Shiftedx commited on
Commit
e560b5f
·
verified ·
1 Parent(s): 2a40954

Add files using upload-large-folder tool

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 +1,81 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ base_model: Jackrong/Qwopus3.6-35B-A3B-Coder
4
+ base_model_relation: quantized
5
+ library_name: mlx
6
+ pipeline_tag: image-text-to-text
7
+ tags:
8
+ - mlx
9
+ - mlx-vlm
10
+ - mxfp4
11
+ - safetensors
12
+ - lm-studio
13
+ - apple-silicon
14
+ - qwen3_5_moe
15
+ - qwen3_6
16
+ - coder
17
+ - agent
18
+ - tool-use
19
+ - function-calling
20
+ - vision
21
+ - thinking-off
22
+ - abliterated
23
+ ---
24
+
25
+ # qwopus3.6-35b-a3b-coder-abliterated-mxfp4-vision-mlx
26
+
27
+ Vision-enabled MLX MXFP4 conversion of [Jackrong/Qwopus3.6-35B-A3B-Coder](https://huggingface.co/Jackrong/Qwopus3.6-35B-A3B-Coder), prepared by Shiftedx for Apple Silicon, MLX, and LM Studio.
28
+
29
+ ## Build Notes
30
+
31
+ - Quantized primary language weights with `mxfp4` at group size 32.
32
+ - Kept MoE router and gate modules in affine 8-bit group size 64 for compatibility.
33
+ - Added Qwen3.5/Qwen3.6-MoE-compatible vision components and validated image grounding locally.
34
+ - Removed source MTP tensors and set MTP/next-token prediction layer counts to 0 for LM Studio compatibility.
35
+ - Set `tool_parser_type` to `qwen3_coder`.
36
+ - Patched the chat template so `enable_thinking` defaults to false when the runtime honors that template variable.
37
+ - Applied a research refusal-direction weight edit using residual-direction orthogonalization.
38
+
39
+ ## Local Validation
40
+
41
+ Validated locally on June 30, 2026 and July 1, 2026 with LM Studio and direct MLX/VLM loading.
42
+
43
+ | Check | Result |
44
+ |---|---|
45
+ | LM Studio load | Passed at 32k context, parallel 1, GPU max. |
46
+ | Basic text completion | Passed; answered `2+2` with `4` and stopped. |
47
+ | Code completion | Passed; produced a simple valid `add(a, b)` function. |
48
+ | Direct MLX/VLM image color smoke | Passed; answered `Red`. |
49
+ | Direct MLX/VLM OCR smoke | Passed; answered `FABLE 42`. |
50
+ | LM Studio OpenAI-compatible image smoke | Passed; answered `Red` and `FABLE 42`. |
51
+ | LM Studio native image smoke | Passed; answered `Red` and `FABLE 42`. |
52
+ | Thinking-off behavior | Smoke checks returned `0` reasoning tokens. |
53
+ | LM Studio logs | No warnings, errors, tracebacks, KV-cache issues, or tokenizer-regex warnings in the validation window. |
54
+
55
+ This is a smoke-validated release, not a full benchmark suite. Broader downstream evaluation is still recommended for production use.
56
+
57
+ ## Recommended LM Studio Defaults
58
+
59
+ After downloading in LM Studio, load the model by repo name:
60
+
61
+ ```bash
62
+ lms load shiftedx/qwopus3.6-35b-a3b-coder-abliterated-mxfp4-vision-mlx --context-length 32768 --parallel 1 --gpu max
63
+ ```
64
+
65
+ Recommended profile defaults:
66
+
67
+ - Preset/template: Qwen3 thinking-compatible Jinja template with `<|im_end|>` stop.
68
+ - Thinking: off by default through the included chat template.
69
+ - Context length: `200000` when memory allows; `32768` was used for local validation.
70
+ - Sampling: temperature `0.6`, top-k `20`, top-p `0.95`, min-p enabled at `0`.
71
+ - Repeat penalty: off by default.
72
+ - Load: parallel `1`, GPU `max`.
73
+
74
+ ## Provenance
75
+
76
+ - Publisher: Shiftedx
77
+ - Source model: https://huggingface.co/Jackrong/Qwopus3.6-35B-A3B-Coder
78
+ - Source license: Apache-2.0
79
+ - Quantization date: June 29, 2026
80
+ - Vision validation completed: July 1, 2026
81
+ - No MTP or draft-model tensors are included.
chat_template.jinja ADDED
@@ -0,0 +1,157 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {%- set image_count = namespace(value=0) %}
2
+ {%- set video_count = namespace(value=0) %}
3
+ {%- macro render_content(content, do_vision_count, is_system_content=false) %}
4
+ {%- if content is string %}
5
+ {{- content }}
6
+ {%- elif content is iterable and content is not mapping %}
7
+ {%- for item in content %}
8
+ {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
9
+ {%- if is_system_content %}
10
+ {{- raise_exception('System message cannot contain images.') }}
11
+ {%- endif %}
12
+ {%- if do_vision_count %}
13
+ {%- set image_count.value = image_count.value + 1 %}
14
+ {%- endif %}
15
+ {%- if add_vision_id %}
16
+ {{- 'Picture ' ~ image_count.value ~ ': ' }}
17
+ {%- endif %}
18
+ {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
19
+ {%- elif 'video' in item or item.type == 'video' %}
20
+ {%- if is_system_content %}
21
+ {{- raise_exception('System message cannot contain videos.') }}
22
+ {%- endif %}
23
+ {%- if do_vision_count %}
24
+ {%- set video_count.value = video_count.value + 1 %}
25
+ {%- endif %}
26
+ {%- if add_vision_id %}
27
+ {{- 'Video ' ~ video_count.value ~ ': ' }}
28
+ {%- endif %}
29
+ {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
30
+ {%- elif 'text' in item %}
31
+ {{- item.text }}
32
+ {%- else %}
33
+ {{- raise_exception('Unexpected item type in content.') }}
34
+ {%- endif %}
35
+ {%- endfor %}
36
+ {%- elif content is none or content is undefined %}
37
+ {{- '' }}
38
+ {%- else %}
39
+ {{- raise_exception('Unexpected content type.') }}
40
+ {%- endif %}
41
+ {%- endmacro %}
42
+ {%- if messages %}
43
+ {%- if tools and tools is iterable and tools is not mapping %}
44
+ {{- '<|im_start|>system\n' }}
45
+ {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
46
+ {%- for tool in tools %}
47
+ {{- "\n" }}
48
+ {{- tool | tojson }}
49
+ {%- endfor %}
50
+ {{- "\n</tools>" }}
51
+ {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
52
+ {%- if messages[0].role == 'system' %}
53
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
54
+ {%- if content %}
55
+ {{- '\n\n' + content }}
56
+ {%- endif %}
57
+ {%- endif %}
58
+ {{- '<|im_end|>\n' }}
59
+ {%- else %}
60
+ {%- if messages[0].role == 'system' %}
61
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
62
+ {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
63
+ {%- endif %}
64
+ {%- endif %}
65
+ {%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
66
+ {%- for message in messages[::-1] %}
67
+ {%- set index = (messages|length - 1) - loop.index0 %}
68
+ {%- if ns.multi_step_tool and message.role == "user" %}
69
+ {%- set content = render_content(message.content, false)|trim %}
70
+ {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
71
+ {%- set ns.multi_step_tool = false %}
72
+ {%- set ns.last_query_index = index %}
73
+ {%- endif %}
74
+ {%- endif %}
75
+ {%- endfor %}
76
+ {%- for message in messages %}
77
+ {%- set content = render_content(message.content, true)|trim %}
78
+ {%- if message.role == "system" %}
79
+ {%- if not loop.first %}
80
+ {{- raise_exception('System message must be at the beginning.') }}
81
+ {%- endif %}
82
+ {%- elif message.role == "user" %}
83
+ {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
84
+ {%- elif message.role == "assistant" %}
85
+ {%- set reasoning_content = '' %}
86
+ {%- if message.reasoning_content is string %}
87
+ {%- set reasoning_content = message.reasoning_content %}
88
+ {%- else %}
89
+ {%- if '</think>' in content %}
90
+ {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
91
+ {%- set content = content.split('</think>')[-1].lstrip('\n') %}
92
+ {%- endif %}
93
+ {%- endif %}
94
+ {%- set reasoning_content = reasoning_content|trim %}
95
+ {%- if (preserve_thinking is defined and preserve_thinking is true) or (loop.index0 > ns.last_query_index) %}
96
+ {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
97
+ {%- else %}
98
+ {{- '<|im_start|>' + message.role + '\n' + content }}
99
+ {%- endif %}
100
+ {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
101
+ {%- for tool_call in message.tool_calls %}
102
+ {%- if tool_call.function is defined %}
103
+ {%- set tool_call = tool_call.function %}
104
+ {%- endif %}
105
+ {%- if loop.first %}
106
+ {%- if content|trim %}
107
+ {{- '\n\n' }}
108
+ {%- endif %}
109
+ {%- else %}
110
+ {{- '\n' }}
111
+ {%- endif %}
112
+ {%- if tool_call.arguments is mapping %}
113
+ {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
114
+ {%- for args_name, args_value in tool_call.arguments|items %}
115
+ {{- '<parameter=' + args_name + '>\n' }}
116
+ {%- set args_value = args_value | string if args_value is string else args_value | tojson | safe %}
117
+ {{- args_value }}
118
+ {{- '\n</parameter>\n' }}
119
+ {%- endfor %}
120
+ {{- '</function>\n</tool_call>' }}
121
+ {%- else %}
122
+ {{- '<tool_call>\n{"name": ' + tool_call.name|tojson + ', "arguments": ' }}
123
+ {%- if tool_call.arguments is string %}
124
+ {{- tool_call.arguments }}
125
+ {%- else %}
126
+ {{- tool_call.arguments | tojson | safe }}
127
+ {%- endif %}
128
+ {{- '}\n</tool_call>' }}
129
+ {%- endif %}
130
+ {%- endfor %}
131
+ {%- endif %}
132
+ {{- '<|im_end|>\n' }}
133
+ {%- elif message.role == "tool" %}
134
+ {%- if loop.previtem and loop.previtem.role != "tool" %}
135
+ {{- '<|im_start|>user' }}
136
+ {%- endif %}
137
+ {{- '\n<tool_response>\n' }}
138
+ {{- content }}
139
+ {{- '\n</tool_response>' }}
140
+ {%- if not loop.last and loop.nextitem.role != "tool" %}
141
+ {{- '<|im_end|>\n' }}
142
+ {%- elif loop.last %}
143
+ {{- '<|im_end|>\n' }}
144
+ {%- endif %}
145
+ {%- else %}
146
+ {{- raise_exception('Unexpected message role.') }}
147
+ {%- endif %}
148
+ {%- endfor %}
149
+ {%- if add_generation_prompt %}
150
+ {{- '<|im_start|>assistant\n' }}
151
+ {%- if enable_thinking is defined and enable_thinking is true %}
152
+ {{- '<think>\n' }}
153
+ {%- else %}
154
+ {{- '<think>\n\n</think>\n\n' }}
155
+ {%- endif %}
156
+ {%- endif %}
157
+ {%- endif %}
config.json ADDED
@@ -0,0 +1,1910 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "Qwen3_5MoeForConditionalGeneration"
4
+ ],
5
+ "torch_dtype": "bfloat16",
6
+ "image_token_id": 248056,
7
+ "model_name": "/root/.cache/kagglehub/models/jackjirong2/qwen3.6-35b-a3b/transformers/default/1",
8
+ "model_type": "qwen3_5_moe",
9
+ "pad_token_id": 248044,
10
+ "text_config": {
11
+ "attention_bias": false,
12
+ "attention_dropout": 0.0,
13
+ "attn_output_gate": true,
14
+ "bos_token_id": 248044,
15
+ "torch_dtype": "bfloat16",
16
+ "eos_token_id": 248044,
17
+ "full_attention_interval": 4,
18
+ "head_dim": 256,
19
+ "hidden_act": "silu",
20
+ "hidden_size": 2048,
21
+ "initializer_range": 0.02,
22
+ "layer_types": [
23
+ "linear_attention",
24
+ "linear_attention",
25
+ "linear_attention",
26
+ "full_attention",
27
+ "linear_attention",
28
+ "linear_attention",
29
+ "linear_attention",
30
+ "full_attention",
31
+ "linear_attention",
32
+ "linear_attention",
33
+ "linear_attention",
34
+ "full_attention",
35
+ "linear_attention",
36
+ "linear_attention",
37
+ "linear_attention",
38
+ "full_attention",
39
+ "linear_attention",
40
+ "linear_attention",
41
+ "linear_attention",
42
+ "full_attention",
43
+ "linear_attention",
44
+ "linear_attention",
45
+ "linear_attention",
46
+ "full_attention",
47
+ "linear_attention",
48
+ "linear_attention",
49
+ "linear_attention",
50
+ "full_attention",
51
+ "linear_attention",
52
+ "linear_attention",
53
+ "linear_attention",
54
+ "full_attention",
55
+ "linear_attention",
56
+ "linear_attention",
57
+ "linear_attention",
58
+ "full_attention",
59
+ "linear_attention",
60
+ "linear_attention",
61
+ "linear_attention",
62
+ "full_attention"
63
+ ],
64
+ "linear_conv_kernel_dim": 4,
65
+ "linear_key_head_dim": 128,
66
+ "linear_num_key_heads": 16,
67
+ "linear_num_value_heads": 32,
68
+ "linear_value_head_dim": 128,
69
+ "mamba_ssm_dtype": "float32",
70
+ "max_position_embeddings": 262144,
71
+ "model_type": "qwen3_5_moe_text",
72
+ "moe_intermediate_size": 512,
73
+ "mtp_num_hidden_layers": 0,
74
+ "mtp_use_dedicated_embeddings": false,
75
+ "num_attention_heads": 16,
76
+ "num_experts": 256,
77
+ "num_experts_per_tok": 8,
78
+ "num_hidden_layers": 40,
79
+ "num_key_value_heads": 2,
80
+ "output_router_logits": false,
81
+ "pad_token_id": null,
82
+ "partial_rotary_factor": 0.25,
83
+ "rms_norm_eps": 1e-06,
84
+ "rope_parameters": {
85
+ "mrope_interleaved": true,
86
+ "mrope_section": [
87
+ 11,
88
+ 11,
89
+ 10
90
+ ],
91
+ "partial_rotary_factor": 0.25,
92
+ "rope_theta": 10000000,
93
+ "type": "default"
94
+ },
95
+ "router_aux_loss_coef": 0.001,
96
+ "shared_expert_intermediate_size": 512,
97
+ "tie_word_embeddings": false,
98
+ "use_cache": true,
99
+ "vocab_size": 248320,
100
+ "num_nextn_predict_layers": 0
101
+ },
102
+ "tie_word_embeddings": false,
103
+ "unsloth_version": "2026.4.8",
104
+ "video_token_id": 248057,
105
+ "vision_end_token_id": 248054,
106
+ "vision_start_token_id": 248053,
107
+ "tool_parser_type": "qwen3_coder",
108
+ "mtp_num_hidden_layers": 0,
109
+ "num_nextn_predict_layers": 0,
110
+ "quantization": {
111
+ "group_size": 32,
112
+ "bits": 4,
113
+ "mode": "mxfp4",
114
+ "language_model.model.layers.0.mlp.gate": {
115
+ "group_size": 64,
116
+ "bits": 8,
117
+ "mode": "affine"
118
+ },
119
+ "language_model.model.layers.0.mlp.shared_expert_gate": {
120
+ "group_size": 64,
121
+ "bits": 8,
122
+ "mode": "affine"
123
+ },
124
+ "language_model.model.layers.1.mlp.gate": {
125
+ "group_size": 64,
126
+ "bits": 8,
127
+ "mode": "affine"
128
+ },
129
+ "language_model.model.layers.1.mlp.shared_expert_gate": {
130
+ "group_size": 64,
131
+ "bits": 8,
132
+ "mode": "affine"
133
+ },
134
+ "language_model.model.layers.2.mlp.gate": {
135
+ "group_size": 64,
136
+ "bits": 8,
137
+ "mode": "affine"
138
+ },
139
+ "language_model.model.layers.2.mlp.shared_expert_gate": {
140
+ "group_size": 64,
141
+ "bits": 8,
142
+ "mode": "affine"
143
+ },
144
+ "language_model.model.layers.3.mlp.gate": {
145
+ "group_size": 64,
146
+ "bits": 8,
147
+ "mode": "affine"
148
+ },
149
+ "language_model.model.layers.3.mlp.shared_expert_gate": {
150
+ "group_size": 64,
151
+ "bits": 8,
152
+ "mode": "affine"
153
+ },
154
+ "language_model.model.layers.4.mlp.gate": {
155
+ "group_size": 64,
156
+ "bits": 8,
157
+ "mode": "affine"
158
+ },
159
+ "language_model.model.layers.4.mlp.shared_expert_gate": {
160
+ "group_size": 64,
161
+ "bits": 8,
162
+ "mode": "affine"
163
+ },
164
+ "language_model.model.layers.5.mlp.gate": {
165
+ "group_size": 64,
166
+ "bits": 8,
167
+ "mode": "affine"
168
+ },
169
+ "language_model.model.layers.5.mlp.shared_expert_gate": {
170
+ "group_size": 64,
171
+ "bits": 8,
172
+ "mode": "affine"
173
+ },
174
+ "language_model.model.layers.6.mlp.gate": {
175
+ "group_size": 64,
176
+ "bits": 8,
177
+ "mode": "affine"
178
+ },
179
+ "language_model.model.layers.6.mlp.shared_expert_gate": {
180
+ "group_size": 64,
181
+ "bits": 8,
182
+ "mode": "affine"
183
+ },
184
+ "language_model.model.layers.7.mlp.gate": {
185
+ "group_size": 64,
186
+ "bits": 8,
187
+ "mode": "affine"
188
+ },
189
+ "language_model.model.layers.7.mlp.shared_expert_gate": {
190
+ "group_size": 64,
191
+ "bits": 8,
192
+ "mode": "affine"
193
+ },
194
+ "language_model.model.layers.8.mlp.gate": {
195
+ "group_size": 64,
196
+ "bits": 8,
197
+ "mode": "affine"
198
+ },
199
+ "language_model.model.layers.8.mlp.shared_expert_gate": {
200
+ "group_size": 64,
201
+ "bits": 8,
202
+ "mode": "affine"
203
+ },
204
+ "language_model.model.layers.9.mlp.gate": {
205
+ "group_size": 64,
206
+ "bits": 8,
207
+ "mode": "affine"
208
+ },
209
+ "language_model.model.layers.9.mlp.shared_expert_gate": {
210
+ "group_size": 64,
211
+ "bits": 8,
212
+ "mode": "affine"
213
+ },
214
+ "language_model.model.layers.10.mlp.gate": {
215
+ "group_size": 64,
216
+ "bits": 8,
217
+ "mode": "affine"
218
+ },
219
+ "language_model.model.layers.10.mlp.shared_expert_gate": {
220
+ "group_size": 64,
221
+ "bits": 8,
222
+ "mode": "affine"
223
+ },
224
+ "language_model.model.layers.11.mlp.gate": {
225
+ "group_size": 64,
226
+ "bits": 8,
227
+ "mode": "affine"
228
+ },
229
+ "language_model.model.layers.11.mlp.shared_expert_gate": {
230
+ "group_size": 64,
231
+ "bits": 8,
232
+ "mode": "affine"
233
+ },
234
+ "language_model.model.layers.12.mlp.gate": {
235
+ "group_size": 64,
236
+ "bits": 8,
237
+ "mode": "affine"
238
+ },
239
+ "language_model.model.layers.12.mlp.shared_expert_gate": {
240
+ "group_size": 64,
241
+ "bits": 8,
242
+ "mode": "affine"
243
+ },
244
+ "language_model.model.layers.13.mlp.gate": {
245
+ "group_size": 64,
246
+ "bits": 8,
247
+ "mode": "affine"
248
+ },
249
+ "language_model.model.layers.13.mlp.shared_expert_gate": {
250
+ "group_size": 64,
251
+ "bits": 8,
252
+ "mode": "affine"
253
+ },
254
+ "language_model.model.layers.14.mlp.gate": {
255
+ "group_size": 64,
256
+ "bits": 8,
257
+ "mode": "affine"
258
+ },
259
+ "language_model.model.layers.14.mlp.shared_expert_gate": {
260
+ "group_size": 64,
261
+ "bits": 8,
262
+ "mode": "affine"
263
+ },
264
+ "language_model.model.layers.15.mlp.gate": {
265
+ "group_size": 64,
266
+ "bits": 8,
267
+ "mode": "affine"
268
+ },
269
+ "language_model.model.layers.15.mlp.shared_expert_gate": {
270
+ "group_size": 64,
271
+ "bits": 8,
272
+ "mode": "affine"
273
+ },
274
+ "language_model.model.layers.16.mlp.gate": {
275
+ "group_size": 64,
276
+ "bits": 8,
277
+ "mode": "affine"
278
+ },
279
+ "language_model.model.layers.16.mlp.shared_expert_gate": {
280
+ "group_size": 64,
281
+ "bits": 8,
282
+ "mode": "affine"
283
+ },
284
+ "language_model.model.layers.17.mlp.gate": {
285
+ "group_size": 64,
286
+ "bits": 8,
287
+ "mode": "affine"
288
+ },
289
+ "language_model.model.layers.17.mlp.shared_expert_gate": {
290
+ "group_size": 64,
291
+ "bits": 8,
292
+ "mode": "affine"
293
+ },
294
+ "language_model.model.layers.18.mlp.gate": {
295
+ "group_size": 64,
296
+ "bits": 8,
297
+ "mode": "affine"
298
+ },
299
+ "language_model.model.layers.18.mlp.shared_expert_gate": {
300
+ "group_size": 64,
301
+ "bits": 8,
302
+ "mode": "affine"
303
+ },
304
+ "language_model.model.layers.19.mlp.gate": {
305
+ "group_size": 64,
306
+ "bits": 8,
307
+ "mode": "affine"
308
+ },
309
+ "language_model.model.layers.19.mlp.shared_expert_gate": {
310
+ "group_size": 64,
311
+ "bits": 8,
312
+ "mode": "affine"
313
+ },
314
+ "language_model.model.layers.20.mlp.gate": {
315
+ "group_size": 64,
316
+ "bits": 8,
317
+ "mode": "affine"
318
+ },
319
+ "language_model.model.layers.20.mlp.shared_expert_gate": {
320
+ "group_size": 64,
321
+ "bits": 8,
322
+ "mode": "affine"
323
+ },
324
+ "language_model.model.layers.21.mlp.gate": {
325
+ "group_size": 64,
326
+ "bits": 8,
327
+ "mode": "affine"
328
+ },
329
+ "language_model.model.layers.21.mlp.shared_expert_gate": {
330
+ "group_size": 64,
331
+ "bits": 8,
332
+ "mode": "affine"
333
+ },
334
+ "language_model.model.layers.22.mlp.gate": {
335
+ "group_size": 64,
336
+ "bits": 8,
337
+ "mode": "affine"
338
+ },
339
+ "language_model.model.layers.22.mlp.shared_expert_gate": {
340
+ "group_size": 64,
341
+ "bits": 8,
342
+ "mode": "affine"
343
+ },
344
+ "language_model.model.layers.23.mlp.gate": {
345
+ "group_size": 64,
346
+ "bits": 8,
347
+ "mode": "affine"
348
+ },
349
+ "language_model.model.layers.23.mlp.shared_expert_gate": {
350
+ "group_size": 64,
351
+ "bits": 8,
352
+ "mode": "affine"
353
+ },
354
+ "language_model.model.layers.24.mlp.gate": {
355
+ "group_size": 64,
356
+ "bits": 8,
357
+ "mode": "affine"
358
+ },
359
+ "language_model.model.layers.24.mlp.shared_expert_gate": {
360
+ "group_size": 64,
361
+ "bits": 8,
362
+ "mode": "affine"
363
+ },
364
+ "language_model.model.layers.25.mlp.gate": {
365
+ "group_size": 64,
366
+ "bits": 8,
367
+ "mode": "affine"
368
+ },
369
+ "language_model.model.layers.25.mlp.shared_expert_gate": {
370
+ "group_size": 64,
371
+ "bits": 8,
372
+ "mode": "affine"
373
+ },
374
+ "language_model.model.layers.26.mlp.gate": {
375
+ "group_size": 64,
376
+ "bits": 8,
377
+ "mode": "affine"
378
+ },
379
+ "language_model.model.layers.26.mlp.shared_expert_gate": {
380
+ "group_size": 64,
381
+ "bits": 8,
382
+ "mode": "affine"
383
+ },
384
+ "language_model.model.layers.27.mlp.gate": {
385
+ "group_size": 64,
386
+ "bits": 8,
387
+ "mode": "affine"
388
+ },
389
+ "language_model.model.layers.27.mlp.shared_expert_gate": {
390
+ "group_size": 64,
391
+ "bits": 8,
392
+ "mode": "affine"
393
+ },
394
+ "language_model.model.layers.28.mlp.gate": {
395
+ "group_size": 64,
396
+ "bits": 8,
397
+ "mode": "affine"
398
+ },
399
+ "language_model.model.layers.28.mlp.shared_expert_gate": {
400
+ "group_size": 64,
401
+ "bits": 8,
402
+ "mode": "affine"
403
+ },
404
+ "language_model.model.layers.29.mlp.gate": {
405
+ "group_size": 64,
406
+ "bits": 8,
407
+ "mode": "affine"
408
+ },
409
+ "language_model.model.layers.29.mlp.shared_expert_gate": {
410
+ "group_size": 64,
411
+ "bits": 8,
412
+ "mode": "affine"
413
+ },
414
+ "language_model.model.layers.30.mlp.gate": {
415
+ "group_size": 64,
416
+ "bits": 8,
417
+ "mode": "affine"
418
+ },
419
+ "language_model.model.layers.30.mlp.shared_expert_gate": {
420
+ "group_size": 64,
421
+ "bits": 8,
422
+ "mode": "affine"
423
+ },
424
+ "language_model.model.layers.31.mlp.gate": {
425
+ "group_size": 64,
426
+ "bits": 8,
427
+ "mode": "affine"
428
+ },
429
+ "language_model.model.layers.31.mlp.shared_expert_gate": {
430
+ "group_size": 64,
431
+ "bits": 8,
432
+ "mode": "affine"
433
+ },
434
+ "language_model.model.layers.32.mlp.gate": {
435
+ "group_size": 64,
436
+ "bits": 8,
437
+ "mode": "affine"
438
+ },
439
+ "language_model.model.layers.32.mlp.shared_expert_gate": {
440
+ "group_size": 64,
441
+ "bits": 8,
442
+ "mode": "affine"
443
+ },
444
+ "language_model.model.layers.33.mlp.gate": {
445
+ "group_size": 64,
446
+ "bits": 8,
447
+ "mode": "affine"
448
+ },
449
+ "language_model.model.layers.33.mlp.shared_expert_gate": {
450
+ "group_size": 64,
451
+ "bits": 8,
452
+ "mode": "affine"
453
+ },
454
+ "language_model.model.layers.34.mlp.gate": {
455
+ "group_size": 64,
456
+ "bits": 8,
457
+ "mode": "affine"
458
+ },
459
+ "language_model.model.layers.34.mlp.shared_expert_gate": {
460
+ "group_size": 64,
461
+ "bits": 8,
462
+ "mode": "affine"
463
+ },
464
+ "language_model.model.layers.35.mlp.gate": {
465
+ "group_size": 64,
466
+ "bits": 8,
467
+ "mode": "affine"
468
+ },
469
+ "language_model.model.layers.35.mlp.shared_expert_gate": {
470
+ "group_size": 64,
471
+ "bits": 8,
472
+ "mode": "affine"
473
+ },
474
+ "language_model.model.layers.36.mlp.gate": {
475
+ "group_size": 64,
476
+ "bits": 8,
477
+ "mode": "affine"
478
+ },
479
+ "language_model.model.layers.36.mlp.shared_expert_gate": {
480
+ "group_size": 64,
481
+ "bits": 8,
482
+ "mode": "affine"
483
+ },
484
+ "language_model.model.layers.37.mlp.gate": {
485
+ "group_size": 64,
486
+ "bits": 8,
487
+ "mode": "affine"
488
+ },
489
+ "language_model.model.layers.37.mlp.shared_expert_gate": {
490
+ "group_size": 64,
491
+ "bits": 8,
492
+ "mode": "affine"
493
+ },
494
+ "language_model.model.layers.38.mlp.gate": {
495
+ "group_size": 64,
496
+ "bits": 8,
497
+ "mode": "affine"
498
+ },
499
+ "language_model.model.layers.38.mlp.shared_expert_gate": {
500
+ "group_size": 64,
501
+ "bits": 8,
502
+ "mode": "affine"
503
+ },
504
+ "language_model.model.layers.39.mlp.gate": {
505
+ "group_size": 64,
506
+ "bits": 8,
507
+ "mode": "affine"
508
+ },
509
+ "language_model.model.layers.39.mlp.shared_expert_gate": {
510
+ "group_size": 64,
511
+ "bits": 8,
512
+ "mode": "affine"
513
+ }
514
+ },
515
+ "quantization_config": {
516
+ "group_size": 32,
517
+ "bits": 4,
518
+ "mode": "mxfp4",
519
+ "language_model.model.layers.0.mlp.gate": {
520
+ "group_size": 64,
521
+ "bits": 8,
522
+ "mode": "affine"
523
+ },
524
+ "language_model.model.layers.0.mlp.shared_expert_gate": {
525
+ "group_size": 64,
526
+ "bits": 8,
527
+ "mode": "affine"
528
+ },
529
+ "language_model.model.layers.1.mlp.gate": {
530
+ "group_size": 64,
531
+ "bits": 8,
532
+ "mode": "affine"
533
+ },
534
+ "language_model.model.layers.1.mlp.shared_expert_gate": {
535
+ "group_size": 64,
536
+ "bits": 8,
537
+ "mode": "affine"
538
+ },
539
+ "language_model.model.layers.2.mlp.gate": {
540
+ "group_size": 64,
541
+ "bits": 8,
542
+ "mode": "affine"
543
+ },
544
+ "language_model.model.layers.2.mlp.shared_expert_gate": {
545
+ "group_size": 64,
546
+ "bits": 8,
547
+ "mode": "affine"
548
+ },
549
+ "language_model.model.layers.3.mlp.gate": {
550
+ "group_size": 64,
551
+ "bits": 8,
552
+ "mode": "affine"
553
+ },
554
+ "language_model.model.layers.3.mlp.shared_expert_gate": {
555
+ "group_size": 64,
556
+ "bits": 8,
557
+ "mode": "affine"
558
+ },
559
+ "language_model.model.layers.4.mlp.gate": {
560
+ "group_size": 64,
561
+ "bits": 8,
562
+ "mode": "affine"
563
+ },
564
+ "language_model.model.layers.4.mlp.shared_expert_gate": {
565
+ "group_size": 64,
566
+ "bits": 8,
567
+ "mode": "affine"
568
+ },
569
+ "language_model.model.layers.5.mlp.gate": {
570
+ "group_size": 64,
571
+ "bits": 8,
572
+ "mode": "affine"
573
+ },
574
+ "language_model.model.layers.5.mlp.shared_expert_gate": {
575
+ "group_size": 64,
576
+ "bits": 8,
577
+ "mode": "affine"
578
+ },
579
+ "language_model.model.layers.6.mlp.gate": {
580
+ "group_size": 64,
581
+ "bits": 8,
582
+ "mode": "affine"
583
+ },
584
+ "language_model.model.layers.6.mlp.shared_expert_gate": {
585
+ "group_size": 64,
586
+ "bits": 8,
587
+ "mode": "affine"
588
+ },
589
+ "language_model.model.layers.7.mlp.gate": {
590
+ "group_size": 64,
591
+ "bits": 8,
592
+ "mode": "affine"
593
+ },
594
+ "language_model.model.layers.7.mlp.shared_expert_gate": {
595
+ "group_size": 64,
596
+ "bits": 8,
597
+ "mode": "affine"
598
+ },
599
+ "language_model.model.layers.8.mlp.gate": {
600
+ "group_size": 64,
601
+ "bits": 8,
602
+ "mode": "affine"
603
+ },
604
+ "language_model.model.layers.8.mlp.shared_expert_gate": {
605
+ "group_size": 64,
606
+ "bits": 8,
607
+ "mode": "affine"
608
+ },
609
+ "language_model.model.layers.9.mlp.gate": {
610
+ "group_size": 64,
611
+ "bits": 8,
612
+ "mode": "affine"
613
+ },
614
+ "language_model.model.layers.9.mlp.shared_expert_gate": {
615
+ "group_size": 64,
616
+ "bits": 8,
617
+ "mode": "affine"
618
+ },
619
+ "language_model.model.layers.10.mlp.gate": {
620
+ "group_size": 64,
621
+ "bits": 8,
622
+ "mode": "affine"
623
+ },
624
+ "language_model.model.layers.10.mlp.shared_expert_gate": {
625
+ "group_size": 64,
626
+ "bits": 8,
627
+ "mode": "affine"
628
+ },
629
+ "language_model.model.layers.11.mlp.gate": {
630
+ "group_size": 64,
631
+ "bits": 8,
632
+ "mode": "affine"
633
+ },
634
+ "language_model.model.layers.11.mlp.shared_expert_gate": {
635
+ "group_size": 64,
636
+ "bits": 8,
637
+ "mode": "affine"
638
+ },
639
+ "language_model.model.layers.12.mlp.gate": {
640
+ "group_size": 64,
641
+ "bits": 8,
642
+ "mode": "affine"
643
+ },
644
+ "language_model.model.layers.12.mlp.shared_expert_gate": {
645
+ "group_size": 64,
646
+ "bits": 8,
647
+ "mode": "affine"
648
+ },
649
+ "language_model.model.layers.13.mlp.gate": {
650
+ "group_size": 64,
651
+ "bits": 8,
652
+ "mode": "affine"
653
+ },
654
+ "language_model.model.layers.13.mlp.shared_expert_gate": {
655
+ "group_size": 64,
656
+ "bits": 8,
657
+ "mode": "affine"
658
+ },
659
+ "language_model.model.layers.14.mlp.gate": {
660
+ "group_size": 64,
661
+ "bits": 8,
662
+ "mode": "affine"
663
+ },
664
+ "language_model.model.layers.14.mlp.shared_expert_gate": {
665
+ "group_size": 64,
666
+ "bits": 8,
667
+ "mode": "affine"
668
+ },
669
+ "language_model.model.layers.15.mlp.gate": {
670
+ "group_size": 64,
671
+ "bits": 8,
672
+ "mode": "affine"
673
+ },
674
+ "language_model.model.layers.15.mlp.shared_expert_gate": {
675
+ "group_size": 64,
676
+ "bits": 8,
677
+ "mode": "affine"
678
+ },
679
+ "language_model.model.layers.16.mlp.gate": {
680
+ "group_size": 64,
681
+ "bits": 8,
682
+ "mode": "affine"
683
+ },
684
+ "language_model.model.layers.16.mlp.shared_expert_gate": {
685
+ "group_size": 64,
686
+ "bits": 8,
687
+ "mode": "affine"
688
+ },
689
+ "language_model.model.layers.17.mlp.gate": {
690
+ "group_size": 64,
691
+ "bits": 8,
692
+ "mode": "affine"
693
+ },
694
+ "language_model.model.layers.17.mlp.shared_expert_gate": {
695
+ "group_size": 64,
696
+ "bits": 8,
697
+ "mode": "affine"
698
+ },
699
+ "language_model.model.layers.18.mlp.gate": {
700
+ "group_size": 64,
701
+ "bits": 8,
702
+ "mode": "affine"
703
+ },
704
+ "language_model.model.layers.18.mlp.shared_expert_gate": {
705
+ "group_size": 64,
706
+ "bits": 8,
707
+ "mode": "affine"
708
+ },
709
+ "language_model.model.layers.19.mlp.gate": {
710
+ "group_size": 64,
711
+ "bits": 8,
712
+ "mode": "affine"
713
+ },
714
+ "language_model.model.layers.19.mlp.shared_expert_gate": {
715
+ "group_size": 64,
716
+ "bits": 8,
717
+ "mode": "affine"
718
+ },
719
+ "language_model.model.layers.20.mlp.gate": {
720
+ "group_size": 64,
721
+ "bits": 8,
722
+ "mode": "affine"
723
+ },
724
+ "language_model.model.layers.20.mlp.shared_expert_gate": {
725
+ "group_size": 64,
726
+ "bits": 8,
727
+ "mode": "affine"
728
+ },
729
+ "language_model.model.layers.21.mlp.gate": {
730
+ "group_size": 64,
731
+ "bits": 8,
732
+ "mode": "affine"
733
+ },
734
+ "language_model.model.layers.21.mlp.shared_expert_gate": {
735
+ "group_size": 64,
736
+ "bits": 8,
737
+ "mode": "affine"
738
+ },
739
+ "language_model.model.layers.22.mlp.gate": {
740
+ "group_size": 64,
741
+ "bits": 8,
742
+ "mode": "affine"
743
+ },
744
+ "language_model.model.layers.22.mlp.shared_expert_gate": {
745
+ "group_size": 64,
746
+ "bits": 8,
747
+ "mode": "affine"
748
+ },
749
+ "language_model.model.layers.23.mlp.gate": {
750
+ "group_size": 64,
751
+ "bits": 8,
752
+ "mode": "affine"
753
+ },
754
+ "language_model.model.layers.23.mlp.shared_expert_gate": {
755
+ "group_size": 64,
756
+ "bits": 8,
757
+ "mode": "affine"
758
+ },
759
+ "language_model.model.layers.24.mlp.gate": {
760
+ "group_size": 64,
761
+ "bits": 8,
762
+ "mode": "affine"
763
+ },
764
+ "language_model.model.layers.24.mlp.shared_expert_gate": {
765
+ "group_size": 64,
766
+ "bits": 8,
767
+ "mode": "affine"
768
+ },
769
+ "language_model.model.layers.25.mlp.gate": {
770
+ "group_size": 64,
771
+ "bits": 8,
772
+ "mode": "affine"
773
+ },
774
+ "language_model.model.layers.25.mlp.shared_expert_gate": {
775
+ "group_size": 64,
776
+ "bits": 8,
777
+ "mode": "affine"
778
+ },
779
+ "language_model.model.layers.26.mlp.gate": {
780
+ "group_size": 64,
781
+ "bits": 8,
782
+ "mode": "affine"
783
+ },
784
+ "language_model.model.layers.26.mlp.shared_expert_gate": {
785
+ "group_size": 64,
786
+ "bits": 8,
787
+ "mode": "affine"
788
+ },
789
+ "language_model.model.layers.27.mlp.gate": {
790
+ "group_size": 64,
791
+ "bits": 8,
792
+ "mode": "affine"
793
+ },
794
+ "language_model.model.layers.27.mlp.shared_expert_gate": {
795
+ "group_size": 64,
796
+ "bits": 8,
797
+ "mode": "affine"
798
+ },
799
+ "language_model.model.layers.28.mlp.gate": {
800
+ "group_size": 64,
801
+ "bits": 8,
802
+ "mode": "affine"
803
+ },
804
+ "language_model.model.layers.28.mlp.shared_expert_gate": {
805
+ "group_size": 64,
806
+ "bits": 8,
807
+ "mode": "affine"
808
+ },
809
+ "language_model.model.layers.29.mlp.gate": {
810
+ "group_size": 64,
811
+ "bits": 8,
812
+ "mode": "affine"
813
+ },
814
+ "language_model.model.layers.29.mlp.shared_expert_gate": {
815
+ "group_size": 64,
816
+ "bits": 8,
817
+ "mode": "affine"
818
+ },
819
+ "language_model.model.layers.30.mlp.gate": {
820
+ "group_size": 64,
821
+ "bits": 8,
822
+ "mode": "affine"
823
+ },
824
+ "language_model.model.layers.30.mlp.shared_expert_gate": {
825
+ "group_size": 64,
826
+ "bits": 8,
827
+ "mode": "affine"
828
+ },
829
+ "language_model.model.layers.31.mlp.gate": {
830
+ "group_size": 64,
831
+ "bits": 8,
832
+ "mode": "affine"
833
+ },
834
+ "language_model.model.layers.31.mlp.shared_expert_gate": {
835
+ "group_size": 64,
836
+ "bits": 8,
837
+ "mode": "affine"
838
+ },
839
+ "language_model.model.layers.32.mlp.gate": {
840
+ "group_size": 64,
841
+ "bits": 8,
842
+ "mode": "affine"
843
+ },
844
+ "language_model.model.layers.32.mlp.shared_expert_gate": {
845
+ "group_size": 64,
846
+ "bits": 8,
847
+ "mode": "affine"
848
+ },
849
+ "language_model.model.layers.33.mlp.gate": {
850
+ "group_size": 64,
851
+ "bits": 8,
852
+ "mode": "affine"
853
+ },
854
+ "language_model.model.layers.33.mlp.shared_expert_gate": {
855
+ "group_size": 64,
856
+ "bits": 8,
857
+ "mode": "affine"
858
+ },
859
+ "language_model.model.layers.34.mlp.gate": {
860
+ "group_size": 64,
861
+ "bits": 8,
862
+ "mode": "affine"
863
+ },
864
+ "language_model.model.layers.34.mlp.shared_expert_gate": {
865
+ "group_size": 64,
866
+ "bits": 8,
867
+ "mode": "affine"
868
+ },
869
+ "language_model.model.layers.35.mlp.gate": {
870
+ "group_size": 64,
871
+ "bits": 8,
872
+ "mode": "affine"
873
+ },
874
+ "language_model.model.layers.35.mlp.shared_expert_gate": {
875
+ "group_size": 64,
876
+ "bits": 8,
877
+ "mode": "affine"
878
+ },
879
+ "language_model.model.layers.36.mlp.gate": {
880
+ "group_size": 64,
881
+ "bits": 8,
882
+ "mode": "affine"
883
+ },
884
+ "language_model.model.layers.36.mlp.shared_expert_gate": {
885
+ "group_size": 64,
886
+ "bits": 8,
887
+ "mode": "affine"
888
+ },
889
+ "language_model.model.layers.37.mlp.gate": {
890
+ "group_size": 64,
891
+ "bits": 8,
892
+ "mode": "affine"
893
+ },
894
+ "language_model.model.layers.37.mlp.shared_expert_gate": {
895
+ "group_size": 64,
896
+ "bits": 8,
897
+ "mode": "affine"
898
+ },
899
+ "language_model.model.layers.38.mlp.gate": {
900
+ "group_size": 64,
901
+ "bits": 8,
902
+ "mode": "affine"
903
+ },
904
+ "language_model.model.layers.38.mlp.shared_expert_gate": {
905
+ "group_size": 64,
906
+ "bits": 8,
907
+ "mode": "affine"
908
+ },
909
+ "language_model.model.layers.39.mlp.gate": {
910
+ "group_size": 64,
911
+ "bits": 8,
912
+ "mode": "affine"
913
+ },
914
+ "language_model.model.layers.39.mlp.shared_expert_gate": {
915
+ "group_size": 64,
916
+ "bits": 8,
917
+ "mode": "affine"
918
+ }
919
+ },
920
+ "abliteration": {
921
+ "method": "residual_direction_weight_orthogonalization",
922
+ "strength": 2.0,
923
+ "targets": "attention,shared_down,switch_down",
924
+ "layers": "",
925
+ "direction_scope": "global",
926
+ "preserve_column_norm": true,
927
+ "edited_modules": [
928
+ {
929
+ "name": "language_model.model.layers.39.mlp.shared_expert.down_proj",
930
+ "kind": "shared_down",
931
+ "layer": 39,
932
+ "axis": "output",
933
+ "bits": 4,
934
+ "mode": "mxfp4"
935
+ },
936
+ {
937
+ "name": "language_model.model.layers.39.mlp.switch_mlp.down_proj",
938
+ "kind": "switch_down",
939
+ "layer": 39,
940
+ "axis": "output",
941
+ "bits": 4,
942
+ "mode": "mxfp4"
943
+ },
944
+ {
945
+ "name": "language_model.model.layers.39.self_attn.o_proj",
946
+ "kind": "attention",
947
+ "layer": 39,
948
+ "axis": "output",
949
+ "bits": 4,
950
+ "mode": "mxfp4"
951
+ },
952
+ {
953
+ "name": "language_model.model.layers.38.mlp.shared_expert.down_proj",
954
+ "kind": "shared_down",
955
+ "layer": 38,
956
+ "axis": "output",
957
+ "bits": 4,
958
+ "mode": "mxfp4"
959
+ },
960
+ {
961
+ "name": "language_model.model.layers.38.mlp.switch_mlp.down_proj",
962
+ "kind": "switch_down",
963
+ "layer": 38,
964
+ "axis": "output",
965
+ "bits": 4,
966
+ "mode": "mxfp4"
967
+ },
968
+ {
969
+ "name": "language_model.model.layers.38.linear_attn.out_proj",
970
+ "kind": "attention",
971
+ "layer": 38,
972
+ "axis": "output",
973
+ "bits": 4,
974
+ "mode": "mxfp4"
975
+ },
976
+ {
977
+ "name": "language_model.model.layers.37.mlp.shared_expert.down_proj",
978
+ "kind": "shared_down",
979
+ "layer": 37,
980
+ "axis": "output",
981
+ "bits": 4,
982
+ "mode": "mxfp4"
983
+ },
984
+ {
985
+ "name": "language_model.model.layers.37.mlp.switch_mlp.down_proj",
986
+ "kind": "switch_down",
987
+ "layer": 37,
988
+ "axis": "output",
989
+ "bits": 4,
990
+ "mode": "mxfp4"
991
+ },
992
+ {
993
+ "name": "language_model.model.layers.37.linear_attn.out_proj",
994
+ "kind": "attention",
995
+ "layer": 37,
996
+ "axis": "output",
997
+ "bits": 4,
998
+ "mode": "mxfp4"
999
+ },
1000
+ {
1001
+ "name": "language_model.model.layers.36.mlp.shared_expert.down_proj",
1002
+ "kind": "shared_down",
1003
+ "layer": 36,
1004
+ "axis": "output",
1005
+ "bits": 4,
1006
+ "mode": "mxfp4"
1007
+ },
1008
+ {
1009
+ "name": "language_model.model.layers.36.mlp.switch_mlp.down_proj",
1010
+ "kind": "switch_down",
1011
+ "layer": 36,
1012
+ "axis": "output",
1013
+ "bits": 4,
1014
+ "mode": "mxfp4"
1015
+ },
1016
+ {
1017
+ "name": "language_model.model.layers.36.linear_attn.out_proj",
1018
+ "kind": "attention",
1019
+ "layer": 36,
1020
+ "axis": "output",
1021
+ "bits": 4,
1022
+ "mode": "mxfp4"
1023
+ },
1024
+ {
1025
+ "name": "language_model.model.layers.35.mlp.shared_expert.down_proj",
1026
+ "kind": "shared_down",
1027
+ "layer": 35,
1028
+ "axis": "output",
1029
+ "bits": 4,
1030
+ "mode": "mxfp4"
1031
+ },
1032
+ {
1033
+ "name": "language_model.model.layers.35.mlp.switch_mlp.down_proj",
1034
+ "kind": "switch_down",
1035
+ "layer": 35,
1036
+ "axis": "output",
1037
+ "bits": 4,
1038
+ "mode": "mxfp4"
1039
+ },
1040
+ {
1041
+ "name": "language_model.model.layers.35.self_attn.o_proj",
1042
+ "kind": "attention",
1043
+ "layer": 35,
1044
+ "axis": "output",
1045
+ "bits": 4,
1046
+ "mode": "mxfp4"
1047
+ },
1048
+ {
1049
+ "name": "language_model.model.layers.34.mlp.shared_expert.down_proj",
1050
+ "kind": "shared_down",
1051
+ "layer": 34,
1052
+ "axis": "output",
1053
+ "bits": 4,
1054
+ "mode": "mxfp4"
1055
+ },
1056
+ {
1057
+ "name": "language_model.model.layers.34.mlp.switch_mlp.down_proj",
1058
+ "kind": "switch_down",
1059
+ "layer": 34,
1060
+ "axis": "output",
1061
+ "bits": 4,
1062
+ "mode": "mxfp4"
1063
+ },
1064
+ {
1065
+ "name": "language_model.model.layers.34.linear_attn.out_proj",
1066
+ "kind": "attention",
1067
+ "layer": 34,
1068
+ "axis": "output",
1069
+ "bits": 4,
1070
+ "mode": "mxfp4"
1071
+ },
1072
+ {
1073
+ "name": "language_model.model.layers.33.mlp.shared_expert.down_proj",
1074
+ "kind": "shared_down",
1075
+ "layer": 33,
1076
+ "axis": "output",
1077
+ "bits": 4,
1078
+ "mode": "mxfp4"
1079
+ },
1080
+ {
1081
+ "name": "language_model.model.layers.33.mlp.switch_mlp.down_proj",
1082
+ "kind": "switch_down",
1083
+ "layer": 33,
1084
+ "axis": "output",
1085
+ "bits": 4,
1086
+ "mode": "mxfp4"
1087
+ },
1088
+ {
1089
+ "name": "language_model.model.layers.33.linear_attn.out_proj",
1090
+ "kind": "attention",
1091
+ "layer": 33,
1092
+ "axis": "output",
1093
+ "bits": 4,
1094
+ "mode": "mxfp4"
1095
+ },
1096
+ {
1097
+ "name": "language_model.model.layers.32.mlp.shared_expert.down_proj",
1098
+ "kind": "shared_down",
1099
+ "layer": 32,
1100
+ "axis": "output",
1101
+ "bits": 4,
1102
+ "mode": "mxfp4"
1103
+ },
1104
+ {
1105
+ "name": "language_model.model.layers.32.mlp.switch_mlp.down_proj",
1106
+ "kind": "switch_down",
1107
+ "layer": 32,
1108
+ "axis": "output",
1109
+ "bits": 4,
1110
+ "mode": "mxfp4"
1111
+ },
1112
+ {
1113
+ "name": "language_model.model.layers.32.linear_attn.out_proj",
1114
+ "kind": "attention",
1115
+ "layer": 32,
1116
+ "axis": "output",
1117
+ "bits": 4,
1118
+ "mode": "mxfp4"
1119
+ },
1120
+ {
1121
+ "name": "language_model.model.layers.31.mlp.shared_expert.down_proj",
1122
+ "kind": "shared_down",
1123
+ "layer": 31,
1124
+ "axis": "output",
1125
+ "bits": 4,
1126
+ "mode": "mxfp4"
1127
+ },
1128
+ {
1129
+ "name": "language_model.model.layers.31.mlp.switch_mlp.down_proj",
1130
+ "kind": "switch_down",
1131
+ "layer": 31,
1132
+ "axis": "output",
1133
+ "bits": 4,
1134
+ "mode": "mxfp4"
1135
+ },
1136
+ {
1137
+ "name": "language_model.model.layers.31.self_attn.o_proj",
1138
+ "kind": "attention",
1139
+ "layer": 31,
1140
+ "axis": "output",
1141
+ "bits": 4,
1142
+ "mode": "mxfp4"
1143
+ },
1144
+ {
1145
+ "name": "language_model.model.layers.30.mlp.shared_expert.down_proj",
1146
+ "kind": "shared_down",
1147
+ "layer": 30,
1148
+ "axis": "output",
1149
+ "bits": 4,
1150
+ "mode": "mxfp4"
1151
+ },
1152
+ {
1153
+ "name": "language_model.model.layers.30.mlp.switch_mlp.down_proj",
1154
+ "kind": "switch_down",
1155
+ "layer": 30,
1156
+ "axis": "output",
1157
+ "bits": 4,
1158
+ "mode": "mxfp4"
1159
+ },
1160
+ {
1161
+ "name": "language_model.model.layers.30.linear_attn.out_proj",
1162
+ "kind": "attention",
1163
+ "layer": 30,
1164
+ "axis": "output",
1165
+ "bits": 4,
1166
+ "mode": "mxfp4"
1167
+ },
1168
+ {
1169
+ "name": "language_model.model.layers.29.mlp.shared_expert.down_proj",
1170
+ "kind": "shared_down",
1171
+ "layer": 29,
1172
+ "axis": "output",
1173
+ "bits": 4,
1174
+ "mode": "mxfp4"
1175
+ },
1176
+ {
1177
+ "name": "language_model.model.layers.29.mlp.switch_mlp.down_proj",
1178
+ "kind": "switch_down",
1179
+ "layer": 29,
1180
+ "axis": "output",
1181
+ "bits": 4,
1182
+ "mode": "mxfp4"
1183
+ },
1184
+ {
1185
+ "name": "language_model.model.layers.29.linear_attn.out_proj",
1186
+ "kind": "attention",
1187
+ "layer": 29,
1188
+ "axis": "output",
1189
+ "bits": 4,
1190
+ "mode": "mxfp4"
1191
+ },
1192
+ {
1193
+ "name": "language_model.model.layers.28.mlp.shared_expert.down_proj",
1194
+ "kind": "shared_down",
1195
+ "layer": 28,
1196
+ "axis": "output",
1197
+ "bits": 4,
1198
+ "mode": "mxfp4"
1199
+ },
1200
+ {
1201
+ "name": "language_model.model.layers.28.mlp.switch_mlp.down_proj",
1202
+ "kind": "switch_down",
1203
+ "layer": 28,
1204
+ "axis": "output",
1205
+ "bits": 4,
1206
+ "mode": "mxfp4"
1207
+ },
1208
+ {
1209
+ "name": "language_model.model.layers.28.linear_attn.out_proj",
1210
+ "kind": "attention",
1211
+ "layer": 28,
1212
+ "axis": "output",
1213
+ "bits": 4,
1214
+ "mode": "mxfp4"
1215
+ },
1216
+ {
1217
+ "name": "language_model.model.layers.27.mlp.shared_expert.down_proj",
1218
+ "kind": "shared_down",
1219
+ "layer": 27,
1220
+ "axis": "output",
1221
+ "bits": 4,
1222
+ "mode": "mxfp4"
1223
+ },
1224
+ {
1225
+ "name": "language_model.model.layers.27.mlp.switch_mlp.down_proj",
1226
+ "kind": "switch_down",
1227
+ "layer": 27,
1228
+ "axis": "output",
1229
+ "bits": 4,
1230
+ "mode": "mxfp4"
1231
+ },
1232
+ {
1233
+ "name": "language_model.model.layers.27.self_attn.o_proj",
1234
+ "kind": "attention",
1235
+ "layer": 27,
1236
+ "axis": "output",
1237
+ "bits": 4,
1238
+ "mode": "mxfp4"
1239
+ },
1240
+ {
1241
+ "name": "language_model.model.layers.26.mlp.shared_expert.down_proj",
1242
+ "kind": "shared_down",
1243
+ "layer": 26,
1244
+ "axis": "output",
1245
+ "bits": 4,
1246
+ "mode": "mxfp4"
1247
+ },
1248
+ {
1249
+ "name": "language_model.model.layers.26.mlp.switch_mlp.down_proj",
1250
+ "kind": "switch_down",
1251
+ "layer": 26,
1252
+ "axis": "output",
1253
+ "bits": 4,
1254
+ "mode": "mxfp4"
1255
+ },
1256
+ {
1257
+ "name": "language_model.model.layers.26.linear_attn.out_proj",
1258
+ "kind": "attention",
1259
+ "layer": 26,
1260
+ "axis": "output",
1261
+ "bits": 4,
1262
+ "mode": "mxfp4"
1263
+ },
1264
+ {
1265
+ "name": "language_model.model.layers.25.mlp.shared_expert.down_proj",
1266
+ "kind": "shared_down",
1267
+ "layer": 25,
1268
+ "axis": "output",
1269
+ "bits": 4,
1270
+ "mode": "mxfp4"
1271
+ },
1272
+ {
1273
+ "name": "language_model.model.layers.25.mlp.switch_mlp.down_proj",
1274
+ "kind": "switch_down",
1275
+ "layer": 25,
1276
+ "axis": "output",
1277
+ "bits": 4,
1278
+ "mode": "mxfp4"
1279
+ },
1280
+ {
1281
+ "name": "language_model.model.layers.25.linear_attn.out_proj",
1282
+ "kind": "attention",
1283
+ "layer": 25,
1284
+ "axis": "output",
1285
+ "bits": 4,
1286
+ "mode": "mxfp4"
1287
+ },
1288
+ {
1289
+ "name": "language_model.model.layers.24.mlp.shared_expert.down_proj",
1290
+ "kind": "shared_down",
1291
+ "layer": 24,
1292
+ "axis": "output",
1293
+ "bits": 4,
1294
+ "mode": "mxfp4"
1295
+ },
1296
+ {
1297
+ "name": "language_model.model.layers.24.mlp.switch_mlp.down_proj",
1298
+ "kind": "switch_down",
1299
+ "layer": 24,
1300
+ "axis": "output",
1301
+ "bits": 4,
1302
+ "mode": "mxfp4"
1303
+ },
1304
+ {
1305
+ "name": "language_model.model.layers.24.linear_attn.out_proj",
1306
+ "kind": "attention",
1307
+ "layer": 24,
1308
+ "axis": "output",
1309
+ "bits": 4,
1310
+ "mode": "mxfp4"
1311
+ },
1312
+ {
1313
+ "name": "language_model.model.layers.23.mlp.shared_expert.down_proj",
1314
+ "kind": "shared_down",
1315
+ "layer": 23,
1316
+ "axis": "output",
1317
+ "bits": 4,
1318
+ "mode": "mxfp4"
1319
+ },
1320
+ {
1321
+ "name": "language_model.model.layers.23.mlp.switch_mlp.down_proj",
1322
+ "kind": "switch_down",
1323
+ "layer": 23,
1324
+ "axis": "output",
1325
+ "bits": 4,
1326
+ "mode": "mxfp4"
1327
+ },
1328
+ {
1329
+ "name": "language_model.model.layers.23.self_attn.o_proj",
1330
+ "kind": "attention",
1331
+ "layer": 23,
1332
+ "axis": "output",
1333
+ "bits": 4,
1334
+ "mode": "mxfp4"
1335
+ },
1336
+ {
1337
+ "name": "language_model.model.layers.22.mlp.shared_expert.down_proj",
1338
+ "kind": "shared_down",
1339
+ "layer": 22,
1340
+ "axis": "output",
1341
+ "bits": 4,
1342
+ "mode": "mxfp4"
1343
+ },
1344
+ {
1345
+ "name": "language_model.model.layers.22.mlp.switch_mlp.down_proj",
1346
+ "kind": "switch_down",
1347
+ "layer": 22,
1348
+ "axis": "output",
1349
+ "bits": 4,
1350
+ "mode": "mxfp4"
1351
+ },
1352
+ {
1353
+ "name": "language_model.model.layers.22.linear_attn.out_proj",
1354
+ "kind": "attention",
1355
+ "layer": 22,
1356
+ "axis": "output",
1357
+ "bits": 4,
1358
+ "mode": "mxfp4"
1359
+ },
1360
+ {
1361
+ "name": "language_model.model.layers.21.mlp.shared_expert.down_proj",
1362
+ "kind": "shared_down",
1363
+ "layer": 21,
1364
+ "axis": "output",
1365
+ "bits": 4,
1366
+ "mode": "mxfp4"
1367
+ },
1368
+ {
1369
+ "name": "language_model.model.layers.21.mlp.switch_mlp.down_proj",
1370
+ "kind": "switch_down",
1371
+ "layer": 21,
1372
+ "axis": "output",
1373
+ "bits": 4,
1374
+ "mode": "mxfp4"
1375
+ },
1376
+ {
1377
+ "name": "language_model.model.layers.21.linear_attn.out_proj",
1378
+ "kind": "attention",
1379
+ "layer": 21,
1380
+ "axis": "output",
1381
+ "bits": 4,
1382
+ "mode": "mxfp4"
1383
+ },
1384
+ {
1385
+ "name": "language_model.model.layers.20.mlp.shared_expert.down_proj",
1386
+ "kind": "shared_down",
1387
+ "layer": 20,
1388
+ "axis": "output",
1389
+ "bits": 4,
1390
+ "mode": "mxfp4"
1391
+ },
1392
+ {
1393
+ "name": "language_model.model.layers.20.mlp.switch_mlp.down_proj",
1394
+ "kind": "switch_down",
1395
+ "layer": 20,
1396
+ "axis": "output",
1397
+ "bits": 4,
1398
+ "mode": "mxfp4"
1399
+ },
1400
+ {
1401
+ "name": "language_model.model.layers.20.linear_attn.out_proj",
1402
+ "kind": "attention",
1403
+ "layer": 20,
1404
+ "axis": "output",
1405
+ "bits": 4,
1406
+ "mode": "mxfp4"
1407
+ },
1408
+ {
1409
+ "name": "language_model.model.layers.19.mlp.shared_expert.down_proj",
1410
+ "kind": "shared_down",
1411
+ "layer": 19,
1412
+ "axis": "output",
1413
+ "bits": 4,
1414
+ "mode": "mxfp4"
1415
+ },
1416
+ {
1417
+ "name": "language_model.model.layers.19.mlp.switch_mlp.down_proj",
1418
+ "kind": "switch_down",
1419
+ "layer": 19,
1420
+ "axis": "output",
1421
+ "bits": 4,
1422
+ "mode": "mxfp4"
1423
+ },
1424
+ {
1425
+ "name": "language_model.model.layers.19.self_attn.o_proj",
1426
+ "kind": "attention",
1427
+ "layer": 19,
1428
+ "axis": "output",
1429
+ "bits": 4,
1430
+ "mode": "mxfp4"
1431
+ },
1432
+ {
1433
+ "name": "language_model.model.layers.18.mlp.shared_expert.down_proj",
1434
+ "kind": "shared_down",
1435
+ "layer": 18,
1436
+ "axis": "output",
1437
+ "bits": 4,
1438
+ "mode": "mxfp4"
1439
+ },
1440
+ {
1441
+ "name": "language_model.model.layers.18.mlp.switch_mlp.down_proj",
1442
+ "kind": "switch_down",
1443
+ "layer": 18,
1444
+ "axis": "output",
1445
+ "bits": 4,
1446
+ "mode": "mxfp4"
1447
+ },
1448
+ {
1449
+ "name": "language_model.model.layers.18.linear_attn.out_proj",
1450
+ "kind": "attention",
1451
+ "layer": 18,
1452
+ "axis": "output",
1453
+ "bits": 4,
1454
+ "mode": "mxfp4"
1455
+ },
1456
+ {
1457
+ "name": "language_model.model.layers.17.mlp.shared_expert.down_proj",
1458
+ "kind": "shared_down",
1459
+ "layer": 17,
1460
+ "axis": "output",
1461
+ "bits": 4,
1462
+ "mode": "mxfp4"
1463
+ },
1464
+ {
1465
+ "name": "language_model.model.layers.17.mlp.switch_mlp.down_proj",
1466
+ "kind": "switch_down",
1467
+ "layer": 17,
1468
+ "axis": "output",
1469
+ "bits": 4,
1470
+ "mode": "mxfp4"
1471
+ },
1472
+ {
1473
+ "name": "language_model.model.layers.17.linear_attn.out_proj",
1474
+ "kind": "attention",
1475
+ "layer": 17,
1476
+ "axis": "output",
1477
+ "bits": 4,
1478
+ "mode": "mxfp4"
1479
+ },
1480
+ {
1481
+ "name": "language_model.model.layers.16.mlp.shared_expert.down_proj",
1482
+ "kind": "shared_down",
1483
+ "layer": 16,
1484
+ "axis": "output",
1485
+ "bits": 4,
1486
+ "mode": "mxfp4"
1487
+ },
1488
+ {
1489
+ "name": "language_model.model.layers.16.mlp.switch_mlp.down_proj",
1490
+ "kind": "switch_down",
1491
+ "layer": 16,
1492
+ "axis": "output",
1493
+ "bits": 4,
1494
+ "mode": "mxfp4"
1495
+ },
1496
+ {
1497
+ "name": "language_model.model.layers.16.linear_attn.out_proj",
1498
+ "kind": "attention",
1499
+ "layer": 16,
1500
+ "axis": "output",
1501
+ "bits": 4,
1502
+ "mode": "mxfp4"
1503
+ },
1504
+ {
1505
+ "name": "language_model.model.layers.15.mlp.shared_expert.down_proj",
1506
+ "kind": "shared_down",
1507
+ "layer": 15,
1508
+ "axis": "output",
1509
+ "bits": 4,
1510
+ "mode": "mxfp4"
1511
+ },
1512
+ {
1513
+ "name": "language_model.model.layers.15.mlp.switch_mlp.down_proj",
1514
+ "kind": "switch_down",
1515
+ "layer": 15,
1516
+ "axis": "output",
1517
+ "bits": 4,
1518
+ "mode": "mxfp4"
1519
+ },
1520
+ {
1521
+ "name": "language_model.model.layers.15.self_attn.o_proj",
1522
+ "kind": "attention",
1523
+ "layer": 15,
1524
+ "axis": "output",
1525
+ "bits": 4,
1526
+ "mode": "mxfp4"
1527
+ },
1528
+ {
1529
+ "name": "language_model.model.layers.14.mlp.shared_expert.down_proj",
1530
+ "kind": "shared_down",
1531
+ "layer": 14,
1532
+ "axis": "output",
1533
+ "bits": 4,
1534
+ "mode": "mxfp4"
1535
+ },
1536
+ {
1537
+ "name": "language_model.model.layers.14.mlp.switch_mlp.down_proj",
1538
+ "kind": "switch_down",
1539
+ "layer": 14,
1540
+ "axis": "output",
1541
+ "bits": 4,
1542
+ "mode": "mxfp4"
1543
+ },
1544
+ {
1545
+ "name": "language_model.model.layers.14.linear_attn.out_proj",
1546
+ "kind": "attention",
1547
+ "layer": 14,
1548
+ "axis": "output",
1549
+ "bits": 4,
1550
+ "mode": "mxfp4"
1551
+ },
1552
+ {
1553
+ "name": "language_model.model.layers.13.mlp.shared_expert.down_proj",
1554
+ "kind": "shared_down",
1555
+ "layer": 13,
1556
+ "axis": "output",
1557
+ "bits": 4,
1558
+ "mode": "mxfp4"
1559
+ },
1560
+ {
1561
+ "name": "language_model.model.layers.13.mlp.switch_mlp.down_proj",
1562
+ "kind": "switch_down",
1563
+ "layer": 13,
1564
+ "axis": "output",
1565
+ "bits": 4,
1566
+ "mode": "mxfp4"
1567
+ },
1568
+ {
1569
+ "name": "language_model.model.layers.13.linear_attn.out_proj",
1570
+ "kind": "attention",
1571
+ "layer": 13,
1572
+ "axis": "output",
1573
+ "bits": 4,
1574
+ "mode": "mxfp4"
1575
+ },
1576
+ {
1577
+ "name": "language_model.model.layers.12.mlp.shared_expert.down_proj",
1578
+ "kind": "shared_down",
1579
+ "layer": 12,
1580
+ "axis": "output",
1581
+ "bits": 4,
1582
+ "mode": "mxfp4"
1583
+ },
1584
+ {
1585
+ "name": "language_model.model.layers.12.mlp.switch_mlp.down_proj",
1586
+ "kind": "switch_down",
1587
+ "layer": 12,
1588
+ "axis": "output",
1589
+ "bits": 4,
1590
+ "mode": "mxfp4"
1591
+ },
1592
+ {
1593
+ "name": "language_model.model.layers.12.linear_attn.out_proj",
1594
+ "kind": "attention",
1595
+ "layer": 12,
1596
+ "axis": "output",
1597
+ "bits": 4,
1598
+ "mode": "mxfp4"
1599
+ },
1600
+ {
1601
+ "name": "language_model.model.layers.11.mlp.shared_expert.down_proj",
1602
+ "kind": "shared_down",
1603
+ "layer": 11,
1604
+ "axis": "output",
1605
+ "bits": 4,
1606
+ "mode": "mxfp4"
1607
+ },
1608
+ {
1609
+ "name": "language_model.model.layers.11.mlp.switch_mlp.down_proj",
1610
+ "kind": "switch_down",
1611
+ "layer": 11,
1612
+ "axis": "output",
1613
+ "bits": 4,
1614
+ "mode": "mxfp4"
1615
+ },
1616
+ {
1617
+ "name": "language_model.model.layers.11.self_attn.o_proj",
1618
+ "kind": "attention",
1619
+ "layer": 11,
1620
+ "axis": "output",
1621
+ "bits": 4,
1622
+ "mode": "mxfp4"
1623
+ },
1624
+ {
1625
+ "name": "language_model.model.layers.10.mlp.shared_expert.down_proj",
1626
+ "kind": "shared_down",
1627
+ "layer": 10,
1628
+ "axis": "output",
1629
+ "bits": 4,
1630
+ "mode": "mxfp4"
1631
+ },
1632
+ {
1633
+ "name": "language_model.model.layers.10.mlp.switch_mlp.down_proj",
1634
+ "kind": "switch_down",
1635
+ "layer": 10,
1636
+ "axis": "output",
1637
+ "bits": 4,
1638
+ "mode": "mxfp4"
1639
+ },
1640
+ {
1641
+ "name": "language_model.model.layers.10.linear_attn.out_proj",
1642
+ "kind": "attention",
1643
+ "layer": 10,
1644
+ "axis": "output",
1645
+ "bits": 4,
1646
+ "mode": "mxfp4"
1647
+ },
1648
+ {
1649
+ "name": "language_model.model.layers.9.mlp.shared_expert.down_proj",
1650
+ "kind": "shared_down",
1651
+ "layer": 9,
1652
+ "axis": "output",
1653
+ "bits": 4,
1654
+ "mode": "mxfp4"
1655
+ },
1656
+ {
1657
+ "name": "language_model.model.layers.9.mlp.switch_mlp.down_proj",
1658
+ "kind": "switch_down",
1659
+ "layer": 9,
1660
+ "axis": "output",
1661
+ "bits": 4,
1662
+ "mode": "mxfp4"
1663
+ },
1664
+ {
1665
+ "name": "language_model.model.layers.9.linear_attn.out_proj",
1666
+ "kind": "attention",
1667
+ "layer": 9,
1668
+ "axis": "output",
1669
+ "bits": 4,
1670
+ "mode": "mxfp4"
1671
+ },
1672
+ {
1673
+ "name": "language_model.model.layers.8.mlp.shared_expert.down_proj",
1674
+ "kind": "shared_down",
1675
+ "layer": 8,
1676
+ "axis": "output",
1677
+ "bits": 4,
1678
+ "mode": "mxfp4"
1679
+ },
1680
+ {
1681
+ "name": "language_model.model.layers.8.mlp.switch_mlp.down_proj",
1682
+ "kind": "switch_down",
1683
+ "layer": 8,
1684
+ "axis": "output",
1685
+ "bits": 4,
1686
+ "mode": "mxfp4"
1687
+ },
1688
+ {
1689
+ "name": "language_model.model.layers.8.linear_attn.out_proj",
1690
+ "kind": "attention",
1691
+ "layer": 8,
1692
+ "axis": "output",
1693
+ "bits": 4,
1694
+ "mode": "mxfp4"
1695
+ },
1696
+ {
1697
+ "name": "language_model.model.layers.7.mlp.shared_expert.down_proj",
1698
+ "kind": "shared_down",
1699
+ "layer": 7,
1700
+ "axis": "output",
1701
+ "bits": 4,
1702
+ "mode": "mxfp4"
1703
+ },
1704
+ {
1705
+ "name": "language_model.model.layers.7.mlp.switch_mlp.down_proj",
1706
+ "kind": "switch_down",
1707
+ "layer": 7,
1708
+ "axis": "output",
1709
+ "bits": 4,
1710
+ "mode": "mxfp4"
1711
+ },
1712
+ {
1713
+ "name": "language_model.model.layers.7.self_attn.o_proj",
1714
+ "kind": "attention",
1715
+ "layer": 7,
1716
+ "axis": "output",
1717
+ "bits": 4,
1718
+ "mode": "mxfp4"
1719
+ },
1720
+ {
1721
+ "name": "language_model.model.layers.6.mlp.shared_expert.down_proj",
1722
+ "kind": "shared_down",
1723
+ "layer": 6,
1724
+ "axis": "output",
1725
+ "bits": 4,
1726
+ "mode": "mxfp4"
1727
+ },
1728
+ {
1729
+ "name": "language_model.model.layers.6.mlp.switch_mlp.down_proj",
1730
+ "kind": "switch_down",
1731
+ "layer": 6,
1732
+ "axis": "output",
1733
+ "bits": 4,
1734
+ "mode": "mxfp4"
1735
+ },
1736
+ {
1737
+ "name": "language_model.model.layers.6.linear_attn.out_proj",
1738
+ "kind": "attention",
1739
+ "layer": 6,
1740
+ "axis": "output",
1741
+ "bits": 4,
1742
+ "mode": "mxfp4"
1743
+ },
1744
+ {
1745
+ "name": "language_model.model.layers.5.mlp.shared_expert.down_proj",
1746
+ "kind": "shared_down",
1747
+ "layer": 5,
1748
+ "axis": "output",
1749
+ "bits": 4,
1750
+ "mode": "mxfp4"
1751
+ },
1752
+ {
1753
+ "name": "language_model.model.layers.5.mlp.switch_mlp.down_proj",
1754
+ "kind": "switch_down",
1755
+ "layer": 5,
1756
+ "axis": "output",
1757
+ "bits": 4,
1758
+ "mode": "mxfp4"
1759
+ },
1760
+ {
1761
+ "name": "language_model.model.layers.5.linear_attn.out_proj",
1762
+ "kind": "attention",
1763
+ "layer": 5,
1764
+ "axis": "output",
1765
+ "bits": 4,
1766
+ "mode": "mxfp4"
1767
+ },
1768
+ {
1769
+ "name": "language_model.model.layers.4.mlp.shared_expert.down_proj",
1770
+ "kind": "shared_down",
1771
+ "layer": 4,
1772
+ "axis": "output",
1773
+ "bits": 4,
1774
+ "mode": "mxfp4"
1775
+ },
1776
+ {
1777
+ "name": "language_model.model.layers.4.mlp.switch_mlp.down_proj",
1778
+ "kind": "switch_down",
1779
+ "layer": 4,
1780
+ "axis": "output",
1781
+ "bits": 4,
1782
+ "mode": "mxfp4"
1783
+ },
1784
+ {
1785
+ "name": "language_model.model.layers.4.linear_attn.out_proj",
1786
+ "kind": "attention",
1787
+ "layer": 4,
1788
+ "axis": "output",
1789
+ "bits": 4,
1790
+ "mode": "mxfp4"
1791
+ },
1792
+ {
1793
+ "name": "language_model.model.layers.3.mlp.shared_expert.down_proj",
1794
+ "kind": "shared_down",
1795
+ "layer": 3,
1796
+ "axis": "output",
1797
+ "bits": 4,
1798
+ "mode": "mxfp4"
1799
+ },
1800
+ {
1801
+ "name": "language_model.model.layers.3.mlp.switch_mlp.down_proj",
1802
+ "kind": "switch_down",
1803
+ "layer": 3,
1804
+ "axis": "output",
1805
+ "bits": 4,
1806
+ "mode": "mxfp4"
1807
+ },
1808
+ {
1809
+ "name": "language_model.model.layers.3.self_attn.o_proj",
1810
+ "kind": "attention",
1811
+ "layer": 3,
1812
+ "axis": "output",
1813
+ "bits": 4,
1814
+ "mode": "mxfp4"
1815
+ },
1816
+ {
1817
+ "name": "language_model.model.layers.2.mlp.shared_expert.down_proj",
1818
+ "kind": "shared_down",
1819
+ "layer": 2,
1820
+ "axis": "output",
1821
+ "bits": 4,
1822
+ "mode": "mxfp4"
1823
+ },
1824
+ {
1825
+ "name": "language_model.model.layers.2.mlp.switch_mlp.down_proj",
1826
+ "kind": "switch_down",
1827
+ "layer": 2,
1828
+ "axis": "output",
1829
+ "bits": 4,
1830
+ "mode": "mxfp4"
1831
+ },
1832
+ {
1833
+ "name": "language_model.model.layers.2.linear_attn.out_proj",
1834
+ "kind": "attention",
1835
+ "layer": 2,
1836
+ "axis": "output",
1837
+ "bits": 4,
1838
+ "mode": "mxfp4"
1839
+ },
1840
+ {
1841
+ "name": "language_model.model.layers.1.mlp.shared_expert.down_proj",
1842
+ "kind": "shared_down",
1843
+ "layer": 1,
1844
+ "axis": "output",
1845
+ "bits": 4,
1846
+ "mode": "mxfp4"
1847
+ },
1848
+ {
1849
+ "name": "language_model.model.layers.1.mlp.switch_mlp.down_proj",
1850
+ "kind": "switch_down",
1851
+ "layer": 1,
1852
+ "axis": "output",
1853
+ "bits": 4,
1854
+ "mode": "mxfp4"
1855
+ },
1856
+ {
1857
+ "name": "language_model.model.layers.1.linear_attn.out_proj",
1858
+ "kind": "attention",
1859
+ "layer": 1,
1860
+ "axis": "output",
1861
+ "bits": 4,
1862
+ "mode": "mxfp4"
1863
+ },
1864
+ {
1865
+ "name": "language_model.model.layers.0.mlp.shared_expert.down_proj",
1866
+ "kind": "shared_down",
1867
+ "layer": 0,
1868
+ "axis": "output",
1869
+ "bits": 4,
1870
+ "mode": "mxfp4"
1871
+ },
1872
+ {
1873
+ "name": "language_model.model.layers.0.mlp.switch_mlp.down_proj",
1874
+ "kind": "switch_down",
1875
+ "layer": 0,
1876
+ "axis": "output",
1877
+ "bits": 4,
1878
+ "mode": "mxfp4"
1879
+ },
1880
+ {
1881
+ "name": "language_model.model.layers.0.linear_attn.out_proj",
1882
+ "kind": "attention",
1883
+ "layer": 0,
1884
+ "axis": "output",
1885
+ "bits": 4,
1886
+ "mode": "mxfp4"
1887
+ }
1888
+ ],
1889
+ "edited_module_count": 120,
1890
+ "direction_description": "completion-projected refusal direction; source direction file is not required for inference"
1891
+ },
1892
+ "vision_config": {
1893
+ "deepstack_visual_indexes": [],
1894
+ "depth": 27,
1895
+ "dtype": "bfloat16",
1896
+ "hidden_act": "gelu_pytorch_tanh",
1897
+ "hidden_size": 1152,
1898
+ "in_channels": 3,
1899
+ "initializer_range": 0.02,
1900
+ "intermediate_size": 4304,
1901
+ "model_type": "qwen3_5_moe",
1902
+ "num_heads": 16,
1903
+ "num_position_embeddings": 2304,
1904
+ "out_hidden_size": 2048,
1905
+ "patch_size": 16,
1906
+ "spatial_merge_size": 2,
1907
+ "temporal_patch_size": 2
1908
+ },
1909
+ "publisher": "shiftedx"
1910
+ }
model-00001-of-00004.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8e1450f4f405025e7557d34acadb4904f09c03ca17a4e816fc6da2beabcbfade
3
+ size 5349133901
model-00002-of-00004.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ec3dc4b9a6754b2d28ff1428aded584e08359563530ff1d5d3994fc9b3453233
3
+ size 5365849645
model-00003-of-00004.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:520a8e3fb2a51c51ed01c2833183da164131528bb2add1ab980638e2cd1b6311
3
+ size 5365849659
model-00004-of-00004.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7f3f3434ef9e5e0f4ad019ff306ead5e57e511c5ddbbe536ed437a20e025a829
3
+ size 2345680871
model-vision-00001-of-00001.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d4579339122fdc5e06774fa0c188bb5b79410a8ab821343cec7a6122a0588014
3
+ size 893179536
model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
preprocessor_config.json ADDED
@@ -0,0 +1,21 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "size": {
3
+ "longest_edge": 16777216,
4
+ "shortest_edge": 65536
5
+ },
6
+ "patch_size": 16,
7
+ "temporal_patch_size": 2,
8
+ "merge_size": 2,
9
+ "image_mean": [
10
+ 0.5,
11
+ 0.5,
12
+ 0.5
13
+ ],
14
+ "image_std": [
15
+ 0.5,
16
+ 0.5,
17
+ 0.5
18
+ ],
19
+ "processor_class": "Qwen3VLProcessor",
20
+ "image_processor_type": "Qwen2VLImageProcessorFast"
21
+ }
processor_config.json ADDED
@@ -0,0 +1,50 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "image_processor": {
3
+ "do_convert_rgb": true,
4
+ "do_normalize": true,
5
+ "do_rescale": true,
6
+ "image_mean": [
7
+ 0.5,
8
+ 0.5,
9
+ 0.5
10
+ ],
11
+ "image_processor_type": "Qwen3VLImageProcessor",
12
+ "image_std": [
13
+ 0.5,
14
+ 0.5,
15
+ 0.5
16
+ ],
17
+ "max_pixels": 16777216,
18
+ "merge_size": 2,
19
+ "min_pixels": 65536,
20
+ "patch_size": 16,
21
+ "rescale_factor": 0.00392156862745098,
22
+ "temporal_patch_size": 2
23
+ },
24
+ "processor_class": "Qwen3VLProcessor",
25
+ "video_processor": {
26
+ "do_convert_rgb": true,
27
+ "do_normalize": true,
28
+ "do_rescale": true,
29
+ "fps": 2.0,
30
+ "image_mean": [
31
+ 0.5,
32
+ 0.5,
33
+ 0.5
34
+ ],
35
+ "image_std": [
36
+ 0.5,
37
+ 0.5,
38
+ 0.5
39
+ ],
40
+ "max_frames": 768,
41
+ "max_pixels": 25165824,
42
+ "merge_size": 2,
43
+ "min_frames": 4,
44
+ "min_pixels": 4096,
45
+ "patch_size": 16,
46
+ "rescale_factor": 0.00392156862745098,
47
+ "temporal_patch_size": 2,
48
+ "video_processor_type": "Qwen3VLVideoProcessor"
49
+ }
50
+ }
tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:87a7830d63fcf43bf241c3c5242e96e62dd3fdc29224ca26fed8ea333db72de4
3
+ size 19989343
tokenizer_config.json ADDED
@@ -0,0 +1,301 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_prefix_space": false,
3
+ "audio_bos_token": "<|audio_start|>",
4
+ "audio_eos_token": "<|audio_end|>",
5
+ "audio_token": "<|audio_pad|>",
6
+ "backend": "tokenizers",
7
+ "bos_token": null,
8
+ "clean_up_tokenization_spaces": false,
9
+ "eos_token": "<|im_end|>",
10
+ "errors": "replace",
11
+ "fix_mistral_regex": true,
12
+ "image_token": "<|image_pad|>",
13
+ "is_local": true,
14
+ "model_max_length": 262144,
15
+ "model_specific_special_tokens": {
16
+ "audio_bos_token": "<|audio_start|>",
17
+ "audio_eos_token": "<|audio_end|>",
18
+ "audio_token": "<|audio_pad|>",
19
+ "image_token": "<|image_pad|>",
20
+ "video_token": "<|video_pad|>",
21
+ "vision_bos_token": "<|vision_start|>",
22
+ "vision_eos_token": "<|vision_end|>"
23
+ },
24
+ "pad_token": "<|endoftext|>",
25
+ "padding_side": "left",
26
+ "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
27
+ "processor_class": "Qwen3VLProcessor",
28
+ "split_special_tokens": false,
29
+ "tokenizer_class": "TokenizersBackend",
30
+ "unk_token": null,
31
+ "video_token": "<|video_pad|>",
32
+ "vision_bos_token": "<|vision_start|>",
33
+ "vision_eos_token": "<|vision_end|>",
34
+ "added_tokens_decoder": {
35
+ "248044": {
36
+ "content": "<|endoftext|>",
37
+ "single_word": false,
38
+ "lstrip": false,
39
+ "rstrip": false,
40
+ "normalized": false,
41
+ "special": true
42
+ },
43
+ "248045": {
44
+ "content": "<|im_start|>",
45
+ "single_word": false,
46
+ "lstrip": false,
47
+ "rstrip": false,
48
+ "normalized": false,
49
+ "special": true
50
+ },
51
+ "248046": {
52
+ "content": "<|im_end|>",
53
+ "single_word": false,
54
+ "lstrip": false,
55
+ "rstrip": false,
56
+ "normalized": false,
57
+ "special": true
58
+ },
59
+ "248047": {
60
+ "content": "<|object_ref_start|>",
61
+ "single_word": false,
62
+ "lstrip": false,
63
+ "rstrip": false,
64
+ "normalized": false,
65
+ "special": true
66
+ },
67
+ "248048": {
68
+ "content": "<|object_ref_end|>",
69
+ "single_word": false,
70
+ "lstrip": false,
71
+ "rstrip": false,
72
+ "normalized": false,
73
+ "special": true
74
+ },
75
+ "248049": {
76
+ "content": "<|box_start|>",
77
+ "single_word": false,
78
+ "lstrip": false,
79
+ "rstrip": false,
80
+ "normalized": false,
81
+ "special": true
82
+ },
83
+ "248050": {
84
+ "content": "<|box_end|>",
85
+ "single_word": false,
86
+ "lstrip": false,
87
+ "rstrip": false,
88
+ "normalized": false,
89
+ "special": true
90
+ },
91
+ "248051": {
92
+ "content": "<|quad_start|>",
93
+ "single_word": false,
94
+ "lstrip": false,
95
+ "rstrip": false,
96
+ "normalized": false,
97
+ "special": true
98
+ },
99
+ "248052": {
100
+ "content": "<|quad_end|>",
101
+ "single_word": false,
102
+ "lstrip": false,
103
+ "rstrip": false,
104
+ "normalized": false,
105
+ "special": true
106
+ },
107
+ "248053": {
108
+ "content": "<|vision_start|>",
109
+ "single_word": false,
110
+ "lstrip": false,
111
+ "rstrip": false,
112
+ "normalized": false,
113
+ "special": true
114
+ },
115
+ "248054": {
116
+ "content": "<|vision_end|>",
117
+ "single_word": false,
118
+ "lstrip": false,
119
+ "rstrip": false,
120
+ "normalized": false,
121
+ "special": true
122
+ },
123
+ "248055": {
124
+ "content": "<|vision_pad|>",
125
+ "single_word": false,
126
+ "lstrip": false,
127
+ "rstrip": false,
128
+ "normalized": false,
129
+ "special": true
130
+ },
131
+ "248056": {
132
+ "content": "<|image_pad|>",
133
+ "single_word": false,
134
+ "lstrip": false,
135
+ "rstrip": false,
136
+ "normalized": false,
137
+ "special": true
138
+ },
139
+ "248057": {
140
+ "content": "<|video_pad|>",
141
+ "single_word": false,
142
+ "lstrip": false,
143
+ "rstrip": false,
144
+ "normalized": false,
145
+ "special": true
146
+ },
147
+ "248058": {
148
+ "content": "<tool_call>",
149
+ "single_word": false,
150
+ "lstrip": false,
151
+ "rstrip": false,
152
+ "normalized": false,
153
+ "special": false
154
+ },
155
+ "248059": {
156
+ "content": "</tool_call>",
157
+ "single_word": false,
158
+ "lstrip": false,
159
+ "rstrip": false,
160
+ "normalized": false,
161
+ "special": false
162
+ },
163
+ "248060": {
164
+ "content": "<|fim_prefix|>",
165
+ "single_word": false,
166
+ "lstrip": false,
167
+ "rstrip": false,
168
+ "normalized": false,
169
+ "special": false
170
+ },
171
+ "248061": {
172
+ "content": "<|fim_middle|>",
173
+ "single_word": false,
174
+ "lstrip": false,
175
+ "rstrip": false,
176
+ "normalized": false,
177
+ "special": false
178
+ },
179
+ "248062": {
180
+ "content": "<|fim_suffix|>",
181
+ "single_word": false,
182
+ "lstrip": false,
183
+ "rstrip": false,
184
+ "normalized": false,
185
+ "special": false
186
+ },
187
+ "248063": {
188
+ "content": "<|fim_pad|>",
189
+ "single_word": false,
190
+ "lstrip": false,
191
+ "rstrip": false,
192
+ "normalized": false,
193
+ "special": false
194
+ },
195
+ "248064": {
196
+ "content": "<|repo_name|>",
197
+ "single_word": false,
198
+ "lstrip": false,
199
+ "rstrip": false,
200
+ "normalized": false,
201
+ "special": false
202
+ },
203
+ "248065": {
204
+ "content": "<|file_sep|>",
205
+ "single_word": false,
206
+ "lstrip": false,
207
+ "rstrip": false,
208
+ "normalized": false,
209
+ "special": false
210
+ },
211
+ "248066": {
212
+ "content": "<tool_response>",
213
+ "single_word": false,
214
+ "lstrip": false,
215
+ "rstrip": false,
216
+ "normalized": false,
217
+ "special": false
218
+ },
219
+ "248067": {
220
+ "content": "</tool_response>",
221
+ "single_word": false,
222
+ "lstrip": false,
223
+ "rstrip": false,
224
+ "normalized": false,
225
+ "special": false
226
+ },
227
+ "248068": {
228
+ "content": "<think>",
229
+ "single_word": false,
230
+ "lstrip": false,
231
+ "rstrip": false,
232
+ "normalized": false,
233
+ "special": false
234
+ },
235
+ "248069": {
236
+ "content": "</think>",
237
+ "single_word": false,
238
+ "lstrip": false,
239
+ "rstrip": false,
240
+ "normalized": false,
241
+ "special": false
242
+ },
243
+ "248070": {
244
+ "content": "<|audio_start|>",
245
+ "single_word": false,
246
+ "lstrip": false,
247
+ "rstrip": false,
248
+ "normalized": false,
249
+ "special": true
250
+ },
251
+ "248071": {
252
+ "content": "<|audio_end|>",
253
+ "single_word": false,
254
+ "lstrip": false,
255
+ "rstrip": false,
256
+ "normalized": false,
257
+ "special": true
258
+ },
259
+ "248072": {
260
+ "content": "<tts_pad>",
261
+ "single_word": false,
262
+ "lstrip": false,
263
+ "rstrip": false,
264
+ "normalized": false,
265
+ "special": true
266
+ },
267
+ "248073": {
268
+ "content": "<tts_text_bos>",
269
+ "single_word": false,
270
+ "lstrip": false,
271
+ "rstrip": false,
272
+ "normalized": false,
273
+ "special": true
274
+ },
275
+ "248074": {
276
+ "content": "<tts_text_eod>",
277
+ "single_word": false,
278
+ "lstrip": false,
279
+ "rstrip": false,
280
+ "normalized": false,
281
+ "special": true
282
+ },
283
+ "248075": {
284
+ "content": "<tts_text_bos_single>",
285
+ "single_word": false,
286
+ "lstrip": false,
287
+ "rstrip": false,
288
+ "normalized": false,
289
+ "special": true
290
+ },
291
+ "248076": {
292
+ "content": "<|audio_pad|>",
293
+ "single_word": false,
294
+ "lstrip": false,
295
+ "rstrip": false,
296
+ "normalized": false,
297
+ "special": true
298
+ }
299
+ },
300
+ "chat_template": "{%- set image_count = namespace(value=0) %}\n{%- set video_count = namespace(value=0) %}\n{%- macro render_content(content, do_vision_count, is_system_content=false) %}\n {%- if content is string %}\n {{- content }}\n {%- elif content is iterable and content is not mapping %}\n {%- for item in content %}\n {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}\n {%- if is_system_content %}\n {{- raise_exception('System message cannot contain images.') }}\n {%- endif %}\n {%- if do_vision_count %}\n {%- set image_count.value = image_count.value + 1 %}\n {%- endif %}\n {%- if add_vision_id %}\n {{- 'Picture ' ~ image_count.value ~ ': ' }}\n {%- endif %}\n {{- '<|vision_start|><|image_pad|><|vision_end|>' }}\n {%- elif 'video' in item or item.type == 'video' %}\n {%- if is_system_content %}\n {{- raise_exception('System message cannot contain videos.') }}\n {%- endif %}\n {%- if do_vision_count %}\n {%- set video_count.value = video_count.value + 1 %}\n {%- endif %}\n {%- if add_vision_id %}\n {{- 'Video ' ~ video_count.value ~ ': ' }}\n {%- endif %}\n {{- '<|vision_start|><|video_pad|><|vision_end|>' }}\n {%- elif 'text' in item %}\n {{- item.text }}\n {%- else %}\n {{- raise_exception('Unexpected item type in content.') }}\n {%- endif %}\n {%- endfor %}\n {%- elif content is none or content is undefined %}\n {{- '' }}\n {%- else %}\n {{- raise_exception('Unexpected content type.') }}\n {%- endif %}\n{%- endmacro %}\n{%- if messages %}\n{%- if tools and tools is iterable and tools is not mapping %}\n {{- '<|im_start|>system\\n' }}\n {{- \"# Tools\\n\\nYou have access to the following functions:\\n\\n<tools>\" }}\n {%- for tool in tools %}\n {{- \"\\n\" }}\n {{- tool | tojson }}\n {%- endfor %}\n {{- \"\\n</tools>\" }}\n {{- '\\n\\nIf you choose to call a function ONLY reply in the following format with NO suffix:\\n\\n<tool_call>\\n<function=example_function_name>\\n<parameter=example_parameter_1>\\nvalue_1\\n</parameter>\\n<parameter=example_parameter_2>\\nThis is the value for the second parameter\\nthat can span\\nmultiple lines\\n</parameter>\\n</function>\\n</tool_call>\\n\\n<IMPORTANT>\\nReminder:\\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\\n- Required parameters MUST be specified\\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\\n</IMPORTANT>' }}\n {%- if messages[0].role == 'system' %}\n {%- set content = render_content(messages[0].content, false, true)|trim %}\n {%- if content %}\n {{- '\\n\\n' + content }}\n {%- endif %}\n {%- endif %}\n {{- '<|im_end|>\\n' }}\n{%- else %}\n {%- if messages[0].role == 'system' %}\n {%- set content = render_content(messages[0].content, false, true)|trim %}\n {{- '<|im_start|>system\\n' + content + '<|im_end|>\\n' }}\n {%- endif %}\n{%- endif %}\n{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}\n{%- for message in messages[::-1] %}\n {%- set index = (messages|length - 1) - loop.index0 %}\n {%- if ns.multi_step_tool and message.role == \"user\" %}\n {%- set content = render_content(message.content, false)|trim %}\n {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}\n {%- set ns.multi_step_tool = false %}\n {%- set ns.last_query_index = index %}\n {%- endif %}\n {%- endif %}\n{%- endfor %}\n{%- for message in messages %}\n {%- set content = render_content(message.content, true)|trim %}\n {%- if message.role == \"system\" %}\n {%- if not loop.first %}\n {{- raise_exception('System message must be at the beginning.') }}\n {%- endif %}\n {%- elif message.role == \"user\" %}\n {{- '<|im_start|>' + message.role + '\\n' + content + '<|im_end|>' + '\\n' }}\n {%- elif message.role == \"assistant\" %}\n {%- set reasoning_content = '' %}\n {%- if message.reasoning_content is string %}\n {%- set reasoning_content = message.reasoning_content %}\n {%- else %}\n {%- if '</think>' in content %}\n {%- set reasoning_content = content.split('</think>')[0].rstrip('\\n').split('<think>')[-1].lstrip('\\n') %}\n {%- set content = content.split('</think>')[-1].lstrip('\\n') %}\n {%- endif %}\n {%- endif %}\n {%- set reasoning_content = reasoning_content|trim %}\n {%- if (preserve_thinking is defined and preserve_thinking is true) or (loop.index0 > ns.last_query_index) %}\n {{- '<|im_start|>' + message.role + '\\n<think>\\n' + reasoning_content + '\\n</think>\\n\\n' + content }}\n {%- else %}\n {{- '<|im_start|>' + message.role + '\\n' + content }}\n {%- endif %}\n {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}\n {%- for tool_call in message.tool_calls %}\n {%- if tool_call.function is defined %}\n {%- set tool_call = tool_call.function %}\n {%- endif %}\n {%- if loop.first %}\n {%- if content|trim %}\n {{- '\\n\\n' }}\n {%- endif %}\n {%- else %}\n {{- '\\n' }}\n {%- endif %}\n {%- if tool_call.arguments is mapping %}\n {{- '<tool_call>\\n<function=' + tool_call.name + '>\\n' }}\n {%- for args_name, args_value in tool_call.arguments|items %}\n {{- '<parameter=' + args_name + '>\\n' }}\n {%- set args_value = args_value | string if args_value is string else args_value | tojson | safe %}\n {{- args_value }}\n {{- '\\n</parameter>\\n' }}\n {%- endfor %}\n {{- '</function>\\n</tool_call>' }}\n {%- else %}\n {{- '<tool_call>\\n{\"name\": ' + tool_call.name|tojson + ', \"arguments\": ' }}\n {%- if tool_call.arguments is string %}\n {{- tool_call.arguments }}\n {%- else %}\n {{- tool_call.arguments | tojson | safe }}\n {%- endif %}\n {{- '}\\n</tool_call>' }}\n {%- endif %}\n {%- endfor %}\n {%- endif %}\n {{- '<|im_end|>\\n' }}\n {%- elif message.role == \"tool\" %}\n {%- if loop.previtem and loop.previtem.role != \"tool\" %}\n {{- '<|im_start|>user' }}\n {%- endif %}\n {{- '\\n<tool_response>\\n' }}\n {{- content }}\n {{- '\\n</tool_response>' }}\n {%- if not loop.last and loop.nextitem.role != \"tool\" %}\n {{- '<|im_end|>\\n' }}\n {%- elif loop.last %}\n {{- '<|im_end|>\\n' }}\n {%- endif %}\n {%- else %}\n {{- raise_exception('Unexpected message role.') }}\n {%- endif %}\n{%- endfor %}\n{%- if add_generation_prompt %}\n {{- '<|im_start|>assistant\\n' }}\n {%- if enable_thinking is defined and enable_thinking is true %}\n {{- '<think>\\n' }}\n {%- else %}\n {{- '<think>\\n\\n</think>\\n\\n' }}\n {%- endif %}\n{%- endif %}\n{%- endif %}\n"
301
+ }
video_preprocessor_config.json ADDED
@@ -0,0 +1,21 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "size": {
3
+ "longest_edge": 25165824,
4
+ "shortest_edge": 4096
5
+ },
6
+ "patch_size": 16,
7
+ "temporal_patch_size": 2,
8
+ "merge_size": 2,
9
+ "image_mean": [
10
+ 0.5,
11
+ 0.5,
12
+ 0.5
13
+ ],
14
+ "image_std": [
15
+ 0.5,
16
+ 0.5,
17
+ 0.5
18
+ ],
19
+ "processor_class": "Qwen3VLProcessor",
20
+ "video_processor_type": "Qwen3VLVideoProcessor"
21
+ }