Mapika commited on
Commit
09e9e33
·
verified ·
1 Parent(s): dcd2155

decider-12b v1: Gemma-4-12B-it (707f0a3b, unchanged weights) with the decider chat readout and per-type temperatures

Browse files
.gitattributes CHANGED
@@ -33,3 +33,4 @@ saved_model/**/* 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
 
 
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
+ tokenizer.json filter=lfs diff=lfs merge=lfs -text
README.md ADDED
@@ -0,0 +1,83 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ base_model: google/gemma-4-12B-it
4
+ language: [en]
5
+ pipeline_tag: text-classification
6
+ tags: [decision-model, calibrated, structured-output, system-one, one-pass]
7
+ ---
8
+
9
+ # decider-12b: typed decisions with calibrated probabilities in one forward pass, on Gemma-4-12B-it
10
+
11
+ decider-12b is [google/gemma-4-12B-it](https://huggingface.co/google/gemma-4-12B-it) (revision `707f0a3b`) with its weights
12
+ unchanged, read through the decider readout. It is not trained further.
13
+
14
+ - **Input:** a state and one or more typed questions (Choice, Noul yes/no, Score), each with an explicit option list.
15
+ - **Output:** a probability distribution over the options for every question, from one forward pass. There is no decoding and
16
+ no parsing.
17
+ - **Prompt:** the model's chat template, with thinking off. The option letter is read at the answer slot, with the model's
18
+ final-logit softcapping applied.
19
+ - **Temperatures:** one per answer type, in `decider_config.json`.
20
+
21
+ This repository holds the bf16 weights as published by Google, the chat template, and `decider_config.json`.
22
+
23
+ ## Usage
24
+
25
+ Needs `decider-ai>=1.7.0` (1.7.0 adds Gemma's final-logit softcapping; older versions read this model too sharply).
26
+
27
+ ```python
28
+ from decider.infer import Decider
29
+ d = Decider("Mapika/decider-12b")
30
+ d.system_one(state, {"refund": {"type": "noul", "instructions": "Is the refund allowed under the policy?",
31
+ "criteria": {"true": "allowed", "false": "not allowed"}}})
32
+ ```
33
+
34
+ Serving: `DECIDER_MODEL=Mapika/decider-12b uvicorn decider.serve:app` (POST `/v1/systemone`, the System One wire format).
35
+
36
+ ## Temperatures
37
+
38
+ | type | T | fitted on (our own rows, no JevBench or Decision Index items) |
39
+ |---|---|---|
40
+ | Choice | 4.0 | NLL and top-label ECE over 2,113 Choice rows (human-labelled public sets, held-out generated decision families, teacher-written document questions) |
41
+ | Noul | 1.0 | 685 Noul rows plus the 204-item fit half of a fresh authored yes/no set |
42
+ | Score | 3.5 | 246 Score rows |
43
+
44
+ Choice accuracy does not depend on T, so its temperature is chosen for calibration only:
45
+ - ECE 0.243 at T 1.0, 0.026 at T 4.0.
46
+
47
+ Noul is kept sharp. JevBench v1.5 counts a yes/no answer with P(yes) between 0.2 and 0.8 as wrong:
48
+ - on the fresh set, T 2.0 would lower ECE from 0.140 to 0.096 but cost 16 chance-corrected points.
49
+
50
+ Score is read as the expected level:
51
+ - T 3.5 gives chance-corrected competence 59.9 (60.4 at best, 57.1 at T 1.0), with top-label ECE 0.044 (0.241 at T 1.0).
52
+
53
+ ## Measurements (decider-ai 1.7.0, one B300)
54
+
55
+ | set | decider-12b | decider-4b v2 | decider-35b-a3b |
56
+ |---|---|---|---|
57
+ | JevBench public items, accuracy easy / standard / hard | 1.000 / 0.986 / 0.730 | 1.000 / 0.986 / 0.676 | 1.000 / 0.972 / 0.676 |
58
+ | JevBench public hard tier, top-label ECE | 0.106 | 0.104 | 0.157 |
59
+ | fresh 399-item yes/no set, v1.5 chance-corrected Noul score | 61.2 | 1.0 | 21.0 |
60
+ | the same, argmax accuracy / share of answers between 0.2 and 0.8 | 0.822 / 4 % | 0.717 / 38 % | 0.737 / 21 % |
61
+ | latency per decision, serial, p50 / p95 | 20 ms / 155 ms | 33 ms / 55 ms | |
62
+
63
+ About the fresh yes/no set:
64
+ - 319 items were written and reviewed for this model family's evaluation. 80 are generated state-tracking logs with
65
+ computed answers.
66
+ - The set was not used for anything but the Noul temperature, which was fitted on its fit half.
67
+
68
+ About the JevBench public items:
69
+ - They were read for measurement only.
70
+ - They were not used for training, temperatures or any choice in this repository.
71
+
72
+ ## Limitations
73
+
74
+ - The model is stock Gemma-4-12B-it. Its strengths and failure modes are the base model's. The one change is the readout: one
75
+ forward pass with the probabilities read at a fixed answer slot.
76
+ - Choice probabilities at T 4.0 are calibrated on our rows. Per-dataset calibration still varies (knowledge multiple choice is
77
+ underconfident, hard reasoning overconfident).
78
+ - Weakest measured area: state-tracking logs, with argmax accuracy 0.69 on the fresh set.
79
+
80
+ ## Licence
81
+
82
+ Weights: google/gemma-4-12B-it, Apache-2.0 as published by Google.
83
+ Readout code: Apache-2.0 (github.com/Mapika/decider).
chat_template.jinja ADDED
@@ -0,0 +1,390 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {#
2
+ Template: Google Gemma 4 Canonical Chat Template
3
+ Author: Google Gemma Engineering Team
4
+ Published: 2026-07-09
5
+ Context: Fixed tool-calling loops, turn closures, and thinking content-ordering.
6
+ #}
7
+ {%- macro format_parameters(properties, required, filter_keys=false) -%}
8
+ {%- set standard_keys = ['description', 'type', 'properties', 'required', 'nullable'] -%}
9
+ {%- set ns = namespace(found_first=false) -%}
10
+ {%- for key, value in properties | dictsort -%}
11
+ {%- set add_comma = false -%}
12
+ {%- if not filter_keys or key not in standard_keys -%}
13
+ {%- if ns.found_first %},{% endif -%}
14
+ {%- set ns.found_first = true -%}
15
+ {{ key }}:{
16
+ {%- if value['description'] -%}
17
+ description:<|"|>{{ value['description'] }}<|"|>
18
+ {%- set add_comma = true -%}
19
+ {%- endif -%}
20
+ {%- if value['type'] | upper == 'STRING' -%}
21
+ {%- if value['enum'] -%}
22
+ {%- if add_comma %},{%- else -%} {%- set add_comma = true -%} {% endif -%}
23
+ enum:{{ format_argument(value['enum']) }}
24
+ {%- endif -%}
25
+ {%- elif value['type'] | upper == 'ARRAY' -%}
26
+ {%- if value['items'] is mapping and value['items'] -%}
27
+ {%- if add_comma %},{%- else -%} {%- set add_comma = true -%} {% endif -%}
28
+ items:{
29
+ {%- set ns_items = namespace(found_first=false) -%}
30
+ {%- for item_key, item_value in value['items'] | dictsort -%}
31
+ {%- if item_value is not none -%}
32
+ {%- if ns_items.found_first %},{% endif -%}
33
+ {%- set ns_items.found_first = true -%}
34
+ {%- if item_key == 'properties' -%}
35
+ properties:{
36
+ {%- if item_value is mapping -%}
37
+ {{- format_parameters(item_value, value['items']['required'] | default([])) -}}
38
+ {%- endif -%}
39
+ }
40
+ {%- elif item_key == 'required' -%}
41
+ required:[
42
+ {%- for req_item in item_value -%}
43
+ <|"|>{{- req_item -}}<|"|>
44
+ {%- if not loop.last %},{% endif -%}
45
+ {%- endfor -%}
46
+ ]
47
+ {%- elif item_key == 'type' -%}
48
+ {%- if item_value is string -%}
49
+ type:{{ format_argument(item_value | upper) }}
50
+ {%- else -%}
51
+ type:{{ format_argument(item_value | map('upper') | list) }}
52
+ {%- endif -%}
53
+ {%- else -%}
54
+ {{ item_key }}:{{ format_argument(item_value) }}
55
+ {%- endif -%}
56
+ {%- endif -%}
57
+ {%- endfor -%}
58
+ }
59
+ {%- endif -%}
60
+ {%- endif -%}
61
+ {%- if value['nullable'] %}
62
+ {%- if add_comma %},{%- else -%} {%- set add_comma = true -%} {% endif -%}
63
+ nullable:true
64
+ {%- endif -%}
65
+ {%- if value['type'] | upper == 'OBJECT' -%}
66
+ {%- if value['properties'] is defined and value['properties'] is mapping -%}
67
+ {%- if add_comma %},{%- else -%} {%- set add_comma = true -%} {% endif -%}
68
+ properties:{
69
+ {{- format_parameters(value['properties'], value['required'] | default([])) -}}
70
+ }
71
+ {%- elif value is mapping -%}
72
+ {%- if add_comma %},{%- else -%} {%- set add_comma = true -%} {% endif -%}
73
+ properties:{
74
+ {{- format_parameters(value, value['required'] | default([]), filter_keys=true) -}}
75
+ }
76
+ {%- endif -%}
77
+ {%- if value['required'] -%}
78
+ {%- if add_comma %},{%- else -%} {%- set add_comma = true -%} {% endif -%}
79
+ required:[
80
+ {%- for item in value['required'] | default([]) -%}
81
+ <|"|>{{- item -}}<|"|>
82
+ {%- if not loop.last %},{% endif -%}
83
+ {%- endfor -%}
84
+ ]
85
+ {%- endif -%}
86
+ {%- endif -%}
87
+ {%- if add_comma %},{%- else -%} {%- set add_comma = true -%} {% endif -%}
88
+ type:<|"|>{{ value['type'] | upper }}<|"|>}
89
+ {%- endif -%}
90
+ {%- endfor -%}
91
+ {%- endmacro -%}
92
+ {%- macro format_function_declaration(tool_data) -%}
93
+ declaration:{{- tool_data['function']['name'] -}}{description:<|"|>{{- tool_data['function']['description'] -}}<|"|>
94
+ {%- set params = tool_data['function']['parameters'] -%}
95
+ {%- if params -%}
96
+ ,parameters:{
97
+ {%- if params['properties'] -%}
98
+ properties:{ {{- format_parameters(params['properties'], params['required']) -}} },
99
+ {%- endif -%}
100
+ {%- if params['required'] -%}
101
+ required:[
102
+ {%- for item in params['required'] -%}
103
+ <|"|>{{- item -}}<|"|>
104
+ {{- ',' if not loop.last -}}
105
+ {%- endfor -%}
106
+ ],
107
+ {%- endif -%}
108
+ {%- if params['type'] -%}
109
+ type:<|"|>{{- params['type'] | upper -}}<|"|>}
110
+ {%- endif -%}
111
+ {%- endif -%}
112
+ {%- if 'response' in tool_data['function'] -%}
113
+ {%- set response_declaration = tool_data['function']['response'] -%}
114
+ ,response:{
115
+ {%- if response_declaration['description'] -%}
116
+ description:<|"|>{{- response_declaration['description'] -}}<|"|>,
117
+ {%- endif -%}
118
+ {%- if response_declaration['type'] | upper == 'OBJECT' -%}
119
+ type:<|"|>{{- response_declaration['type'] | upper -}}<|"|>}
120
+ {%- endif -%}
121
+ {%- endif -%}
122
+ }
123
+ {%- endmacro -%}
124
+ {%- macro format_argument(argument, escape_keys=True) -%}
125
+ {%- if argument is none -%}
126
+ {{- 'null' -}}
127
+ {%- elif argument is string -%}
128
+ {{- '<|"|>' + argument + '<|"|>' -}}
129
+ {%- elif argument is boolean -%}
130
+ {{- 'true' if argument else 'false' -}}
131
+ {%- elif argument is mapping -%}
132
+ {{- '{' -}}
133
+ {%- set ns = namespace(found_first=false) -%}
134
+ {%- for key, value in argument | dictsort -%}
135
+ {%- if ns.found_first %},{% endif -%}
136
+ {%- set ns.found_first = true -%}
137
+ {%- if escape_keys -%}
138
+ {{- '<|"|>' + key + '<|"|>' -}}
139
+ {%- else -%}
140
+ {{- key -}}
141
+ {%- endif -%}
142
+ :{{- format_argument(value, escape_keys=escape_keys) -}}
143
+ {%- endfor -%}
144
+ {{- '}' -}}
145
+ {%- elif argument is sequence -%}
146
+ {{- '[' -}}
147
+ {%- for item in argument -%}
148
+ {{- format_argument(item, escape_keys=escape_keys) -}}
149
+ {%- if not loop.last %},{% endif -%}
150
+ {%- endfor -%}
151
+ {{- ']' -}}
152
+ {%- else -%}
153
+ {{- argument -}}
154
+ {%- endif -%}
155
+ {%- endmacro -%}
156
+ {%- macro strip_thinking(text) -%}
157
+ {%- set ns = namespace(result='') -%}
158
+ {%- for part in text.split('<channel|>') -%}
159
+ {%- if '<|channel>' in part -%}
160
+ {%- set ns.result = ns.result + part.split('<|channel>')[0] -%}
161
+ {%- else -%}
162
+ {%- set ns.result = ns.result + part -%}
163
+ {%- endif -%}
164
+ {%- endfor -%}
165
+ {{- ns.result | trim -}}
166
+ {%- endmacro -%}
167
+
168
+ {%- macro format_tool_response_block(tool_name, response) -%}
169
+ {{- '<|tool_response>' -}}
170
+ {%- if response is mapping -%}
171
+ {{- 'response:' + tool_name + '{' -}}
172
+ {%- for key, value in response | dictsort -%}
173
+ {{- key -}}:{{- format_argument(value, escape_keys=False) -}}
174
+ {%- if not loop.last %},{% endif -%}
175
+ {%- endfor -%}
176
+ {{- '}' -}}
177
+ {%- else -%}
178
+ {{- 'response:' + tool_name + '{value:' + format_argument(response, escape_keys=False) + '}' -}}
179
+ {%- endif -%}
180
+ {{- '<tool_response|>' -}}
181
+ {%- endmacro -%}
182
+
183
+ {#- ===== SETUP ===== -#}
184
+ {%- set ns = namespace(prev_message_type=None, prev_non_tool_role=None) -%}
185
+ {%- set loop_messages = messages -%}
186
+ {%- set enable_thinking = enable_thinking | default(false) -%}
187
+ {%- set preserve_thinking = preserve_thinking | default(false) -%}
188
+ {{- bos_token -}}
189
+ {#- Handle System/Tool Definitions Block -#}
190
+ {%- if enable_thinking or tools or (messages and messages[0]['role'] in ['system', 'developer']) -%}
191
+ {{- '<|turn>system\n' -}}
192
+ {#- Inject Thinking token at the very top of the FIRST system turn -#}
193
+ {%- if enable_thinking -%}
194
+ {{- '<|think|>\n' -}}
195
+ {%- set ns.prev_message_type = 'think' -%}
196
+ {%- endif -%}
197
+ {%- if messages and messages[0]['role'] in ['system', 'developer'] -%}
198
+ {%- if messages[0]['content'] is string -%}
199
+ {{- messages[0]['content'] | trim -}}
200
+ {%- elif messages[0]['content'] is sequence -%}
201
+ {%- for item in messages[0]['content'] -%}
202
+ {{- item['text'] | trim + ' '-}}
203
+ {%- endfor -%}
204
+ {%- endif -%}
205
+ {%- set loop_messages = messages[1:] -%}
206
+ {%- endif -%}
207
+ {%- if tools -%}
208
+ {%- for tool in tools %}
209
+ {{- '<|tool>' -}}
210
+ {{- format_function_declaration(tool) | trim -}}
211
+ {{- '<tool|>' -}}
212
+ {%- endfor %}
213
+ {%- set ns.prev_message_type = 'tool' -%}
214
+ {%- endif -%}
215
+ {{- '<turn|>\n' -}}
216
+ {%- endif %}
217
+
218
+ {#- Pre-scan: find last user message index for reasoning guard -#}
219
+ {%- set ns_turn = namespace(last_user_idx=-1) -%}
220
+ {%- for i in range(loop_messages | length) -%}
221
+ {%- if loop_messages[i]['role'] == 'user' -%}
222
+ {%- set ns_turn.last_user_idx = i -%}
223
+ {%- endif -%}
224
+ {%- endfor -%}
225
+
226
+ {#- Loop through messages -#}
227
+ {%- for message in loop_messages -%}
228
+ {%- if message['role'] != 'tool' -%}
229
+ {%- set ns.prev_message_type = None -%}
230
+ {%- set role = 'model' if message['role'] == 'assistant' else message['role'] -%}
231
+ {#- Detect continuation using tracked state — O(1) instead of O(n) backward scan -#}
232
+ {%- set continue_same_model_turn = (role == 'model' and ns.prev_non_tool_role == 'assistant') -%}
233
+ {%- if not continue_same_model_turn -%}
234
+ {{- '<|turn>' + role + '\n' }}
235
+
236
+ {%- endif -%}
237
+
238
+ {#- Render reasoning/reasoning_content as thinking channel -#}
239
+ {%- set thinking_text = message.get('reasoning') or message.get('reasoning_content') -%}
240
+ {%- set thinking_gate = (loop.index0 > ns_turn.last_user_idx) or (preserve_thinking and message.get('tool_calls')) -%}
241
+ {%- if thinking_text and thinking_gate -%}
242
+ {{- '<|channel>thought\n' + thinking_text + '\n<channel|>' -}}
243
+ {%- endif -%}
244
+
245
+ {%- if message.get('tool_calls') -%}
246
+ {%- for tool_call in message.get('tool_calls') -%}
247
+ {%- set function = tool_call['function'] -%}
248
+ {{- '<|tool_call>call:' + function['name'] + '{' -}}
249
+ {%- if function['arguments'] is mapping -%}
250
+ {%- set ns_args = namespace(found_first=false) -%}
251
+ {%- for key, value in function['arguments'] | dictsort -%}
252
+ {%- if ns_args.found_first %},{% endif -%}
253
+ {%- set ns_args.found_first = true -%}
254
+ {{- key -}}:{{- format_argument(value, escape_keys=False) -}}
255
+ {%- endfor -%}
256
+ {%- elif function['arguments'] is none -%}
257
+ {%- else -%}
258
+ {{- raise_exception(
259
+ "chat_template: tool_calls[].function.arguments must be a "
260
+ "JSON object (mapping), not a string. Deserialize arguments "
261
+ "before passing to the template."
262
+ ) -}}
263
+ {%- endif -%}
264
+ {{- '}<tool_call|>' -}}
265
+ {%- endfor -%}
266
+ {%- set ns.prev_message_type = 'tool_call' -%}
267
+ {%- endif -%}
268
+
269
+ {%- set ns_tr_out = namespace(flag=false) -%}
270
+ {%- if message.get('tool_responses') -%}
271
+ {#- Legacy: tool_responses embedded on the assistant message (Google/Gemma native) -#}
272
+ {%- for tool_response in message.get('tool_responses') -%}
273
+ {{- format_tool_response_block(tool_response['name'] | default('unknown', true), tool_response['response']) -}}
274
+ {%- set ns_tr_out.flag = true -%}
275
+ {%- set ns.prev_message_type = 'tool_response' -%}
276
+ {%- endfor -%}
277
+ {%- elif message.get('tool_calls') -%}
278
+ {#- OpenAI Chat Completions: forward-scan consecutive role:tool messages -#}
279
+ {%- set ns_tool_scan = namespace(stopped=false) -%}
280
+ {%- for k in range(loop.index0 + 1, loop_messages | length) -%}
281
+ {%- if ns_tool_scan.stopped -%}
282
+ {%- elif loop_messages[k]['role'] != 'tool' -%}
283
+ {%- set ns_tool_scan.stopped = true -%}
284
+ {%- else -%}
285
+ {%- set follow = loop_messages[k] -%}
286
+ {#- Resolve tool_call_id to function name -#}
287
+ {%- set ns_tname = namespace(name=follow.get('name') or 'unknown') -%}
288
+ {%- for tc in message.get('tool_calls') -%}
289
+ {%- if tc.get('id') == follow.get('tool_call_id') -%}
290
+ {%- set ns_tname.name = tc['function']['name'] -%}
291
+ {%- endif -%}
292
+ {%- endfor -%}
293
+ {#- Handle content as string or content-parts array -#}
294
+ {%- set tool_body = follow.get('content') -%}
295
+ {%- if tool_body is string -%}
296
+ {{- format_tool_response_block(ns_tname.name, tool_body) -}}
297
+ {%- elif tool_body is sequence and tool_body is not string -%}
298
+ {%- set ns_txt = namespace(s='') -%}
299
+ {%- for part in tool_body -%}
300
+ {%- if part.get('type') == 'text' -%}
301
+ {%- set ns_txt.s = ns_txt.s + (part.get('text') | default('')) -%}
302
+ {%- endif -%}
303
+ {%- endfor -%}
304
+ {{- format_tool_response_block(ns_tname.name, ns_txt.s) -}}
305
+ {%- for part in tool_body -%}
306
+ {%- if part.get('type') in ['image', 'image_url'] -%}
307
+ {{- '<|image|>' -}}
308
+ {%- elif part.get('type') in ['audio', 'input_audio'] -%}
309
+ {{- '<|audio|>' -}}
310
+ {%- elif part.get('type') == 'video' -%}
311
+ {{- '<|video|>' -}}
312
+ {%- endif -%}
313
+ {%- endfor -%}
314
+ {%- else -%}
315
+ {{- format_tool_response_block(ns_tname.name, tool_body) -}}
316
+ {%- endif -%}
317
+ {%- set ns_tr_out.flag = true -%}
318
+ {%- set ns.prev_message_type = 'tool_response' -%}
319
+ {%- endif -%}
320
+ {%- endfor -%}
321
+ {%- endif -%}
322
+
323
+ {%- set captured_content -%}
324
+ {%- if message.get('content') is string -%}
325
+ {%- if role == 'model' -%}
326
+ {{- strip_thinking(message['content']) -}}
327
+ {%- else -%}
328
+ {{- message['content'] | trim -}}
329
+ {%- endif -%}
330
+ {%- elif message.get('content') is sequence -%}
331
+ {%- for item in message['content'] -%}
332
+ {%- if item.get('type') == 'text' -%}
333
+ {%- if role == 'model' -%}
334
+ {{- strip_thinking(item['text']) -}}
335
+ {%- else -%}
336
+ {{- item['text'] | trim -}}
337
+ {%- endif -%}
338
+ {%- elif item.get('type') in ['image', 'image_url'] -%}
339
+ {{- '<|image|>' -}}
340
+ {%- elif item.get('type') in ['audio', 'input_audio'] -%}
341
+ {{- '<|audio|>' -}}
342
+ {%- elif item.get('type') == 'video' -%}
343
+ {{- '<|video|>' -}}
344
+ {%- endif -%}
345
+ {%- endfor -%}
346
+ {%- endif -%}
347
+ {%- endset -%}
348
+
349
+ {{- captured_content -}}
350
+ {%- set has_content = captured_content | trim | length > 0 -%}
351
+
352
+ {#- Forward-scan: find next non-tool message role for continuation detection -#}
353
+ {%- set next_nt = namespace(role=None, found=false) -%}
354
+ {%- for j in range(loop.index0 + 1, loop_messages | length) -%}
355
+ {%- if not next_nt.found -%}
356
+ {%- if loop_messages[j]['role'] != 'tool' -%}
357
+ {%- set next_nt.role = loop_messages[j]['role'] -%}
358
+ {%- set next_nt.found = true -%}
359
+ {%- endif -%}
360
+ {%- endif -%}
361
+ {%- endfor -%}
362
+
363
+ {%- set continues_into_next = (
364
+ role == 'model'
365
+ and next_nt.role == 'assistant'
366
+ and (not message.get('tool_calls') or ns_tr_out.flag)
367
+ ) -%}
368
+
369
+ {%- if ns.prev_message_type == 'tool_call' and not ns_tr_out.flag -%}
370
+ {{- '<|tool_response>' -}}
371
+ {%- elif continues_into_next -%}
372
+ {%- elif not (ns_tr_out.flag and not has_content and not next_nt.found) -%}
373
+ {{- '<turn|>\n' -}}
374
+ {%- endif -%}
375
+
376
+ {#- Track previous non-tool role for next iteration (avoids O(n) backward scan) -#}
377
+ {%- set ns.prev_non_tool_role = message['role'] -%}
378
+ {%- endif -%}
379
+ {%- endfor -%}
380
+
381
+ {%- if add_generation_prompt -%}
382
+ {%- if ns.prev_message_type != 'tool_response' and ns.prev_message_type != 'tool_call' -%}
383
+ {{- '<|turn>model\n' -}}
384
+ {%- if not enable_thinking -%}
385
+ {{- '<|channel>thought\n<channel|>' -}}
386
+ {%- endif -%}
387
+ {%- elif ns.prev_message_type == 'tool_response' and enable_thinking -%}
388
+ {{- '<|channel>thought\n' -}}
389
+ {%- endif -%}
390
+ {%- endif -%}
config.json ADDED
@@ -0,0 +1,172 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "Gemma4UnifiedForConditionalGeneration"
4
+ ],
5
+ "audio_config": {
6
+ "_name_or_path": "",
7
+ "architectures": null,
8
+ "audio_embed_dim": 640,
9
+ "audio_samples_per_token": 640,
10
+ "chunk_size_feed_forward": 0,
11
+ "dtype": null,
12
+ "hidden_size": 640,
13
+ "id2label": {
14
+ "0": "LABEL_0",
15
+ "1": "LABEL_1"
16
+ },
17
+ "initializer_range": 0.02,
18
+ "is_encoder_decoder": false,
19
+ "label2id": {
20
+ "LABEL_0": 0,
21
+ "LABEL_1": 1
22
+ },
23
+ "model_type": "gemma4_unified_audio",
24
+ "output_attentions": false,
25
+ "output_hidden_states": false,
26
+ "output_proj_dims": 640,
27
+ "problem_type": null,
28
+ "return_dict": true,
29
+ "rms_norm_eps": 1e-06
30
+ },
31
+ "audio_token_id": 258881,
32
+ "boa_token_id": 256000,
33
+ "boi_token_id": 255999,
34
+ "dtype": "bfloat16",
35
+ "eoa_token_index": 258883,
36
+ "eoi_token_id": 258882,
37
+ "eos_token_id": [
38
+ 1,
39
+ 106
40
+ ],
41
+ "image_token_id": 258880,
42
+ "initializer_range": 0.02,
43
+ "model_type": "gemma4_unified",
44
+ "text_config": {
45
+ "attention_bias": false,
46
+ "attention_dropout": 0.0,
47
+ "attention_k_eq_v": true,
48
+ "bos_token_id": 2,
49
+ "enable_moe_block": false,
50
+ "eos_token_id": 1,
51
+ "final_logit_softcapping": 30.0,
52
+ "global_head_dim": 512,
53
+ "head_dim": 256,
54
+ "hidden_activation": "gelu_pytorch_tanh",
55
+ "hidden_size": 3840,
56
+ "hidden_size_per_layer_input": 0,
57
+ "initializer_range": 0.02,
58
+ "intermediate_size": 15360,
59
+ "layer_types": [
60
+ "sliding_attention",
61
+ "sliding_attention",
62
+ "sliding_attention",
63
+ "sliding_attention",
64
+ "sliding_attention",
65
+ "full_attention",
66
+ "sliding_attention",
67
+ "sliding_attention",
68
+ "sliding_attention",
69
+ "sliding_attention",
70
+ "sliding_attention",
71
+ "full_attention",
72
+ "sliding_attention",
73
+ "sliding_attention",
74
+ "sliding_attention",
75
+ "sliding_attention",
76
+ "sliding_attention",
77
+ "full_attention",
78
+ "sliding_attention",
79
+ "sliding_attention",
80
+ "sliding_attention",
81
+ "sliding_attention",
82
+ "sliding_attention",
83
+ "full_attention",
84
+ "sliding_attention",
85
+ "sliding_attention",
86
+ "sliding_attention",
87
+ "sliding_attention",
88
+ "sliding_attention",
89
+ "full_attention",
90
+ "sliding_attention",
91
+ "sliding_attention",
92
+ "sliding_attention",
93
+ "sliding_attention",
94
+ "sliding_attention",
95
+ "full_attention",
96
+ "sliding_attention",
97
+ "sliding_attention",
98
+ "sliding_attention",
99
+ "sliding_attention",
100
+ "sliding_attention",
101
+ "full_attention",
102
+ "sliding_attention",
103
+ "sliding_attention",
104
+ "sliding_attention",
105
+ "sliding_attention",
106
+ "sliding_attention",
107
+ "full_attention"
108
+ ],
109
+ "max_position_embeddings": 262144,
110
+ "model_type": "gemma4_unified_text",
111
+ "moe_intermediate_size": null,
112
+ "num_attention_heads": 16,
113
+ "num_experts": null,
114
+ "num_global_key_value_heads": 1,
115
+ "num_hidden_layers": 48,
116
+ "num_key_value_heads": 8,
117
+ "num_kv_shared_layers": 0,
118
+ "pad_token_id": 0,
119
+ "rms_norm_eps": 1e-06,
120
+ "rope_parameters": {
121
+ "full_attention": {
122
+ "partial_rotary_factor": 0.25,
123
+ "rope_theta": 1000000.0,
124
+ "rope_type": "proportional"
125
+ },
126
+ "sliding_attention": {
127
+ "rope_theta": 10000.0,
128
+ "rope_type": "default"
129
+ }
130
+ },
131
+ "sliding_window": 1024,
132
+ "tie_word_embeddings": true,
133
+ "top_k_experts": null,
134
+ "use_bidirectional_attention": "vision",
135
+ "use_cache": true,
136
+ "use_double_wide_mlp": false,
137
+ "vocab_size": 262144,
138
+ "vocab_size_per_layer_input": 262144
139
+ },
140
+ "tie_word_embeddings": true,
141
+ "transformers_version": "5.10.0.dev0",
142
+ "video_token_id": 258884,
143
+ "vision_config": {
144
+ "_name_or_path": "",
145
+ "architectures": null,
146
+ "chunk_size_feed_forward": 0,
147
+ "dtype": null,
148
+ "id2label": {
149
+ "0": "LABEL_0",
150
+ "1": "LABEL_1"
151
+ },
152
+ "initializer_range": 0.02,
153
+ "is_encoder_decoder": false,
154
+ "label2id": {
155
+ "LABEL_0": 0,
156
+ "LABEL_1": 1
157
+ },
158
+ "mm_embed_dim": 3840,
159
+ "mm_posemb_size": 1120,
160
+ "model_patch_size": 48,
161
+ "model_type": "gemma4_unified_vision",
162
+ "num_soft_tokens": 280,
163
+ "output_attentions": false,
164
+ "output_hidden_states": false,
165
+ "output_proj_dims": 3840,
166
+ "patch_size": 16,
167
+ "pooling_kernel_size": 3,
168
+ "problem_type": null,
169
+ "return_dict": true,
170
+ "rms_norm_eps": 1e-06
171
+ }
172
+ }
decider_config.json ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "version": "12b-v1",
3
+ "layout": "chat",
4
+ "temperature": 4.0,
5
+ "temperature_by_type": {"choice": 4.0, "noul": 1.0, "score": 3.5},
6
+ "neutralize_none": false,
7
+ "isolated_levels": false,
8
+ "base": "google/gemma-4-12B-it (revision 707f0a3b), unmodified weights",
9
+ "requires": "decider-ai>=1.7.0 (final-logit softcapping)"
10
+ }
generation_config.json ADDED
@@ -0,0 +1,18 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token_id": 2,
3
+ "do_sample": true,
4
+ "eos_token_id": [
5
+ 1,
6
+ 106,
7
+ 50
8
+ ],
9
+ "pad_token_id": 0,
10
+ "suppress_tokens": [
11
+ 258883,
12
+ 258882
13
+ ],
14
+ "temperature": 1.0,
15
+ "top_k": 64,
16
+ "top_p": 0.95,
17
+ "transformers_version": "5.10.0.dev0"
18
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5a84cb313260ac447237b890387116dfa8682e49a6b44bc585ae8353abbff18d
3
+ size 23919549408
processor_config.json ADDED
@@ -0,0 +1,62 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "audio_ms_per_token": 40,
3
+ "audio_seq_length": 750,
4
+ "feature_extractor": {
5
+ "audio_samples_per_token": 640,
6
+ "feature_extractor_type": "Gemma4UnifiedAudioFeatureExtractor",
7
+ "feature_size": 640,
8
+ "padding_side": "right",
9
+ "padding_value": 0.0,
10
+ "return_attention_mask": true,
11
+ "sampling_rate": 16000
12
+ },
13
+ "image_processor": {
14
+ "do_convert_rgb": true,
15
+ "do_normalize": false,
16
+ "do_rescale": true,
17
+ "do_resize": true,
18
+ "image_mean": [
19
+ 0.0,
20
+ 0.0,
21
+ 0.0
22
+ ],
23
+ "image_processor_type": "Gemma4UnifiedImageProcessor",
24
+ "image_std": [
25
+ 1.0,
26
+ 1.0,
27
+ 1.0
28
+ ],
29
+ "max_soft_tokens": 280,
30
+ "patch_size": 16,
31
+ "pooling_kernel_size": 3,
32
+ "resample": 3,
33
+ "rescale_factor": 0.00392156862745098
34
+ },
35
+ "image_seq_length": 280,
36
+ "processor_class": "Gemma4UnifiedProcessor",
37
+ "video_processor": {
38
+ "do_convert_rgb": true,
39
+ "do_normalize": true,
40
+ "do_rescale": true,
41
+ "do_resize": true,
42
+ "do_sample_frames": true,
43
+ "image_mean": [
44
+ 0.0,
45
+ 0.0,
46
+ 0.0
47
+ ],
48
+ "image_std": [
49
+ 1.0,
50
+ 1.0,
51
+ 1.0
52
+ ],
53
+ "max_soft_tokens": 70,
54
+ "num_frames": 32,
55
+ "patch_size": 16,
56
+ "pooling_kernel_size": 3,
57
+ "resample": 3,
58
+ "rescale_factor": 0.00392156862745098,
59
+ "return_metadata": false,
60
+ "video_processor_type": "Gemma4UnifiedVideoProcessor"
61
+ }
62
+ }
tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cc8d3a0ce36466ccc1278bf987df5f71db1719b9ca6b4118264f45cb627bfe0f
3
+ size 32169626
tokenizer_config.json ADDED
@@ -0,0 +1,120 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "audio_token": "<|audio|>",
3
+ "backend": "tokenizers",
4
+ "boa_token": "<|audio>",
5
+ "boi_token": "<|image>",
6
+ "bos_token": "<bos>",
7
+ "eoa_token": "<audio|>",
8
+ "eoc_token": "<channel|>",
9
+ "eoi_token": "<image|>",
10
+ "eos_token": "<eos>",
11
+ "eot_token": "<turn|>",
12
+ "escape_token": "<|\"|>",
13
+ "etc_token": "<tool_call|>",
14
+ "etd_token": "<tool|>",
15
+ "etr_token": "<tool_response|>",
16
+ "extra_special_tokens": [
17
+ "<|video|>"
18
+ ],
19
+ "image_token": "<|image|>",
20
+ "mask_token": "<mask>",
21
+ "model_max_length": 1000000000000000019884624838656,
22
+ "pad_token": "<pad>",
23
+ "padding_side": "left",
24
+ "processor_class": "Gemma4UnifiedProcessor",
25
+ "response_schema": {
26
+ "properties": {
27
+ "content": {
28
+ "type": "string"
29
+ },
30
+ "role": {
31
+ "const": "assistant"
32
+ },
33
+ "thinking": {
34
+ "type": "string"
35
+ },
36
+ "tool_calls": {
37
+ "items": {
38
+ "properties": {
39
+ "function": {
40
+ "properties": {
41
+ "arguments": {
42
+ "additionalProperties": {},
43
+ "type": "object",
44
+ "x-parser": "gemma4-tool-call"
45
+ },
46
+ "name": {
47
+ "type": "string"
48
+ }
49
+ },
50
+ "type": "object",
51
+ "x-regex": "call\\:(?P<name>\\w+)(?P<arguments>\\{.*\\})"
52
+ },
53
+ "type": {
54
+ "const": "function"
55
+ }
56
+ },
57
+ "type": "object"
58
+ },
59
+ "type": "array",
60
+ "x-regex-iterator": "<\\|tool_call>(.*?)<tool_call\\|>"
61
+ }
62
+ },
63
+ "type": "object",
64
+ "x-regex": "(\\<\\|channel\\>thought\\n(?P<thinking>.*?)\\<channel\\|\\>)?(?P<tool_calls>\\<\\|tool_call\\>.*\\<tool_call\\|\\>)?(?P<content>(?:(?!\\<turn\\|\\>)(?!\\<\\|tool_response\\>).)+)?(?:\\<turn\\|\\>|\\<\\|tool_response\\>)?"
65
+ },
66
+ "response_template": {
67
+ "defaults": {
68
+ "role": "assistant"
69
+ },
70
+ "fields": {
71
+ "content": {
72
+ "close": [
73
+ "<turn|>",
74
+ "<|tool_response>",
75
+ "<eos>"
76
+ ],
77
+ "content": "text"
78
+ },
79
+ "thinking": {
80
+ "close": "<channel|>",
81
+ "content": "text",
82
+ "open": "<|channel>thought\n"
83
+ },
84
+ "tool_calls": {
85
+ "close": "<tool_call|>",
86
+ "content": "json",
87
+ "content_args": {
88
+ "string_delims": [
89
+ [
90
+ "<|\"|>",
91
+ "<|\"|>"
92
+ ]
93
+ ],
94
+ "unquoted_keys": true
95
+ },
96
+ "open_pattern": "<\\|tool_call>call:(?P<name>\\w+)",
97
+ "repeats": true,
98
+ "transform": {
99
+ "function": {
100
+ "arguments": "{content}",
101
+ "name": "{name}"
102
+ },
103
+ "type": "function"
104
+ }
105
+ }
106
+ },
107
+ "start_anchor": [
108
+ "<|turn>model\n",
109
+ "<tool_response|>"
110
+ ]
111
+ },
112
+ "soc_token": "<|channel>",
113
+ "sot_token": "<|turn>",
114
+ "stc_token": "<|tool_call>",
115
+ "std_token": "<|tool>",
116
+ "str_token": "<|tool_response>",
117
+ "think_token": "<|think|>",
118
+ "tokenizer_class": "GemmaTokenizer",
119
+ "unk_token": "<unk>"
120
+ }