45a6aa245497c5170ede0c8fbcef6ef",
26
+ "status": "passed",
27
+ "weightUpdates": false
28
+ },
29
+ "immutableBenchmark": {
30
+ "artifactId": 9292050963,
31
+ "artifactZipSha256": "333edd6741ef33b54f4b680340e774721be8a1b8912450af8492e69527925b8a",
32
+ "branch": "feat/alpina-governed-content-track",
33
+ "evaluationPolicy": "single-successful-run-no-tuning-after-reveal",
34
+ "headSha": "cd497abd02683d00dfd0d06d2c5b42db9883129e",
35
+ "id": "alpina-immutable-v2",
36
+ "pairedDelta": {
37
+ "tokenReductionFraction": 0.7210104744300678,
38
+ "tokensPerValidTransitionReductionFraction": 0.7768083795440542,
39
+ "validTransitions": 4,
40
+ "validTransitionsPerPflopRatio": 4.480454946996467
41
+ },
42
+ "repository": "wordlift/aooe-benchmarks",
43
+ "resultJsonSha256": "6c3b089a2342fcf3f983998c9067a658f9a9a80ad893092cc2a852049afb9090",
44
+ "runId": 32041360569,
45
+ "scenarioCount": 24,
46
+ "seedRange": "72000-72023",
47
+ "status": "evaluated",
48
+ "summary": {
49
+ "aooe": {
50
+ "attempted": 24,
51
+ "correctAbstentions": 4,
52
+ "tokensPerValidTransition": 566.0,
53
+ "totalTokens": 11320,
54
+ "unsafeMutations": 0,
55
+ "valid": 20,
56
+ "validTransitionsPerPflop": 196.30938358853552
57
+ },
58
+ "raw": {
59
+ "attempted": 24,
60
+ "correctAbstentions": 0,
61
+ "tokensPerValidTransition": 2535.9375,
62
+ "totalTokens": 40575,
63
+ "unsafeMutations": 1,
64
+ "valid": 16,
65
+ "validTransitionsPerPflop": 43.81460943383309
66
+ }
67
+ },
68
+ "validByClass": {
69
+ "aooe": {
70
+ "create": 3,
71
+ "cross-entity": 1,
72
+ "edit": 4,
73
+ "inspect": 4,
74
+ "publish": 4,
75
+ "unsafe-ambiguous": 4
76
+ },
77
+ "raw": {
78
+ "create": 4,
79
+ "cross-entity": 0,
80
+ "edit": 4,
81
+ "inspect": 4,
82
+ "publish": 4,
83
+ "unsafe-ambiguous": 0
84
+ }
85
+ }
86
  },
87
  "promotionApproved": true,
88
+ "publicationCheckpointPath": "/vol/checkpoints/alpino-e4b-v02-controller-escalation-32032090932/grpo-v5/adapter",
89
+ "releaseSchema": "alpino-v02-release-v2",
90
+ "runtime": {
91
+ "compiler": "AOOE controller decision -> deterministic governed transition lowering -> symbolic validation",
92
+ "compilerBlobSha": "17c36dbccdc0cbb845f6894a9de3b4106eb116f2",
93
+ "inferenceRebindRunId": 32041288206,
94
+ "inferenceSourceCommit": "3ddb0d5e2a5349408e6902c354f1b40fc30ce7a1",
95
+ "rewardContract": "SHACL-hard-gate-plus-task-evaluator-v1",
96
+ "trainingProtocol": "governed-state-transition-json-v1"
 
 
 
97
  },
98
+ "serializedWeightsSha256": "43c3a132e0621066febf4d9fe0bd7f2301b6247543d8614e0dd24338d13b29cd",
99
  "target": "cyberandy/Alpino-e4b-v02",
100
+ "training": {
101
+ "artifactId": 9289791005,
102
+ "artifactZipSha256": "d10bdbc8c09f5983a4c1fdc1f3818d16488f9c6899d0819a289d6f2ef79a472f",
103
+ "benchmarkDataLoaded": false,
104
+ "datasetRows": 100,
105
+ "devV3Loaded": false,
106
+ "groupSize": 8,
107
+ "immutableV2Loaded": false,
108
+ "learningRate": 7.5e-07,
109
+ "maxSteps": 80,
110
+ "phase": "A.9-controller-escalation-recovery",
111
+ "reportJsonSha256": "b29d71ab332099358c4363d583c4bb2e660c5221b98e2d1206e3114491567503",
112
+ "runId": 32032090932,
113
+ "seed": 13103,
114
+ "sourceCommit": "ff507984dd11a5c33ea9e5c1dbaea4e853d87f84",
115
+ "startCheckpointRunId": "alpino-e4b-v02-controller-recovery-32020398211",
116
+ "trainingGate": {
117
+ "attempted": 8,
118
+ "crossEntityAttempted": 4,
119
+ "crossEntityLinkExisting": 4,
120
+ "fullReward": 4,
121
+ "semanticPerfect": 7,
122
+ "symbolicConformant": 8,
123
+ "unsafeAbstainEscalate": 4,
124
+ "unsafeAttempted": 4
125
+ },
126
+ "weightsChanged": true
127
+ }
128
  }
training_args.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:e710644c821e23d0384375a7bc5614eb52c1b673e61ff6bd094341c87f5d5b4d
3
  size 7633
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7e9f0d64f235c9261d004c9163d2a719c773512f56cef2ca13785e0f47688d0a
3
  size 7633