Hugging Face's logo

nightmedia
/
unsloth-Qwen3-Coder-30B-A3B-Instruct-qx4-mlx

Text Generation
MLX
Safetensors
qwen3_moe
unsloth
conversational
4-bit precision
Model card Files
xet
Community
nightmedia commited on
Commit
ea089a7
·
verified ·
1 Parent(s): 2438f9a

Add files using upload-large-folder tool

Browse files
Files changed (17) hide show
  1. .gitattributes +1 -0
  2. README.md +38 -0
  3. added_tokens.json +28 -0
  4. chat_template.jinja +135 -0
  5. config.json +3138 -0
  6. generation_config.json +12 -0
  7. merges.txt +0 -0
  8. model-00001-of-00004.safetensors +3 -0
  9. model-00002-of-00004.safetensors +3 -0
  10. model-00003-of-00004.safetensors +3 -0
  11. model-00004-of-00004.safetensors +3 -0
  12. model.safetensors.index.json +0 -0
  13. qwen3coder_tool_parser.py +675 -0
  14. special_tokens_map.json +31 -0
  15. tokenizer.json +3 -0
  16. tokenizer_config.json +240 -0
  17. vocab.json +0 -0
.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 +1,38 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ tags:
3
+ - unsloth
4
+ - mlx
5
+ base_model: unsloth/Qwen3-Coder-30B-A3B-Instruct
6
+ library_name: mlx
7
+ license: apache-2.0
8
+ license_link: https://huggingface.co/Qwen/Qwen3-Coder-30B-A3B-Instruct/blob/main/LICENSE
9
+ pipeline_tag: text-generation
10
+ ---
11
+
12
+ # unsloth-Qwen3-Coder-30B-A3B-Instruct-qx4-mlx
13
+
14
+ This model [unsloth-Qwen3-Coder-30B-A3B-Instruct-qx4-mlx](https://huggingface.co/unsloth-Qwen3-Coder-30B-A3B-Instruct-qx4-mlx) was
15
+ converted to MLX format from [unsloth/Qwen3-Coder-30B-A3B-Instruct](https://huggingface.co/unsloth/Qwen3-Coder-30B-A3B-Instruct)
16
+ using mlx-lm version **0.26.3**.
17
+
18
+ ## Use with mlx
19
+
20
+ ```bash
21
+ pip install mlx-lm
22
+ ```
23
+
24
+ ```python
25
+ from mlx_lm import load, generate
26
+
27
+ model, tokenizer = load("unsloth-Qwen3-Coder-30B-A3B-Instruct-qx4-mlx")
28
+
29
+ prompt = "hello"
30
+
31
+ if tokenizer.chat_template is not None:
32
+ messages = [{"role": "user", "content": prompt}]
33
+ prompt = tokenizer.apply_chat_template(
34
+ messages, add_generation_prompt=True
35
+ )
36
+
37
+ response = generate(model, tokenizer, prompt=prompt, verbose=True)
38
+ ```
added_tokens.json ADDED
@@ -0,0 +1,28 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "</think>": 151668,
3
+ "</tool_call>": 151658,
4
+ "</tool_response>": 151666,
5
+ "<think>": 151667,
6
+ "<tool_call>": 151657,
7
+ "<tool_response>": 151665,
8
+ "<|box_end|>": 151649,
9
+ "<|box_start|>": 151648,
10
+ "<|endoftext|>": 151643,
11
+ "<|file_sep|>": 151664,
12
+ "<|fim_middle|>": 151660,
13
+ "<|fim_pad|>": 151662,
14
+ "<|fim_prefix|>": 151659,
15
+ "<|fim_suffix|>": 151661,
16
+ "<|im_end|>": 151645,
17
+ "<|im_start|>": 151644,
18
+ "<|image_pad|>": 151655,
19
+ "<|object_ref_end|>": 151647,
20
+ "<|object_ref_start|>": 151646,
21
+ "<|quad_end|>": 151651,
22
+ "<|quad_start|>": 151650,
23
+ "<|repo_name|>": 151663,
24
+ "<|video_pad|>": 151656,
25
+ "<|vision_end|>": 151653,
26
+ "<|vision_pad|>": 151654,
27
+ "<|vision_start|>": 151652
28
+ }
chat_template.jinja ADDED
@@ -0,0 +1,135 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {# Copyright 2025-present Unsloth. Apache 2.0 License. Unsloth Chat template fixes #}
2
+ {% macro render_item_list(item_list, tag_name='required') %}
3
+ {%- if item_list is defined and item_list is iterable and item_list | length > 0 %}
4
+ {%- if tag_name %}{{- '\n<' ~ tag_name ~ '>' -}}{% endif %}
5
+ {{- '[' }}
6
+ {%- for item in item_list -%}
7
+ {%- if loop.index > 1 %}{{- ", "}}{% endif -%}
8
+ {%- if item is string -%}
9
+ {{ "`" ~ item ~ "`" }}
10
+ {%- else -%}
11
+ {{ item }}
12
+ {%- endif -%}
13
+ {%- endfor -%}
14
+ {{- ']' }}
15
+ {%- if tag_name %}{{- '</' ~ tag_name ~ '>' -}}{% endif %}
16
+ {%- endif %}
17
+ {% endmacro %}
18
+
19
+ {%- if messages[0]["role"] == "system" %}
20
+ {%- set system_message = messages[0]["content"] %}
21
+ {%- set loop_messages = messages[1:] %}
22
+ {%- else %}
23
+ {%- set loop_messages = messages %}
24
+ {%- endif %}
25
+
26
+ {%- if not tools is defined %}
27
+ {%- set tools = [] %}
28
+ {%- endif %}
29
+
30
+ {%- if system_message is defined %}
31
+ {{- "<|im_start|>system\n" + system_message }}
32
+ {%- else %}
33
+ {%- if tools is iterable and tools | length > 0 %}
34
+ {{- "<|im_start|>system\nYou are Qwen, a helpful AI assistant that can interact with a computer to solve tasks." }}
35
+ {%- endif %}
36
+ {%- endif %}
37
+ {%- if tools is iterable and tools | length > 0 %}
38
+ {{- "\n\nYou have access to the following functions:\n\n" }}
39
+ {{- "<tools>" }}
40
+ {%- for tool in tools %}
41
+ {%- if tool.function is defined %}
42
+ {%- set tool = tool.function %}
43
+ {%- endif %}
44
+ {{- "\n<function>\n<name>" ~ tool.name ~ "</name>" }}
45
+ {{- '\n<description>' ~ (tool.description | trim) ~ '</description>' }}
46
+ {{- '\n<parameters>' }}
47
+ {%- for param_name, param_fields in tool.parameters.properties|items %}
48
+ {{- '\n<parameter>' }}
49
+ {{- '\n<name>' ~ param_name ~ '</name>' }}
50
+ {%- if param_fields.type is defined %}
51
+ {{- '\n<type>' ~ (param_fields.type | string) ~ '</type>' }}
52
+ {%- endif %}
53
+ {%- if param_fields.description is defined %}
54
+ {{- '\n<description>' ~ (param_fields.description | trim) ~ '</description>' }}
55
+ {%- endif %}
56
+ {{- render_item_list(param_fields.enum, 'enum') }}
57
+ {%- set handled_keys = ['type', 'description', 'enum', 'required'] %}
58
+ {%- for json_key, json_value in param_fields|items %}
59
+ {%- if json_key not in handled_keys %}
60
+ {%- set normed_json_key = json_key|string %}
61
+ {%- if json_value is mapping %}
62
+ {{- '\n<' ~ normed_json_key ~ '>' ~ (json_value | tojson | safe) ~ '</' ~ normed_json_key ~ '>' }}
63
+ {%- else %}
64
+ {{- '\n<' ~ normed_json_key ~ '>' ~ (json_value | string) ~ '</' ~ normed_json_key ~ '>' }}
65
+ {%- endif %}
66
+ {%- endif %}
67
+ {%- endfor %}
68
+ {{- render_item_list(param_fields.required, 'required') }}
69
+ {{- '\n</parameter>' }}
70
+ {%- endfor %}
71
+ {{- render_item_list(tool.parameters.required, 'required') }}
72
+ {{- '\n</parameters>' }}
73
+ {%- if tool.return is defined %}
74
+ {%- if tool.return is mapping %}
75
+ {{- '\n<return>' ~ (tool.return | tojson | safe) ~ '</return>' }}
76
+ {%- else %}
77
+ {{- '\n<return>' ~ (tool.return | string) ~ '</return>' }}
78
+ {%- endif %}
79
+ {%- endif %}
80
+ {{- '\n</function>' }}
81
+ {%- endfor %}
82
+ {{- "\n</tools>" }}
83
+ {{- '\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>' }}
84
+ {%- endif %}
85
+ {%- if system_message is defined %}
86
+ {{- '<|im_end|>\n' }}
87
+ {%- else %}
88
+ {%- if tools is iterable and tools | length > 0 %}
89
+ {{- '<|im_end|>\n' }}
90
+ {%- endif %}
91
+ {%- endif %}
92
+ {%- for message in loop_messages %}
93
+ {%- if message.role == "assistant" and message.tool_calls is defined and message.tool_calls is iterable and message.tool_calls | length > 0 %}
94
+ {{- '<|im_start|>' + message.role }}
95
+ {%- if message.content is defined and message.content is string and message.content | trim | length > 0 %}
96
+ {{- '\n' + message.content | trim + '\n' }}
97
+ {%- endif %}
98
+ {%- for tool_call in message.tool_calls %}
99
+ {%- if tool_call.function is defined %}
100
+ {%- set tool_call = tool_call.function %}
101
+ {%- endif %}
102
+ {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
103
+ {%- if tool_call.arguments is defined %}
104
+ {%- for args_name, args_value in tool_call.arguments|items %}
105
+ {{- '<parameter=' + args_name + '>\n' }}
106
+ {%- set args_value = args_value if args_value is string else args_value | string %}
107
+ {{- args_value }}
108
+ {{- '\n</parameter>\n' }}
109
+ {%- endfor %}
110
+ {%- endif %}
111
+ {{- '</function>\n</tool_call>' }}
112
+ {%- endfor %}
113
+ {{- '<|im_end|>\n' }}
114
+ {%- elif message.role == "user" or message.role == "system" or message.role == "assistant" %}
115
+ {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
116
+ {%- elif message.role == "tool" %}
117
+ {%- if loop.previtem and loop.previtem.role != "tool" %}
118
+ {{- '<|im_start|>user\n' }}
119
+ {%- endif %}
120
+ {{- '<tool_response>\n' }}
121
+ {{- message.content }}
122
+ {{- '\n</tool_response>\n' }}
123
+ {%- if not loop.last and loop.nextitem.role != "tool" %}
124
+ {{- '<|im_end|>\n' }}
125
+ {%- elif loop.last %}
126
+ {{- '<|im_end|>\n' }}
127
+ {%- endif %}
128
+ {%- else %}
129
+ {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>\n' }}
130
+ {%- endif %}
131
+ {%- endfor %}
132
+ {%- if add_generation_prompt %}
133
+ {{- '<|im_start|>assistant\n' }}
134
+ {%- endif %}
135
+ {# Copyright 2025-present Unsloth. Apache 2.0 License. Unsloth Chat template fixes #}
config.json ADDED
@@ -0,0 +1,3138 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "Qwen3MoeForCausalLM"
4
+ ],
5
+ "attention_bias": false,
6
+ "attention_dropout": 0.0,
7
+ "decoder_sparse_step": 1,
8
+ "eos_token_id": 151645,
9
+ "head_dim": 128,
10
+ "hidden_act": "silu",
11
+ "hidden_size": 2048,
12
+ "initializer_range": 0.02,
13
+ "intermediate_size": 5472,
14
+ "max_position_embeddings": 262144,
15
+ "max_window_layers": 28,
16
+ "mlp_only_layers": [],
17
+ "model_type": "qwen3_moe",
18
+ "moe_intermediate_size": 768,
19
+ "norm_topk_prob": true,
20
+ "num_attention_heads": 32,
21
+ "num_experts": 128,
22
+ "num_experts_per_tok": 8,
23
+ "num_hidden_layers": 48,
24
+ "num_key_value_heads": 4,
25
+ "output_router_logits": false,
26
+ "pad_token_id": 151654,
27
+ "qkv_bias": false,
28
+ "quantization": {
29
+ "group_size": 64,
30
+ "bits": 4,
31
+ "model.embed_tokens": {
32
+ "group_size": 64,
33
+ "bits": 4
34
+ },
35
+ "model.layers.0.self_attn.q_proj": {
36
+ "group_size": 32,
37
+ "bits": 8
38
+ },
39
+ "model.layers.0.self_attn.k_proj": {
40
+ "group_size": 32,
41
+ "bits": 8
42
+ },
43
+ "model.layers.0.self_attn.v_proj": {
44
+ "group_size": 32,
45
+ "bits": 8
46
+ },
47
+ "model.layers.0.self_attn.o_proj": {
48
+ "group_size": 32,
49
+ "bits": 8
50
+ },
51
+ "model.layers.0.mlp.gate": {
52
+ "group_size": 32,
53
+ "bits": 8
54
+ },
55
+ "model.layers.0.mlp.switch_mlp.gate_proj": {
56
+ "group_size": 32,
57
+ "bits": 8
58
+ },
59
+ "model.layers.0.mlp.switch_mlp.up_proj": {
60
+ "group_size": 32,
61
+ "bits": 8
62
+ },
63
+ "model.layers.0.mlp.switch_mlp.down_proj": {
64
+ "group_size": 32,
65
+ "bits": 8
66
+ },
67
+ "model.layers.1.self_attn.q_proj": {
68
+ "group_size": 64,
69
+ "bits": 4
70
+ },
71
+ "model.layers.1.self_attn.k_proj": {
72
+ "group_size": 64,
73
+ "bits": 4
74
+ },
75
+ "model.layers.1.self_attn.v_proj": {
76
+ "group_size": 32,
77
+ "bits": 8
78
+ },
79
+ "model.layers.1.self_attn.o_proj": {
80
+ "group_size": 64,
81
+ "bits": 4
82
+ },
83
+ "model.layers.1.mlp.gate": {
84
+ "group_size": 64,
85
+ "bits": 4
86
+ },
87
+ "model.layers.1.mlp.switch_mlp.gate_proj": {
88
+ "group_size": 64,
89
+ "bits": 4
90
+ },
91
+ "model.layers.1.mlp.switch_mlp.up_proj": {
92
+ "group_size": 64,
93
+ "bits": 4
94
+ },
95
+ "model.layers.1.mlp.switch_mlp.down_proj": {
96
+ "group_size": 32,
97
+ "bits": 6
98
+ },
99
+ "model.layers.2.self_attn.q_proj": {
100
+ "group_size": 64,
101
+ "bits": 4
102
+ },
103
+ "model.layers.2.self_attn.k_proj": {
104
+ "group_size": 64,
105
+ "bits": 4
106
+ },
107
+ "model.layers.2.self_attn.v_proj": {
108
+ "group_size": 32,
109
+ "bits": 8
110
+ },
111
+ "model.layers.2.self_attn.o_proj": {
112
+ "group_size": 64,
113
+ "bits": 4
114
+ },
115
+ "model.layers.2.mlp.gate": {
116
+ "group_size": 64,
117
+ "bits": 4
118
+ },
119
+ "model.layers.2.mlp.switch_mlp.gate_proj": {
120
+ "group_size": 64,
121
+ "bits": 4
122
+ },
123
+ "model.layers.2.mlp.switch_mlp.up_proj": {
124
+ "group_size": 64,
125
+ "bits": 4
126
+ },
127
+ "model.layers.2.mlp.switch_mlp.down_proj": {
128
+ "group_size": 32,
129
+ "bits": 6
130
+ },
131
+ "model.layers.3.self_attn.q_proj": {
132
+ "group_size": 64,
133
+ "bits": 4
134
+ },
135
+ "model.layers.3.self_attn.k_proj": {
136
+ "group_size": 64,
137
+ "bits": 4
138
+ },
139
+ "model.layers.3.self_attn.v_proj": {
140
+ "group_size": 32,
141
+ "bits": 8
142
+ },
143
+ "model.layers.3.self_attn.o_proj": {
144
+ "group_size": 64,
145
+ "bits": 4
146
+ },
147
+ "model.layers.3.mlp.gate": {
148
+ "group_size": 64,
149
+ "bits": 4
150
+ },
151
+ "model.layers.3.mlp.switch_mlp.gate_proj": {
152
+ "group_size": 64,
153
+ "bits": 4
154
+ },
155
+ "model.layers.3.mlp.switch_mlp.up_proj": {
156
+ "group_size": 64,
157
+ "bits": 4
158
+ },
159
+ "model.layers.3.mlp.switch_mlp.down_proj": {
160
+ "group_size": 32,
161
+ "bits": 6
162
+ },
163
+ "model.layers.4.self_attn.q_proj": {
164
+ "group_size": 64,
165
+ "bits": 4
166
+ },
167
+ "model.layers.4.self_attn.k_proj": {
168
+ "group_size": 64,
169
+ "bits": 4
170
+ },
171
+ "model.layers.4.self_attn.v_proj": {
172
+ "group_size": 32,
173
+ "bits": 8
174
+ },
175
+ "model.layers.4.self_attn.o_proj": {
176
+ "group_size": 64,
177
+ "bits": 4
178
+ },
179
+ "model.layers.4.mlp.gate": {
180
+ "group_size": 64,
181
+ "bits": 4
182
+ },
183
+ "model.layers.4.mlp.switch_mlp.gate_proj": {
184
+ "group_size": 64,
185
+ "bits": 4
186
+ },
187
+ "model.layers.4.mlp.switch_mlp.up_proj": {
188
+ "group_size": 64,
189
+ "bits": 4
190
+ },
191
+ "model.layers.4.mlp.switch_mlp.down_proj": {
192
+ "group_size": 32,
193
+ "bits": 6
194
+ },
195
+ "model.layers.5.self_attn.q_proj": {
196
+ "group_size": 64,
197
+ "bits": 4
198
+ },
199
+ "model.layers.5.self_attn.k_proj": {
200
+ "group_size": 64,
201
+ "bits": 4
202
+ },
203
+ "model.layers.5.self_attn.v_proj": {
204
+ "group_size": 32,
205
+ "bits": 8
206
+ },
207
+ "model.layers.5.self_attn.o_proj": {
208
+ "group_size": 64,
209
+ "bits": 4
210
+ },
211
+ "model.layers.5.mlp.gate": {
212
+ "group_size": 64,
213
+ "bits": 4
214
+ },
215
+ "model.layers.5.mlp.switch_mlp.gate_proj": {
216
+ "group_size": 64,
217
+ "bits": 4
218
+ },
219
+ "model.layers.5.mlp.switch_mlp.up_proj": {
220
+ "group_size": 64,
221
+ "bits": 4
222
+ },
223
+ "model.layers.5.mlp.switch_mlp.down_proj": {
224
+ "group_size": 32,
225
+ "bits": 6
226
+ },
227
+ "model.layers.6.self_attn.q_proj": {
228
+ "group_size": 64,
229
+ "bits": 4
230
+ },
231
+ "model.layers.6.self_attn.k_proj": {
232
+ "group_size": 64,
233
+ "bits": 4
234
+ },
235
+ "model.layers.6.self_attn.v_proj": {
236
+ "group_size": 64,
237
+ "bits": 4
238
+ },
239
+ "model.layers.6.self_attn.o_proj": {
240
+ "group_size": 64,
241
+ "bits": 4
242
+ },
243
+ "model.layers.6.mlp.gate": {
244
+ "group_size": 64,
245
+ "bits": 4
246
+ },
247
+ "model.layers.6.mlp.switch_mlp.gate_proj": {
248
+ "group_size": 64,
249
+ "bits": 4
250
+ },
251
+ "model.layers.6.mlp.switch_mlp.up_proj": {
252
+ "group_size": 64,
253
+ "bits": 4
254
+ },
255
+ "model.layers.6.mlp.switch_mlp.down_proj": {
256
+ "group_size": 64,
257
+ "bits": 4
258
+ },
259
+ "model.layers.7.self_attn.q_proj": {
260
+ "group_size": 64,
261
+ "bits": 4
262
+ },
263
+ "model.layers.7.self_attn.k_proj": {
264
+ "group_size": 64,
265
+ "bits": 4
266
+ },
267
+ "model.layers.7.self_attn.v_proj": {
268
+ "group_size": 64,
269
+ "bits": 4
270
+ },
271
+ "model.layers.7.self_attn.o_proj": {
272
+ "group_size": 64,
273
+ "bits": 4
274
+ },
275
+ "model.layers.7.mlp.gate": {
276
+ "group_size": 64,
277
+ "bits": 4
278
+ },
279
+ "model.layers.7.mlp.switch_mlp.gate_proj": {
280
+ "group_size": 64,
281
+ "bits": 4
282
+ },
283
+ "model.layers.7.mlp.switch_mlp.up_proj": {
284
+ "group_size": 64,
285
+ "bits": 4
286
+ },
287
+ "model.layers.7.mlp.switch_mlp.down_proj": {
288
+ "group_size": 64,
289
+ "bits": 4
290
+ },
291
+ "model.layers.8.self_attn.q_proj": {
292
+ "group_size": 64,
293
+ "bits": 4
294
+ },
295
+ "model.layers.8.self_attn.k_proj": {
296
+ "group_size": 64,
297
+ "bits": 4
298
+ },
299
+ "model.layers.8.self_attn.v_proj": {
300
+ "group_size": 32,
301
+ "bits": 6
302
+ },
303
+ "model.layers.8.self_attn.o_proj": {
304
+ "group_size": 64,
305
+ "bits": 4
306
+ },
307
+ "model.layers.8.mlp.gate": {
308
+ "group_size": 64,
309
+ "bits": 4
310
+ },
311
+ "model.layers.8.mlp.switch_mlp.gate_proj": {
312
+ "group_size": 64,
313
+ "bits": 4
314
+ },
315
+ "model.layers.8.mlp.switch_mlp.up_proj": {
316
+ "group_size": 64,
317
+ "bits": 4
318
+ },
319
+ "model.layers.8.mlp.switch_mlp.down_proj": {
320
+ "group_size": 32,
321
+ "bits": 6
322
+ },
323
+ "model.layers.9.self_attn.q_proj": {
324
+ "group_size": 64,
325
+ "bits": 4
326
+ },
327
+ "model.layers.9.self_attn.k_proj": {
328
+ "group_size": 64,
329
+ "bits": 4
330
+ },
331
+ "model.layers.9.self_attn.v_proj": {
332
+ "group_size": 64,
333
+ "bits": 4
334
+ },
335
+ "model.layers.9.self_attn.o_proj": {
336
+ "group_size": 64,
337
+ "bits": 4
338
+ },
339
+ "model.layers.9.mlp.gate": {
340
+ "group_size": 64,
341
+ "bits": 4
342
+ },
343
+ "model.layers.9.mlp.switch_mlp.gate_proj": {
344
+ "group_size": 64,
345
+ "bits": 4
346
+ },
347
+ "model.layers.9.mlp.switch_mlp.up_proj": {
348
+ "group_size": 64,
349
+ "bits": 4
350
+ },
351
+ "model.layers.9.mlp.switch_mlp.down_proj": {
352
+ "group_size": 64,
353
+ "bits": 4
354
+ },
355
+ "model.layers.10.self_attn.q_proj": {
356
+ "group_size": 64,
357
+ "bits": 4
358
+ },
359
+ "model.layers.10.self_attn.k_proj": {
360
+ "group_size": 64,
361
+ "bits": 4
362
+ },
363
+ "model.layers.10.self_attn.v_proj": {
364
+ "group_size": 64,
365
+ "bits": 4
366
+ },
367
+ "model.layers.10.self_attn.o_proj": {
368
+ "group_size": 64,
369
+ "bits": 4
370
+ },
371
+ "model.layers.10.mlp.gate": {
372
+ "group_size": 64,
373
+ "bits": 4
374
+ },
375
+ "model.layers.10.mlp.switch_mlp.gate_proj": {
376
+ "group_size": 64,
377
+ "bits": 4
378
+ },
379
+ "model.layers.10.mlp.switch_mlp.up_proj": {
380
+ "group_size": 64,
381
+ "bits": 4
382
+ },
383
+ "model.layers.10.mlp.switch_mlp.down_proj": {
384
+ "group_size": 64,
385
+ "bits": 4
386
+ },
387
+ "model.layers.11.self_attn.q_proj": {
388
+ "group_size": 64,
389
+ "bits": 4
390
+ },
391
+ "model.layers.11.self_attn.k_proj": {
392
+ "group_size": 64,
393
+ "bits": 4
394
+ },
395
+ "model.layers.11.self_attn.v_proj": {
396
+ "group_size": 32,
397
+ "bits": 6
398
+ },
399
+ "model.layers.11.self_attn.o_proj": {
400
+ "group_size": 64,
401
+ "bits": 4
402
+ },
403
+ "model.layers.11.mlp.gate": {
404
+ "group_size": 64,
405
+ "bits": 4
406
+ },
407
+ "model.layers.11.mlp.switch_mlp.gate_proj": {
408
+ "group_size": 64,
409
+ "bits": 4
410
+ },
411
+ "model.layers.11.mlp.switch_mlp.up_proj": {
412
+ "group_size": 64,
413
+ "bits": 4
414
+ },
415
+ "model.layers.11.mlp.switch_mlp.down_proj": {
416
+ "group_size": 32,
417
+ "bits": 6
418
+ },
419
+ "model.layers.12.self_attn.q_proj": {
420
+ "group_size": 64,
421
+ "bits": 4
422
+ },
423
+ "model.layers.12.self_attn.k_proj": {
424
+ "group_size": 64,
425
+ "bits": 4
426
+ },
427
+ "model.layers.12.self_attn.v_proj": {
428
+ "group_size": 64,
429
+ "bits": 4
430
+ },
431
+ "model.layers.12.self_attn.o_proj": {
432
+ "group_size": 64,
433
+ "bits": 4
434
+ },
435
+ "model.layers.12.mlp.gate": {
436
+ "group_size": 64,
437
+ "bits": 4
438
+ },
439
+ "model.layers.12.mlp.switch_mlp.gate_proj": {
440
+ "group_size": 64,
441
+ "bits": 4
442
+ },
443
+ "model.layers.12.mlp.switch_mlp.up_proj": {
444
+ "group_size": 64,
445
+ "bits": 4
446
+ },
447
+ "model.layers.12.mlp.switch_mlp.down_proj": {
448
+ "group_size": 64,
449
+ "bits": 4
450
+ },
451
+ "model.layers.13.self_attn.q_proj": {
452
+ "group_size": 64,
453
+ "bits": 4
454
+ },
455
+ "model.layers.13.self_attn.k_proj": {
456
+ "group_size": 64,
457
+ "bits": 4
458
+ },
459
+ "model.layers.13.self_attn.v_proj": {
460
+ "group_size": 64,
461
+ "bits": 4
462
+ },
463
+ "model.layers.13.self_attn.o_proj": {
464
+ "group_size": 64,
465
+ "bits": 4
466
+ },
467
+ "model.layers.13.mlp.gate": {
468
+ "group_size": 64,
469
+ "bits": 4
470
+ },
471
+ "model.layers.13.mlp.switch_mlp.gate_proj": {
472
+ "group_size": 64,
473
+ "bits": 4
474
+ },
475
+ "model.layers.13.mlp.switch_mlp.up_proj": {
476
+ "group_size": 64,
477
+ "bits": 4
478
+ },
479
+ "model.layers.13.mlp.switch_mlp.down_proj": {
480
+ "group_size": 64,
481
+ "bits": 4
482
+ },
483
+ "model.layers.14.self_attn.q_proj": {
484
+ "group_size": 64,
485
+ "bits": 4
486
+ },
487
+ "model.layers.14.self_attn.k_proj": {
488
+ "group_size": 64,
489
+ "bits": 4
490
+ },
491
+ "model.layers.14.self_attn.v_proj": {
492
+ "group_size": 32,
493
+ "bits": 6
494
+ },
495
+ "model.layers.14.self_attn.o_proj": {
496
+ "group_size": 64,
497
+ "bits": 4
498
+ },
499
+ "model.layers.14.mlp.gate": {
500
+ "group_size": 64,
501
+ "bits": 4
502
+ },
503
+ "model.layers.14.mlp.switch_mlp.gate_proj": {
504
+ "group_size": 64,
505
+ "bits": 4
506
+ },
507
+ "model.layers.14.mlp.switch_mlp.up_proj": {
508
+ "group_size": 64,
509
+ "bits": 4
510
+ },
511
+ "model.layers.14.mlp.switch_mlp.down_proj": {
512
+ "group_size": 32,
513
+ "bits": 6
514
+ },
515
+ "model.layers.15.self_attn.q_proj": {
516
+ "group_size": 64,
517
+ "bits": 4
518
+ },
519
+ "model.layers.15.self_attn.k_proj": {
520
+ "group_size": 64,
521
+ "bits": 4
522
+ },
523
+ "model.layers.15.self_attn.v_proj": {
524
+ "group_size": 64,
525
+ "bits": 4
526
+ },
527
+ "model.layers.15.self_attn.o_proj": {
528
+ "group_size": 64,
529
+ "bits": 4
530
+ },
531
+ "model.layers.15.mlp.gate": {
532
+ "group_size": 64,
533
+ "bits": 4
534
+ },
535
+ "model.layers.15.mlp.switch_mlp.gate_proj": {
536
+ "group_size": 64,
537
+ "bits": 4
538
+ },
539
+ "model.layers.15.mlp.switch_mlp.up_proj": {
540
+ "group_size": 64,
541
+ "bits": 4
542
+ },
543
+ "model.layers.15.mlp.switch_mlp.down_proj": {
544
+ "group_size": 64,
545
+ "bits": 4
546
+ },
547
+ "model.layers.16.self_attn.q_proj": {
548
+ "group_size": 64,
549
+ "bits": 4
550
+ },
551
+ "model.layers.16.self_attn.k_proj": {
552
+ "group_size": 64,
553
+ "bits": 4
554
+ },
555
+ "model.layers.16.self_attn.v_proj": {
556
+ "group_size": 64,
557
+ "bits": 4
558
+ },
559
+ "model.layers.16.self_attn.o_proj": {
560
+ "group_size": 64,
561
+ "bits": 4
562
+ },
563
+ "model.layers.16.mlp.gate": {
564
+ "group_size": 64,
565
+ "bits": 4
566
+ },
567
+ "model.layers.16.mlp.switch_mlp.gate_proj": {
568
+ "group_size": 64,
569
+ "bits": 4
570
+ },
571
+ "model.layers.16.mlp.switch_mlp.up_proj": {
572
+ "group_size": 64,
573
+ "bits": 4
574
+ },
575
+ "model.layers.16.mlp.switch_mlp.down_proj": {
576
+ "group_size": 64,
577
+ "bits": 4
578
+ },
579
+ "model.layers.17.self_attn.q_proj": {
580
+ "group_size": 64,
581
+ "bits": 4
582
+ },
583
+ "model.layers.17.self_attn.k_proj": {
584
+ "group_size": 64,
585
+ "bits": 4
586
+ },
587
+ "model.layers.17.self_attn.v_proj": {
588
+ "group_size": 32,
589
+ "bits": 6
590
+ },
591
+ "model.layers.17.self_attn.o_proj": {
592
+ "group_size": 64,
593
+ "bits": 4
594
+ },
595
+ "model.layers.17.mlp.gate": {
596
+ "group_size": 64,
597
+ "bits": 4
598
+ },
599
+ "model.layers.17.mlp.switch_mlp.gate_proj": {
600
+ "group_size": 64,
601
+ "bits": 4
602
+ },
603
+ "model.layers.17.mlp.switch_mlp.up_proj": {
604
+ "group_size": 64,
605
+ "bits": 4
606
+ },
607
+ "model.layers.17.mlp.switch_mlp.down_proj": {
608
+ "group_size": 32,
609
+ "bits": 6
610
+ },
611
+ "model.layers.18.self_attn.q_proj": {
612
+ "group_size": 64,
613
+ "bits": 4
614
+ },
615
+ "model.layers.18.self_attn.k_proj": {
616
+ "group_size": 64,
617
+ "bits": 4
618
+ },
619
+ "model.layers.18.self_attn.v_proj": {
620
+ "group_size": 64,
621
+ "bits": 4
622
+ },
623
+ "model.layers.18.self_attn.o_proj": {
624
+ "group_size": 64,
625
+ "bits": 4
626
+ },
627
+ "model.layers.18.mlp.gate": {
628
+ "group_size": 64,
629
+ "bits": 4
630
+ },
631
+ "model.layers.18.mlp.switch_mlp.gate_proj": {
632
+ "group_size": 64,
633
+ "bits": 4
634
+ },
635
+ "model.layers.18.mlp.switch_mlp.up_proj": {
636
+ "group_size": 64,
637
+ "bits": 4
638
+ },
639
+ "model.layers.18.mlp.switch_mlp.down_proj": {
640
+ "group_size": 64,
641
+ "bits": 4
642
+ },
643
+ "model.layers.19.self_attn.q_proj": {
644
+ "group_size": 64,
645
+ "bits": 4
646
+ },
647
+ "model.layers.19.self_attn.k_proj": {
648
+ "group_size": 64,
649
+ "bits": 4
650
+ },
651
+ "model.layers.19.self_attn.v_proj": {
652
+ "group_size": 64,
653
+ "bits": 4
654
+ },
655
+ "model.layers.19.self_attn.o_proj": {
656
+ "group_size": 64,
657
+ "bits": 4
658
+ },
659
+ "model.layers.19.mlp.gate": {
660
+ "group_size": 64,
661
+ "bits": 4
662
+ },
663
+ "model.layers.19.mlp.switch_mlp.gate_proj": {
664
+ "group_size": 64,
665
+ "bits": 4
666
+ },
667
+ "model.layers.19.mlp.switch_mlp.up_proj": {
668
+ "group_size": 64,
669
+ "bits": 4
670
+ },
671
+ "model.layers.19.mlp.switch_mlp.down_proj": {
672
+ "group_size": 64,
673
+ "bits": 4
674
+ },
675
+ "model.layers.20.self_attn.q_proj": {
676
+ "group_size": 64,
677
+ "bits": 4
678
+ },
679
+ "model.layers.20.self_attn.k_proj": {
680
+ "group_size": 64,
681
+ "bits": 4
682
+ },
683
+ "model.layers.20.self_attn.v_proj": {
684
+ "group_size": 32,
685
+ "bits": 6
686
+ },
687
+ "model.layers.20.self_attn.o_proj": {
688
+ "group_size": 64,
689
+ "bits": 4
690
+ },
691
+ "model.layers.20.mlp.gate": {
692
+ "group_size": 64,
693
+ "bits": 4
694
+ },
695
+ "model.layers.20.mlp.switch_mlp.gate_proj": {
696
+ "group_size": 64,
697
+ "bits": 4
698
+ },
699
+ "model.layers.20.mlp.switch_mlp.up_proj": {
700
+ "group_size": 64,
701
+ "bits": 4
702
+ },
703
+ "model.layers.20.mlp.switch_mlp.down_proj": {
704
+ "group_size": 32,
705
+ "bits": 6
706
+ },
707
+ "model.layers.21.self_attn.q_proj": {
708
+ "group_size": 64,
709
+ "bits": 4
710
+ },
711
+ "model.layers.21.self_attn.k_proj": {
712
+ "group_size": 64,
713
+ "bits": 4
714
+ },
715
+ "model.layers.21.self_attn.v_proj": {
716
+ "group_size": 64,
717
+ "bits": 4
718
+ },
719
+ "model.layers.21.self_attn.o_proj": {
720
+ "group_size": 64,
721
+ "bits": 4
722
+ },
723
+ "model.layers.21.mlp.gate": {
724
+ "group_size": 64,
725
+ "bits": 4
726
+ },
727
+ "model.layers.21.mlp.switch_mlp.gate_proj": {
728
+ "group_size": 64,
729
+ "bits": 4
730
+ },
731
+ "model.layers.21.mlp.switch_mlp.up_proj": {
732
+ "group_size": 64,
733
+ "bits": 4
734
+ },
735
+ "model.layers.21.mlp.switch_mlp.down_proj": {
736
+ "group_size": 64,
737
+ "bits": 4
738
+ },
739
+ "model.layers.22.self_attn.q_proj": {
740
+ "group_size": 64,
741
+ "bits": 4
742
+ },
743
+ "model.layers.22.self_attn.k_proj": {
744
+ "group_size": 64,
745
+ "bits": 4
746
+ },
747
+ "model.layers.22.self_attn.v_proj": {
748
+ "group_size": 64,
749
+ "bits": 4
750
+ },
751
+ "model.layers.22.self_attn.o_proj": {
752
+ "group_size": 64,
753
+ "bits": 4
754
+ },
755
+ "model.layers.22.mlp.gate": {
756
+ "group_size": 64,
757
+ "bits": 4
758
+ },
759
+ "model.layers.22.mlp.switch_mlp.gate_proj": {
760
+ "group_size": 64,
761
+ "bits": 4
762
+ },
763
+ "model.layers.22.mlp.switch_mlp.up_proj": {
764
+ "group_size": 64,
765
+ "bits": 4
766
+ },
767
+ "model.layers.22.mlp.switch_mlp.down_proj": {
768
+ "group_size": 64,
769
+ "bits": 4
770
+ },
771
+ "model.layers.23.self_attn.q_proj": {
772
+ "group_size": 64,
773
+ "bits": 4
774
+ },
775
+ "model.layers.23.self_attn.k_proj": {
776
+ "group_size": 64,
777
+ "bits": 4
778
+ },
779
+ "model.layers.23.self_attn.v_proj": {
780
+ "group_size": 32,
781
+ "bits": 6
782
+ },
783
+ "model.layers.23.self_attn.o_proj": {
784
+ "group_size": 64,
785
+ "bits": 4
786
+ },
787
+ "model.layers.23.mlp.gate": {
788
+ "group_size": 64,
789
+ "bits": 4
790
+ },
791
+ "model.layers.23.mlp.switch_mlp.gate_proj": {
792
+ "group_size": 64,
793
+ "bits": 4
794
+ },
795
+ "model.layers.23.mlp.switch_mlp.up_proj": {
796
+ "group_size": 64,
797
+ "bits": 4
798
+ },
799
+ "model.layers.23.mlp.switch_mlp.down_proj": {
800
+ "group_size": 32,
801
+ "bits": 6
802
+ },
803
+ "model.layers.24.self_attn.q_proj": {
804
+ "group_size": 64,
805
+ "bits": 4
806
+ },
807
+ "model.layers.24.self_attn.k_proj": {
808
+ "group_size": 64,
809
+ "bits": 4
810
+ },
811
+ "model.layers.24.self_attn.v_proj": {
812
+ "group_size": 64,
813
+ "bits": 4
814
+ },
815
+ "model.layers.24.self_attn.o_proj": {
816
+ "group_size": 64,
817
+ "bits": 4
818
+ },
819
+ "model.layers.24.mlp.gate": {
820
+ "group_size": 64,
821
+ "bits": 4
822
+ },
823
+ "model.layers.24.mlp.switch_mlp.gate_proj": {
824
+ "group_size": 64,
825
+ "bits": 4
826
+ },
827
+ "model.layers.24.mlp.switch_mlp.up_proj": {
828
+ "group_size": 64,
829
+ "bits": 4
830
+ },
831
+ "model.layers.24.mlp.switch_mlp.down_proj": {
832
+ "group_size": 64,
833
+ "bits": 4
834
+ },
835
+ "model.layers.25.self_attn.q_proj": {
836
+ "group_size": 64,
837
+ "bits": 4
838
+ },
839
+ "model.layers.25.self_attn.k_proj": {
840
+ "group_size": 64,
841
+ "bits": 4
842
+ },
843
+ "model.layers.25.self_attn.v_proj": {
844
+ "group_size": 64,
845
+ "bits": 4
846
+ },
847
+ "model.layers.25.self_attn.o_proj": {
848
+ "group_size": 64,
849
+ "bits": 4
850
+ },
851
+ "model.layers.25.mlp.gate": {
852
+ "group_size": 64,
853
+ "bits": 4
854
+ },
855
+ "model.layers.25.mlp.switch_mlp.gate_proj": {
856
+ "group_size": 64,
857
+ "bits": 4
858
+ },
859
+ "model.layers.25.mlp.switch_mlp.up_proj": {
860
+ "group_size": 64,
861
+ "bits": 4
862
+ },
863
+ "model.layers.25.mlp.switch_mlp.down_proj": {
864
+ "group_size": 64,
865
+ "bits": 4
866
+ },
867
+ "model.layers.26.self_attn.q_proj": {
868
+ "group_size": 64,
869
+ "bits": 4
870
+ },
871
+ "model.layers.26.self_attn.k_proj": {
872
+ "group_size": 64,
873
+ "bits": 4
874
+ },
875
+ "model.layers.26.self_attn.v_proj": {
876
+ "group_size": 32,
877
+ "bits": 6
878
+ },
879
+ "model.layers.26.self_attn.o_proj": {
880
+ "group_size": 64,
881
+ "bits": 4
882
+ },
883
+ "model.layers.26.mlp.gate": {
884
+ "group_size": 64,
885
+ "bits": 4
886
+ },
887
+ "model.layers.26.mlp.switch_mlp.gate_proj": {
888
+ "group_size": 64,
889
+ "bits": 4
890
+ },
891
+ "model.layers.26.mlp.switch_mlp.up_proj": {
892
+ "group_size": 64,
893
+ "bits": 4
894
+ },
895
+ "model.layers.26.mlp.switch_mlp.down_proj": {
896
+ "group_size": 32,
897
+ "bits": 6
898
+ },
899
+ "model.layers.27.self_attn.q_proj": {
900
+ "group_size": 64,
901
+ "bits": 4
902
+ },
903
+ "model.layers.27.self_attn.k_proj": {
904
+ "group_size": 64,
905
+ "bits": 4
906
+ },
907
+ "model.layers.27.self_attn.v_proj": {
908
+ "group_size": 64,
909
+ "bits": 4
910
+ },
911
+ "model.layers.27.self_attn.o_proj": {
912
+ "group_size": 64,
913
+ "bits": 4
914
+ },
915
+ "model.layers.27.mlp.gate": {
916
+ "group_size": 64,
917
+ "bits": 4
918
+ },
919
+ "model.layers.27.mlp.switch_mlp.gate_proj": {
920
+ "group_size": 64,
921
+ "bits": 4
922
+ },
923
+ "model.layers.27.mlp.switch_mlp.up_proj": {
924
+ "group_size": 64,
925
+ "bits": 4
926
+ },
927
+ "model.layers.27.mlp.switch_mlp.down_proj": {
928
+ "group_size": 64,
929
+ "bits": 4
930
+ },
931
+ "model.layers.28.self_attn.q_proj": {
932
+ "group_size": 64,
933
+ "bits": 4
934
+ },
935
+ "model.layers.28.self_attn.k_proj": {
936
+ "group_size": 64,
937
+ "bits": 4
938
+ },
939
+ "model.layers.28.self_attn.v_proj": {
940
+ "group_size": 64,
941
+ "bits": 4
942
+ },
943
+ "model.layers.28.self_attn.o_proj": {
944
+ "group_size": 64,
945
+ "bits": 4
946
+ },
947
+ "model.layers.28.mlp.gate": {
948
+ "group_size": 64,
949
+ "bits": 4
950
+ },
951
+ "model.layers.28.mlp.switch_mlp.gate_proj": {
952
+ "group_size": 64,
953
+ "bits": 4
954
+ },
955
+ "model.layers.28.mlp.switch_mlp.up_proj": {
956
+ "group_size": 64,
957
+ "bits": 4
958
+ },
959
+ "model.layers.28.mlp.switch_mlp.down_proj": {
960
+ "group_size": 64,
961
+ "bits": 4
962
+ },
963
+ "model.layers.29.self_attn.q_proj": {
964
+ "group_size": 64,
965
+ "bits": 4
966
+ },
967
+ "model.layers.29.self_attn.k_proj": {
968
+ "group_size": 64,
969
+ "bits": 4
970
+ },
971
+ "model.layers.29.self_attn.v_proj": {
972
+ "group_size": 32,
973
+ "bits": 6
974
+ },
975
+ "model.layers.29.self_attn.o_proj": {
976
+ "group_size": 64,
977
+ "bits": 4
978
+ },
979
+ "model.layers.29.mlp.gate": {
980
+ "group_size": 64,
981
+ "bits": 4
982
+ },
983
+ "model.layers.29.mlp.switch_mlp.gate_proj": {
984
+ "group_size": 64,
985
+ "bits": 4
986
+ },
987
+ "model.layers.29.mlp.switch_mlp.up_proj": {
988
+ "group_size": 64,
989
+ "bits": 4
990
+ },
991
+ "model.layers.29.mlp.switch_mlp.down_proj": {
992
+ "group_size": 32,
993
+ "bits": 6
994
+ },
995
+ "model.layers.30.self_attn.q_proj": {
996
+ "group_size": 64,
997
+ "bits": 4
998
+ },
999
+ "model.layers.30.self_attn.k_proj": {
1000
+ "group_size": 64,
1001
+ "bits": 4
1002
+ },
1003
+ "model.layers.30.self_attn.v_proj": {
1004
+ "group_size": 64,
1005
+ "bits": 4
1006
+ },
1007
+ "model.layers.30.self_attn.o_proj": {
1008
+ "group_size": 64,
1009
+ "bits": 4
1010
+ },
1011
+ "model.layers.30.mlp.gate": {
1012
+ "group_size": 64,
1013
+ "bits": 4
1014
+ },
1015
+ "model.layers.30.mlp.switch_mlp.gate_proj": {
1016
+ "group_size": 64,
1017
+ "bits": 4
1018
+ },
1019
+ "model.layers.30.mlp.switch_mlp.up_proj": {
1020
+ "group_size": 64,
1021
+ "bits": 4
1022
+ },
1023
+ "model.layers.30.mlp.switch_mlp.down_proj": {
1024
+ "group_size": 64,
1025
+ "bits": 4
1026
+ },
1027
+ "model.layers.31.self_attn.q_proj": {
1028
+ "group_size": 64,
1029
+ "bits": 4
1030
+ },
1031
+ "model.layers.31.self_attn.k_proj": {
1032
+ "group_size": 64,
1033
+ "bits": 4
1034
+ },
1035
+ "model.layers.31.self_attn.v_proj": {
1036
+ "group_size": 64,
1037
+ "bits": 4
1038
+ },
1039
+ "model.layers.31.self_attn.o_proj": {
1040
+ "group_size": 64,
1041
+ "bits": 4
1042
+ },
1043
+ "model.layers.31.mlp.gate": {
1044
+ "group_size": 64,
1045
+ "bits": 4
1046
+ },
1047
+ "model.layers.31.mlp.switch_mlp.gate_proj": {
1048
+ "group_size": 64,
1049
+ "bits": 4
1050
+ },
1051
+ "model.layers.31.mlp.switch_mlp.up_proj": {
1052
+ "group_size": 64,
1053
+ "bits": 4
1054
+ },
1055
+ "model.layers.31.mlp.switch_mlp.down_proj": {
1056
+ "group_size": 64,
1057
+ "bits": 4
1058
+ },
1059
+ "model.layers.32.self_attn.q_proj": {
1060
+ "group_size": 64,
1061
+ "bits": 4
1062
+ },
1063
+ "model.layers.32.self_attn.k_proj": {
1064
+ "group_size": 64,
1065
+ "bits": 4
1066
+ },
1067
+ "model.layers.32.self_attn.v_proj": {
1068
+ "group_size": 32,
1069
+ "bits": 6
1070
+ },
1071
+ "model.layers.32.self_attn.o_proj": {
1072
+ "group_size": 64,
1073
+ "bits": 4
1074
+ },
1075
+ "model.layers.32.mlp.gate": {
1076
+ "group_size": 64,
1077
+ "bits": 4
1078
+ },
1079
+ "model.layers.32.mlp.switch_mlp.gate_proj": {
1080
+ "group_size": 64,
1081
+ "bits": 4
1082
+ },
1083
+ "model.layers.32.mlp.switch_mlp.up_proj": {
1084
+ "group_size": 64,
1085
+ "bits": 4
1086
+ },
1087
+ "model.layers.32.mlp.switch_mlp.down_proj": {
1088
+ "group_size": 32,
1089
+ "bits": 6
1090
+ },
1091
+ "model.layers.33.self_attn.q_proj": {
1092
+ "group_size": 64,
1093
+ "bits": 4
1094
+ },
1095
+ "model.layers.33.self_attn.k_proj": {
1096
+ "group_size": 64,
1097
+ "bits": 4
1098
+ },
1099
+ "model.layers.33.self_attn.v_proj": {
1100
+ "group_size": 64,
1101
+ "bits": 4
1102
+ },
1103
+ "model.layers.33.self_attn.o_proj": {
1104
+ "group_size": 64,
1105
+ "bits": 4
1106
+ },
1107
+ "model.layers.33.mlp.gate": {
1108
+ "group_size": 64,
1109
+ "bits": 4
1110
+ },
1111
+ "model.layers.33.mlp.switch_mlp.gate_proj": {
1112
+ "group_size": 64,
1113
+ "bits": 4
1114
+ },
1115
+ "model.layers.33.mlp.switch_mlp.up_proj": {
1116
+ "group_size": 64,
1117
+ "bits": 4
1118
+ },
1119
+ "model.layers.33.mlp.switch_mlp.down_proj": {
1120
+ "group_size": 64,
1121
+ "bits": 4
1122
+ },
1123
+ "model.layers.34.self_attn.q_proj": {
1124
+ "group_size": 64,
1125
+ "bits": 4
1126
+ },
1127
+ "model.layers.34.self_attn.k_proj": {
1128
+ "group_size": 64,
1129
+ "bits": 4
1130
+ },
1131
+ "model.layers.34.self_attn.v_proj": {
1132
+ "group_size": 64,
1133
+ "bits": 4
1134
+ },
1135
+ "model.layers.34.self_attn.o_proj": {
1136
+ "group_size": 64,
1137
+ "bits": 4
1138
+ },
1139
+ "model.layers.34.mlp.gate": {
1140
+ "group_size": 64,
1141
+ "bits": 4
1142
+ },
1143
+ "model.layers.34.mlp.switch_mlp.gate_proj": {
1144
+ "group_size": 64,
1145
+ "bits": 4
1146
+ },
1147
+ "model.layers.34.mlp.switch_mlp.up_proj": {
1148
+ "group_size": 64,
1149
+ "bits": 4
1150
+ },
1151
+ "model.layers.34.mlp.switch_mlp.down_proj": {
1152
+ "group_size": 64,
1153
+ "bits": 4
1154
+ },
1155
+ "model.layers.35.self_attn.q_proj": {
1156
+ "group_size": 64,
1157
+ "bits": 4
1158
+ },
1159
+ "model.layers.35.self_attn.k_proj": {
1160
+ "group_size": 64,
1161
+ "bits": 4
1162
+ },
1163
+ "model.layers.35.self_attn.v_proj": {
1164
+ "group_size": 32,
1165
+ "bits": 6
1166
+ },
1167
+ "model.layers.35.self_attn.o_proj": {
1168
+ "group_size": 64,
1169
+ "bits": 4
1170
+ },
1171
+ "model.layers.35.mlp.gate": {
1172
+ "group_size": 64,
1173
+ "bits": 4
1174
+ },
1175
+ "model.layers.35.mlp.switch_mlp.gate_proj": {
1176
+ "group_size": 64,
1177
+ "bits": 4
1178
+ },
1179
+ "model.layers.35.mlp.switch_mlp.up_proj": {
1180
+ "group_size": 64,
1181
+ "bits": 4
1182
+ },
1183
+ "model.layers.35.mlp.switch_mlp.down_proj": {
1184
+ "group_size": 32,
1185
+ "bits": 6
1186
+ },
1187
+ "model.layers.36.self_attn.q_proj": {
1188
+ "group_size": 64,
1189
+ "bits": 4
1190
+ },
1191
+ "model.layers.36.self_attn.k_proj": {
1192
+ "group_size": 64,
1193
+ "bits": 4
1194
+ },
1195
+ "model.layers.36.self_attn.v_proj": {
1196
+ "group_size": 64,
1197
+ "bits": 4
1198
+ },
1199
+ "model.layers.36.self_attn.o_proj": {
1200
+ "group_size": 64,
1201
+ "bits": 4
1202
+ },
1203
+ "model.layers.36.mlp.gate": {
1204
+ "group_size": 64,
1205
+ "bits": 4
1206
+ },
1207
+ "model.layers.36.mlp.switch_mlp.gate_proj": {
1208
+ "group_size": 64,
1209
+ "bits": 4
1210
+ },
1211
+ "model.layers.36.mlp.switch_mlp.up_proj": {
1212
+ "group_size": 64,
1213
+ "bits": 4
1214
+ },
1215
+ "model.layers.36.mlp.switch_mlp.down_proj": {
1216
+ "group_size": 64,
1217
+ "bits": 4
1218
+ },
1219
+ "model.layers.37.self_attn.q_proj": {
1220
+ "group_size": 64,
1221
+ "bits": 4
1222
+ },
1223
+ "model.layers.37.self_attn.k_proj": {
1224
+ "group_size": 64,
1225
+ "bits": 4
1226
+ },
1227
+ "model.layers.37.self_attn.v_proj": {
1228
+ "group_size": 64,
1229
+ "bits": 4
1230
+ },
1231
+ "model.layers.37.self_attn.o_proj": {
1232
+ "group_size": 64,
1233
+ "bits": 4
1234
+ },
1235
+ "model.layers.37.mlp.gate": {
1236
+ "group_size": 64,
1237
+ "bits": 4
1238
+ },
1239
+ "model.layers.37.mlp.switch_mlp.gate_proj": {
1240
+ "group_size": 64,
1241
+ "bits": 4
1242
+ },
1243
+ "model.layers.37.mlp.switch_mlp.up_proj": {
1244
+ "group_size": 64,
1245
+ "bits": 4
1246
+ },
1247
+ "model.layers.37.mlp.switch_mlp.down_proj": {
1248
+ "group_size": 64,
1249
+ "bits": 4
1250
+ },
1251
+ "model.layers.38.self_attn.q_proj": {
1252
+ "group_size": 64,
1253
+ "bits": 4
1254
+ },
1255
+ "model.layers.38.self_attn.k_proj": {
1256
+ "group_size": 64,
1257
+ "bits": 4
1258
+ },
1259
+ "model.layers.38.self_attn.v_proj": {
1260
+ "group_size": 32,
1261
+ "bits": 6
1262
+ },
1263
+ "model.layers.38.self_attn.o_proj": {
1264
+ "group_size": 64,
1265
+ "bits": 4
1266
+ },
1267
+ "model.layers.38.mlp.gate": {
1268
+ "group_size": 64,
1269
+ "bits": 4
1270
+ },
1271
+ "model.layers.38.mlp.switch_mlp.gate_proj": {
1272
+ "group_size": 64,
1273
+ "bits": 4
1274
+ },
1275
+ "model.layers.38.mlp.switch_mlp.up_proj": {
1276
+ "group_size": 64,
1277
+ "bits": 4
1278
+ },
1279
+ "model.layers.38.mlp.switch_mlp.down_proj": {
1280
+ "group_size": 32,
1281
+ "bits": 6
1282
+ },
1283
+ "model.layers.39.self_attn.q_proj": {
1284
+ "group_size": 64,
1285
+ "bits": 4
1286
+ },
1287
+ "model.layers.39.self_attn.k_proj": {
1288
+ "group_size": 64,
1289
+ "bits": 4
1290
+ },
1291
+ "model.layers.39.self_attn.v_proj": {
1292
+ "group_size": 64,
1293
+ "bits": 4
1294
+ },
1295
+ "model.layers.39.self_attn.o_proj": {
1296
+ "group_size": 64,
1297
+ "bits": 4
1298
+ },
1299
+ "model.layers.39.mlp.gate": {
1300
+ "group_size": 64,
1301
+ "bits": 4
1302
+ },
1303
+ "model.layers.39.mlp.switch_mlp.gate_proj": {
1304
+ "group_size": 64,
1305
+ "bits": 4
1306
+ },
1307
+ "model.layers.39.mlp.switch_mlp.up_proj": {
1308
+ "group_size": 64,
1309
+ "bits": 4
1310
+ },
1311
+ "model.layers.39.mlp.switch_mlp.down_proj": {
1312
+ "group_size": 64,
1313
+ "bits": 4
1314
+ },
1315
+ "model.layers.40.self_attn.q_proj": {
1316
+ "group_size": 64,
1317
+ "bits": 4
1318
+ },
1319
+ "model.layers.40.self_attn.k_proj": {
1320
+ "group_size": 64,
1321
+ "bits": 4
1322
+ },
1323
+ "model.layers.40.self_attn.v_proj": {
1324
+ "group_size": 64,
1325
+ "bits": 4
1326
+ },
1327
+ "model.layers.40.self_attn.o_proj": {
1328
+ "group_size": 64,
1329
+ "bits": 4
1330
+ },
1331
+ "model.layers.40.mlp.gate": {
1332
+ "group_size": 64,
1333
+ "bits": 4
1334
+ },
1335
+ "model.layers.40.mlp.switch_mlp.gate_proj": {
1336
+ "group_size": 64,
1337
+ "bits": 4
1338
+ },
1339
+ "model.layers.40.mlp.switch_mlp.up_proj": {
1340
+ "group_size": 64,
1341
+ "bits": 4
1342
+ },
1343
+ "model.layers.40.mlp.switch_mlp.down_proj": {
1344
+ "group_size": 64,
1345
+ "bits": 4
1346
+ },
1347
+ "model.layers.41.self_attn.q_proj": {
1348
+ "group_size": 64,
1349
+ "bits": 4
1350
+ },
1351
+ "model.layers.41.self_attn.k_proj": {
1352
+ "group_size": 64,
1353
+ "bits": 4
1354
+ },
1355
+ "model.layers.41.self_attn.v_proj": {
1356
+ "group_size": 32,
1357
+ "bits": 6
1358
+ },
1359
+ "model.layers.41.self_attn.o_proj": {
1360
+ "group_size": 64,
1361
+ "bits": 4
1362
+ },
1363
+ "model.layers.41.mlp.gate": {
1364
+ "group_size": 64,
1365
+ "bits": 4
1366
+ },
1367
+ "model.layers.41.mlp.switch_mlp.gate_proj": {
1368
+ "group_size": 64,
1369
+ "bits": 4
1370
+ },
1371
+ "model.layers.41.mlp.switch_mlp.up_proj": {
1372
+ "group_size": 64,
1373
+ "bits": 4
1374
+ },
1375
+ "model.layers.41.mlp.switch_mlp.down_proj": {
1376
+ "group_size": 32,
1377
+ "bits": 6
1378
+ },
1379
+ "model.layers.42.self_attn.q_proj": {
1380
+ "group_size": 64,
1381
+ "bits": 4
1382
+ },
1383
+ "model.layers.42.self_attn.k_proj": {
1384
+ "group_size": 64,
1385
+ "bits": 4
1386
+ },
1387
+ "model.layers.42.self_attn.v_proj": {
1388
+ "group_size": 32,
1389
+ "bits": 6
1390
+ },
1391
+ "model.layers.42.self_attn.o_proj": {
1392
+ "group_size": 64,
1393
+ "bits": 4
1394
+ },
1395
+ "model.layers.42.mlp.gate": {
1396
+ "group_size": 64,
1397
+ "bits": 4
1398
+ },
1399
+ "model.layers.42.mlp.switch_mlp.gate_proj": {
1400
+ "group_size": 64,
1401
+ "bits": 4
1402
+ },
1403
+ "model.layers.42.mlp.switch_mlp.up_proj": {
1404
+ "group_size": 64,
1405
+ "bits": 4
1406
+ },
1407
+ "model.layers.42.mlp.switch_mlp.down_proj": {
1408
+ "group_size": 32,
1409
+ "bits": 6
1410
+ },
1411
+ "model.layers.43.self_attn.q_proj": {
1412
+ "group_size": 64,
1413
+ "bits": 4
1414
+ },
1415
+ "model.layers.43.self_attn.k_proj": {
1416
+ "group_size": 64,
1417
+ "bits": 4
1418
+ },
1419
+ "model.layers.43.self_attn.v_proj": {
1420
+ "group_size": 32,
1421
+ "bits": 6
1422
+ },
1423
+ "model.layers.43.self_attn.o_proj": {
1424
+ "group_size": 64,
1425
+ "bits": 4
1426
+ },
1427
+ "model.layers.43.mlp.gate": {
1428
+ "group_size": 64,
1429
+ "bits": 4
1430
+ },
1431
+ "model.layers.43.mlp.switch_mlp.gate_proj": {
1432
+ "group_size": 64,
1433
+ "bits": 4
1434
+ },
1435
+ "model.layers.43.mlp.switch_mlp.up_proj": {
1436
+ "group_size": 64,
1437
+ "bits": 4
1438
+ },
1439
+ "model.layers.43.mlp.switch_mlp.down_proj": {
1440
+ "group_size": 32,
1441
+ "bits": 6
1442
+ },
1443
+ "model.layers.44.self_attn.q_proj": {
1444
+ "group_size": 64,
1445
+ "bits": 4
1446
+ },
1447
+ "model.layers.44.self_attn.k_proj": {
1448
+ "group_size": 64,
1449
+ "bits": 4
1450
+ },
1451
+ "model.layers.44.self_attn.v_proj": {
1452
+ "group_size": 32,
1453
+ "bits": 6
1454
+ },
1455
+ "model.layers.44.self_attn.o_proj": {
1456
+ "group_size": 64,
1457
+ "bits": 4
1458
+ },
1459
+ "model.layers.44.mlp.gate": {
1460
+ "group_size": 64,
1461
+ "bits": 4
1462
+ },
1463
+ "model.layers.44.mlp.switch_mlp.gate_proj": {
1464
+ "group_size": 64,
1465
+ "bits": 4
1466
+ },
1467
+ "model.layers.44.mlp.switch_mlp.up_proj": {
1468
+ "group_size": 64,
1469
+ "bits": 4
1470
+ },
1471
+ "model.layers.44.mlp.switch_mlp.down_proj": {
1472
+ "group_size": 32,
1473
+ "bits": 6
1474
+ },
1475
+ "model.layers.45.self_attn.q_proj": {
1476
+ "group_size": 64,
1477
+ "bits": 4
1478
+ },
1479
+ "model.layers.45.self_attn.k_proj": {
1480
+ "group_size": 64,
1481
+ "bits": 4
1482
+ },
1483
+ "model.layers.45.self_attn.v_proj": {
1484
+ "group_size": 32,
1485
+ "bits": 6
1486
+ },
1487
+ "model.layers.45.self_attn.o_proj": {
1488
+ "group_size": 64,
1489
+ "bits": 4
1490
+ },
1491
+ "model.layers.45.mlp.gate": {
1492
+ "group_size": 64,
1493
+ "bits": 4
1494
+ },
1495
+ "model.layers.45.mlp.switch_mlp.gate_proj": {
1496
+ "group_size": 64,
1497
+ "bits": 4
1498
+ },
1499
+ "model.layers.45.mlp.switch_mlp.up_proj": {
1500
+ "group_size": 64,
1501
+ "bits": 4
1502
+ },
1503
+ "model.layers.45.mlp.switch_mlp.down_proj": {
1504
+ "group_size": 32,
1505
+ "bits": 6
1506
+ },
1507
+ "model.layers.46.self_attn.q_proj": {
1508
+ "group_size": 64,
1509
+ "bits": 4
1510
+ },
1511
+ "model.layers.46.self_attn.k_proj": {
1512
+ "group_size": 64,
1513
+ "bits": 4
1514
+ },
1515
+ "model.layers.46.self_attn.v_proj": {
1516
+ "group_size": 32,
1517
+ "bits": 6
1518
+ },
1519
+ "model.layers.46.self_attn.o_proj": {
1520
+ "group_size": 64,
1521
+ "bits": 4
1522
+ },
1523
+ "model.layers.46.mlp.gate": {
1524
+ "group_size": 64,
1525
+ "bits": 4
1526
+ },
1527
+ "model.layers.46.mlp.switch_mlp.gate_proj": {
1528
+ "group_size": 64,
1529
+ "bits": 4
1530
+ },
1531
+ "model.layers.46.mlp.switch_mlp.up_proj": {
1532
+ "group_size": 64,
1533
+ "bits": 4
1534
+ },
1535
+ "model.layers.46.mlp.switch_mlp.down_proj": {
1536
+ "group_size": 32,
1537
+ "bits": 6
1538
+ },
1539
+ "model.layers.47.self_attn.q_proj": {
1540
+ "group_size": 64,
1541
+ "bits": 4
1542
+ },
1543
+ "model.layers.47.self_attn.k_proj": {
1544
+ "group_size": 64,
1545
+ "bits": 4
1546
+ },
1547
+ "model.layers.47.self_attn.v_proj": {
1548
+ "group_size": 32,
1549
+ "bits": 6
1550
+ },
1551
+ "model.layers.47.self_attn.o_proj": {
1552
+ "group_size": 64,
1553
+ "bits": 4
1554
+ },
1555
+ "model.layers.47.mlp.gate": {
1556
+ "group_size": 64,
1557
+ "bits": 4
1558
+ },
1559
+ "model.layers.47.mlp.switch_mlp.gate_proj": {
1560
+ "group_size": 64,
1561
+ "bits": 4
1562
+ },
1563
+ "model.layers.47.mlp.switch_mlp.up_proj": {
1564
+ "group_size": 64,
1565
+ "bits": 4
1566
+ },
1567
+ "model.layers.47.mlp.switch_mlp.down_proj": {
1568
+ "group_size": 32,
1569
+ "bits": 6
1570
+ },
1571
+ "lm_head": {
1572
+ "group_size": 32,
1573
+ "bits": 8
1574
+ }
1575
+ },
1576
+ "quantization_config": {
1577
+ "group_size": 64,
1578
+ "bits": 4,
1579
+ "model.embed_tokens": {
1580
+ "group_size": 64,
1581
+ "bits": 4
1582
+ },
1583
+ "model.layers.0.self_attn.q_proj": {
1584
+ "group_size": 32,
1585
+ "bits": 8
1586
+ },
1587
+ "model.layers.0.self_attn.k_proj": {
1588
+ "group_size": 32,
1589
+ "bits": 8
1590
+ },
1591
+ "model.layers.0.self_attn.v_proj": {
1592
+ "group_size": 32,
1593
+ "bits": 8
1594
+ },
1595
+ "model.layers.0.self_attn.o_proj": {
1596
+ "group_size": 32,
1597
+ "bits": 8
1598
+ },
1599
+ "model.layers.0.mlp.gate": {
1600
+ "group_size": 32,
1601
+ "bits": 8
1602
+ },
1603
+ "model.layers.0.mlp.switch_mlp.gate_proj": {
1604
+ "group_size": 32,
1605
+ "bits": 8
1606
+ },
1607
+ "model.layers.0.mlp.switch_mlp.up_proj": {
1608
+ "group_size": 32,
1609
+ "bits": 8
1610
+ },
1611
+ "model.layers.0.mlp.switch_mlp.down_proj": {
1612
+ "group_size": 32,
1613
+ "bits": 8
1614
+ },
1615
+ "model.layers.1.self_attn.q_proj": {
1616
+ "group_size": 64,
1617
+ "bits": 4
1618
+ },
1619
+ "model.layers.1.self_attn.k_proj": {
1620
+ "group_size": 64,
1621
+ "bits": 4
1622
+ },
1623
+ "model.layers.1.self_attn.v_proj": {
1624
+ "group_size": 32,
1625
+ "bits": 8
1626
+ },
1627
+ "model.layers.1.self_attn.o_proj": {
1628
+ "group_size": 64,
1629
+ "bits": 4
1630
+ },
1631
+ "model.layers.1.mlp.gate": {
1632
+ "group_size": 64,
1633
+ "bits": 4
1634
+ },
1635
+ "model.layers.1.mlp.switch_mlp.gate_proj": {
1636
+ "group_size": 64,
1637
+ "bits": 4
1638
+ },
1639
+ "model.layers.1.mlp.switch_mlp.up_proj": {
1640
+ "group_size": 64,
1641
+ "bits": 4
1642
+ },
1643
+ "model.layers.1.mlp.switch_mlp.down_proj": {
1644
+ "group_size": 32,
1645
+ "bits": 6
1646
+ },
1647
+ "model.layers.2.self_attn.q_proj": {
1648
+ "group_size": 64,
1649
+ "bits": 4
1650
+ },
1651
+ "model.layers.2.self_attn.k_proj": {
1652
+ "group_size": 64,
1653
+ "bits": 4
1654
+ },
1655
+ "model.layers.2.self_attn.v_proj": {
1656
+ "group_size": 32,
1657
+ "bits": 8
1658
+ },
1659
+ "model.layers.2.self_attn.o_proj": {
1660
+ "group_size": 64,
1661
+ "bits": 4
1662
+ },
1663
+ "model.layers.2.mlp.gate": {
1664
+ "group_size": 64,
1665
+ "bits": 4
1666
+ },
1667
+ "model.layers.2.mlp.switch_mlp.gate_proj": {
1668
+ "group_size": 64,
1669
+ "bits": 4
1670
+ },
1671
+ "model.layers.2.mlp.switch_mlp.up_proj": {
1672
+ "group_size": 64,
1673
+ "bits": 4
1674
+ },
1675
+ "model.layers.2.mlp.switch_mlp.down_proj": {
1676
+ "group_size": 32,
1677
+ "bits": 6
1678
+ },
1679
+ "model.layers.3.self_attn.q_proj": {
1680
+ "group_size": 64,
1681
+ "bits": 4
1682
+ },
1683
+ "model.layers.3.self_attn.k_proj": {
1684
+ "group_size": 64,
1685
+ "bits": 4
1686
+ },
1687
+ "model.layers.3.self_attn.v_proj": {
1688
+ "group_size": 32,
1689
+ "bits": 8
1690
+ },
1691
+ "model.layers.3.self_attn.o_proj": {
1692
+ "group_size": 64,
1693
+ "bits": 4
1694
+ },
1695
+ "model.layers.3.mlp.gate": {
1696
+ "group_size": 64,
1697
+ "bits": 4
1698
+ },
1699
+ "model.layers.3.mlp.switch_mlp.gate_proj": {
1700
+ "group_size": 64,
1701
+ "bits": 4
1702
+ },
1703
+ "model.layers.3.mlp.switch_mlp.up_proj": {
1704
+ "group_size": 64,
1705
+ "bits": 4
1706
+ },
1707
+ "model.layers.3.mlp.switch_mlp.down_proj": {
1708
+ "group_size": 32,
1709
+ "bits": 6
1710
+ },
1711
+ "model.layers.4.self_attn.q_proj": {
1712
+ "group_size": 64,
1713
+ "bits": 4
1714
+ },
1715
+ "model.layers.4.self_attn.k_proj": {
1716
+ "group_size": 64,
1717
+ "bits": 4
1718
+ },
1719
+ "model.layers.4.self_attn.v_proj": {
1720
+ "group_size": 32,
1721
+ "bits": 8
1722
+ },
1723
+ "model.layers.4.self_attn.o_proj": {
1724
+ "group_size": 64,
1725
+ "bits": 4
1726
+ },
1727
+ "model.layers.4.mlp.gate": {
1728
+ "group_size": 64,
1729
+ "bits": 4
1730
+ },
1731
+ "model.layers.4.mlp.switch_mlp.gate_proj": {
1732
+ "group_size": 64,
1733
+ "bits": 4
1734
+ },
1735
+ "model.layers.4.mlp.switch_mlp.up_proj": {
1736
+ "group_size": 64,
1737
+ "bits": 4
1738
+ },
1739
+ "model.layers.4.mlp.switch_mlp.down_proj": {
1740
+ "group_size": 32,
1741
+ "bits": 6
1742
+ },
1743
+ "model.layers.5.self_attn.q_proj": {
1744
+ "group_size": 64,
1745
+ "bits": 4
1746
+ },
1747
+ "model.layers.5.self_attn.k_proj": {
1748
+ "group_size": 64,
1749
+ "bits": 4
1750
+ },
1751
+ "model.layers.5.self_attn.v_proj": {
1752
+ "group_size": 32,
1753
+ "bits": 8
1754
+ },
1755
+ "model.layers.5.self_attn.o_proj": {
1756
+ "group_size": 64,
1757
+ "bits": 4
1758
+ },
1759
+ "model.layers.5.mlp.gate": {
1760
+ "group_size": 64,
1761
+ "bits": 4
1762
+ },
1763
+ "model.layers.5.mlp.switch_mlp.gate_proj": {
1764
+ "group_size": 64,
1765
+ "bits": 4
1766
+ },
1767
+ "model.layers.5.mlp.switch_mlp.up_proj": {
1768
+ "group_size": 64,
1769
+ "bits": 4
1770
+ },
1771
+ "model.layers.5.mlp.switch_mlp.down_proj": {
1772
+ "group_size": 32,
1773
+ "bits": 6
1774
+ },
1775
+ "model.layers.6.self_attn.q_proj": {
1776
+ "group_size": 64,
1777
+ "bits": 4
1778
+ },
1779
+ "model.layers.6.self_attn.k_proj": {
1780
+ "group_size": 64,
1781
+ "bits": 4
1782
+ },
1783
+ "model.layers.6.self_attn.v_proj": {
1784
+ "group_size": 64,
1785
+ "bits": 4
1786
+ },
1787
+ "model.layers.6.self_attn.o_proj": {
1788
+ "group_size": 64,
1789
+ "bits": 4
1790
+ },
1791
+ "model.layers.6.mlp.gate": {
1792
+ "group_size": 64,
1793
+ "bits": 4
1794
+ },
1795
+ "model.layers.6.mlp.switch_mlp.gate_proj": {
1796
+ "group_size": 64,
1797
+ "bits": 4
1798
+ },
1799
+ "model.layers.6.mlp.switch_mlp.up_proj": {
1800
+ "group_size": 64,
1801
+ "bits": 4
1802
+ },
1803
+ "model.layers.6.mlp.switch_mlp.down_proj": {
1804
+ "group_size": 64,
1805
+ "bits": 4
1806
+ },
1807
+ "model.layers.7.self_attn.q_proj": {
1808
+ "group_size": 64,
1809
+ "bits": 4
1810
+ },
1811
+ "model.layers.7.self_attn.k_proj": {
1812
+ "group_size": 64,
1813
+ "bits": 4
1814
+ },
1815
+ "model.layers.7.self_attn.v_proj": {
1816
+ "group_size": 64,
1817
+ "bits": 4
1818
+ },
1819
+ "model.layers.7.self_attn.o_proj": {
1820
+ "group_size": 64,
1821
+ "bits": 4
1822
+ },
1823
+ "model.layers.7.mlp.gate": {
1824
+ "group_size": 64,
1825
+ "bits": 4
1826
+ },
1827
+ "model.layers.7.mlp.switch_mlp.gate_proj": {
1828
+ "group_size": 64,
1829
+ "bits": 4
1830
+ },
1831
+ "model.layers.7.mlp.switch_mlp.up_proj": {
1832
+ "group_size": 64,
1833
+ "bits": 4
1834
+ },
1835
+ "model.layers.7.mlp.switch_mlp.down_proj": {
1836
+ "group_size": 64,
1837
+ "bits": 4
1838
+ },
1839
+ "model.layers.8.self_attn.q_proj": {
1840
+ "group_size": 64,
1841
+ "bits": 4
1842
+ },
1843
+ "model.layers.8.self_attn.k_proj": {
1844
+ "group_size": 64,
1845
+ "bits": 4
1846
+ },
1847
+ "model.layers.8.self_attn.v_proj": {
1848
+ "group_size": 32,
1849
+ "bits": 6
1850
+ },
1851
+ "model.layers.8.self_attn.o_proj": {
1852
+ "group_size": 64,
1853
+ "bits": 4
1854
+ },
1855
+ "model.layers.8.mlp.gate": {
1856
+ "group_size": 64,
1857
+ "bits": 4
1858
+ },
1859
+ "model.layers.8.mlp.switch_mlp.gate_proj": {
1860
+ "group_size": 64,
1861
+ "bits": 4
1862
+ },
1863
+ "model.layers.8.mlp.switch_mlp.up_proj": {
1864
+ "group_size": 64,
1865
+ "bits": 4
1866
+ },
1867
+ "model.layers.8.mlp.switch_mlp.down_proj": {
1868
+ "group_size": 32,
1869
+ "bits": 6
1870
+ },
1871
+ "model.layers.9.self_attn.q_proj": {
1872
+ "group_size": 64,
1873
+ "bits": 4
1874
+ },
1875
+ "model.layers.9.self_attn.k_proj": {
1876
+ "group_size": 64,
1877
+ "bits": 4
1878
+ },
1879
+ "model.layers.9.self_attn.v_proj": {
1880
+ "group_size": 64,
1881
+ "bits": 4
1882
+ },
1883
+ "model.layers.9.self_attn.o_proj": {
1884
+ "group_size": 64,
1885
+ "bits": 4
1886
+ },
1887
+ "model.layers.9.mlp.gate": {
1888
+ "group_size": 64,
1889
+ "bits": 4
1890
+ },
1891
+ "model.layers.9.mlp.switch_mlp.gate_proj": {
1892
+ "group_size": 64,
1893
+ "bits": 4
1894
+ },
1895
+ "model.layers.9.mlp.switch_mlp.up_proj": {
1896
+ "group_size": 64,
1897
+ "bits": 4
1898
+ },
1899
+ "model.layers.9.mlp.switch_mlp.down_proj": {
1900
+ "group_size": 64,
1901
+ "bits": 4
1902
+ },
1903
+ "model.layers.10.self_attn.q_proj": {
1904
+ "group_size": 64,
1905
+ "bits": 4
1906
+ },
1907
+ "model.layers.10.self_attn.k_proj": {
1908
+ "group_size": 64,
1909
+ "bits": 4
1910
+ },
1911
+ "model.layers.10.self_attn.v_proj": {
1912
+ "group_size": 64,
1913
+ "bits": 4
1914
+ },
1915
+ "model.layers.10.self_attn.o_proj": {
1916
+ "group_size": 64,
1917
+ "bits": 4
1918
+ },
1919
+ "model.layers.10.mlp.gate": {
1920
+ "group_size": 64,
1921
+ "bits": 4
1922
+ },
1923
+ "model.layers.10.mlp.switch_mlp.gate_proj": {
1924
+ "group_size": 64,
1925
+ "bits": 4
1926
+ },
1927
+ "model.layers.10.mlp.switch_mlp.up_proj": {
1928
+ "group_size": 64,
1929
+ "bits": 4
1930
+ },
1931
+ "model.layers.10.mlp.switch_mlp.down_proj": {
1932
+ "group_size": 64,
1933
+ "bits": 4
1934
+ },
1935
+ "model.layers.11.self_attn.q_proj": {
1936
+ "group_size": 64,
1937
+ "bits": 4
1938
+ },
1939
+ "model.layers.11.self_attn.k_proj": {
1940
+ "group_size": 64,
1941
+ "bits": 4
1942
+ },
1943
+ "model.layers.11.self_attn.v_proj": {
1944
+ "group_size": 32,
1945
+ "bits": 6
1946
+ },
1947
+ "model.layers.11.self_attn.o_proj": {
1948
+ "group_size": 64,
1949
+ "bits": 4
1950
+ },
1951
+ "model.layers.11.mlp.gate": {
1952
+ "group_size": 64,
1953
+ "bits": 4
1954
+ },
1955
+ "model.layers.11.mlp.switch_mlp.gate_proj": {
1956
+ "group_size": 64,
1957
+ "bits": 4
1958
+ },
1959
+ "model.layers.11.mlp.switch_mlp.up_proj": {
1960
+ "group_size": 64,
1961
+ "bits": 4
1962
+ },
1963
+ "model.layers.11.mlp.switch_mlp.down_proj": {
1964
+ "group_size": 32,
1965
+ "bits": 6
1966
+ },
1967
+ "model.layers.12.self_attn.q_proj": {
1968
+ "group_size": 64,
1969
+ "bits": 4
1970
+ },
1971
+ "model.layers.12.self_attn.k_proj": {
1972
+ "group_size": 64,
1973
+ "bits": 4
1974
+ },
1975
+ "model.layers.12.self_attn.v_proj": {
1976
+ "group_size": 64,
1977
+ "bits": 4
1978
+ },
1979
+ "model.layers.12.self_attn.o_proj": {
1980
+ "group_size": 64,
1981
+ "bits": 4
1982
+ },
1983
+ "model.layers.12.mlp.gate": {
1984
+ "group_size": 64,
1985
+ "bits": 4
1986
+ },
1987
+ "model.layers.12.mlp.switch_mlp.gate_proj": {
1988
+ "group_size": 64,
1989
+ "bits": 4
1990
+ },
1991
+ "model.layers.12.mlp.switch_mlp.up_proj": {
1992
+ "group_size": 64,
1993
+ "bits": 4
1994
+ },
1995
+ "model.layers.12.mlp.switch_mlp.down_proj": {
1996
+ "group_size": 64,
1997
+ "bits": 4
1998
+ },
1999
+ "model.layers.13.self_attn.q_proj": {
2000
+ "group_size": 64,
2001
+ "bits": 4
2002
+ },
2003
+ "model.layers.13.self_attn.k_proj": {
2004
+ "group_size": 64,
2005
+ "bits": 4
2006
+ },
2007
+ "model.layers.13.self_attn.v_proj": {
2008
+ "group_size": 64,
2009
+ "bits": 4
2010
+ },
2011
+ "model.layers.13.self_attn.o_proj": {
2012
+ "group_size": 64,
2013
+ "bits": 4
2014
+ },
2015
+ "model.layers.13.mlp.gate": {
2016
+ "group_size": 64,
2017
+ "bits": 4
2018
+ },
2019
+ "model.layers.13.mlp.switch_mlp.gate_proj": {
2020
+ "group_size": 64,
2021
+ "bits": 4
2022
+ },
2023
+ "model.layers.13.mlp.switch_mlp.up_proj": {
2024
+ "group_size": 64,
2025
+ "bits": 4
2026
+ },
2027
+ "model.layers.13.mlp.switch_mlp.down_proj": {
2028
+ "group_size": 64,
2029
+ "bits": 4
2030
+ },
2031
+ "model.layers.14.self_attn.q_proj": {
2032
+ "group_size": 64,
2033
+ "bits": 4
2034
+ },
2035
+ "model.layers.14.self_attn.k_proj": {
2036
+ "group_size": 64,
2037
+ "bits": 4
2038
+ },
2039
+ "model.layers.14.self_attn.v_proj": {
2040
+ "group_size": 32,
2041
+ "bits": 6
2042
+ },
2043
+ "model.layers.14.self_attn.o_proj": {
2044
+ "group_size": 64,
2045
+ "bits": 4
2046
+ },
2047
+ "model.layers.14.mlp.gate": {
2048
+ "group_size": 64,
2049
+ "bits": 4
2050
+ },
2051
+ "model.layers.14.mlp.switch_mlp.gate_proj": {
2052
+ "group_size": 64,
2053
+ "bits": 4
2054
+ },
2055
+ "model.layers.14.mlp.switch_mlp.up_proj": {
2056
+ "group_size": 64,
2057
+ "bits": 4
2058
+ },
2059
+ "model.layers.14.mlp.switch_mlp.down_proj": {
2060
+ "group_size": 32,
2061
+ "bits": 6
2062
+ },
2063
+ "model.layers.15.self_attn.q_proj": {
2064
+ "group_size": 64,
2065
+ "bits": 4
2066
+ },
2067
+ "model.layers.15.self_attn.k_proj": {
2068
+ "group_size": 64,
2069
+ "bits": 4
2070
+ },
2071
+ "model.layers.15.self_attn.v_proj": {
2072
+ "group_size": 64,
2073
+ "bits": 4
2074
+ },
2075
+ "model.layers.15.self_attn.o_proj": {
2076
+ "group_size": 64,
2077
+ "bits": 4
2078
+ },
2079
+ "model.layers.15.mlp.gate": {
2080
+ "group_size": 64,
2081
+ "bits": 4
2082
+ },
2083
+ "model.layers.15.mlp.switch_mlp.gate_proj": {
2084
+ "group_size": 64,
2085
+ "bits": 4
2086
+ },
2087
+ "model.layers.15.mlp.switch_mlp.up_proj": {
2088
+ "group_size": 64,
2089
+ "bits": 4
2090
+ },
2091
+ "model.layers.15.mlp.switch_mlp.down_proj": {
2092
+ "group_size": 64,
2093
+ "bits": 4
2094
+ },
2095
+ "model.layers.16.self_attn.q_proj": {
2096
+ "group_size": 64,
2097
+ "bits": 4
2098
+ },
2099
+ "model.layers.16.self_attn.k_proj": {
2100
+ "group_size": 64,
2101
+ "bits": 4
2102
+ },
2103
+ "model.layers.16.self_attn.v_proj": {
2104
+ "group_size": 64,
2105
+ "bits": 4
2106
+ },
2107
+ "model.layers.16.self_attn.o_proj": {
2108
+ "group_size": 64,
2109
+ "bits": 4
2110
+ },
2111
+ "model.layers.16.mlp.gate": {
2112
+ "group_size": 64,
2113
+ "bits": 4
2114
+ },
2115
+ "model.layers.16.mlp.switch_mlp.gate_proj": {
2116
+ "group_size": 64,
2117
+ "bits": 4
2118
+ },
2119
+ "model.layers.16.mlp.switch_mlp.up_proj": {
2120
+ "group_size": 64,
2121
+ "bits": 4
2122
+ },
2123
+ "model.layers.16.mlp.switch_mlp.down_proj": {
2124
+ "group_size": 64,
2125
+ "bits": 4
2126
+ },
2127
+ "model.layers.17.self_attn.q_proj": {
2128
+ "group_size": 64,
2129
+ "bits": 4
2130
+ },
2131
+ "model.layers.17.self_attn.k_proj": {
2132
+ "group_size": 64,
2133
+ "bits": 4
2134
+ },
2135
+ "model.layers.17.self_attn.v_proj": {
2136
+ "group_size": 32,
2137
+ "bits": 6
2138
+ },
2139
+ "model.layers.17.self_attn.o_proj": {
2140
+ "group_size": 64,
2141
+ "bits": 4
2142
+ },
2143
+ "model.layers.17.mlp.gate": {
2144
+ "group_size": 64,
2145
+ "bits": 4
2146
+ },
2147
+ "model.layers.17.mlp.switch_mlp.gate_proj": {
2148
+ "group_size": 64,
2149
+ "bits": 4
2150
+ },
2151
+ "model.layers.17.mlp.switch_mlp.up_proj": {
2152
+ "group_size": 64,
2153
+ "bits": 4
2154
+ },
2155
+ "model.layers.17.mlp.switch_mlp.down_proj": {
2156
+ "group_size": 32,
2157
+ "bits": 6
2158
+ },
2159
+ "model.layers.18.self_attn.q_proj": {
2160
+ "group_size": 64,
2161
+ "bits": 4
2162
+ },
2163
+ "model.layers.18.self_attn.k_proj": {
2164
+ "group_size": 64,
2165
+ "bits": 4
2166
+ },
2167
+ "model.layers.18.self_attn.v_proj": {
2168
+ "group_size": 64,
2169
+ "bits": 4
2170
+ },
2171
+ "model.layers.18.self_attn.o_proj": {
2172
+ "group_size": 64,
2173
+ "bits": 4
2174
+ },
2175
+ "model.layers.18.mlp.gate": {
2176
+ "group_size": 64,
2177
+ "bits": 4
2178
+ },
2179
+ "model.layers.18.mlp.switch_mlp.gate_proj": {
2180
+ "group_size": 64,
2181
+ "bits": 4
2182
+ },
2183
+ "model.layers.18.mlp.switch_mlp.up_proj": {
2184
+ "group_size": 64,
2185
+ "bits": 4
2186
+ },
2187
+ "model.layers.18.mlp.switch_mlp.down_proj": {
2188
+ "group_size": 64,
2189
+ "bits": 4
2190
+ },
2191
+ "model.layers.19.self_attn.q_proj": {
2192
+ "group_size": 64,
2193
+ "bits": 4
2194
+ },
2195
+ "model.layers.19.self_attn.k_proj": {
2196
+ "group_size": 64,
2197
+ "bits": 4
2198
+ },
2199
+ "model.layers.19.self_attn.v_proj": {
2200
+ "group_size": 64,
2201
+ "bits": 4
2202
+ },
2203
+ "model.layers.19.self_attn.o_proj": {
2204
+ "group_size": 64,
2205
+ "bits": 4
2206
+ },
2207
+ "model.layers.19.mlp.gate": {
2208
+ "group_size": 64,
2209
+ "bits": 4
2210
+ },
2211
+ "model.layers.19.mlp.switch_mlp.gate_proj": {
2212
+ "group_size": 64,
2213
+ "bits": 4
2214
+ },
2215
+ "model.layers.19.mlp.switch_mlp.up_proj": {
2216
+ "group_size": 64,
2217
+ "bits": 4
2218
+ },
2219
+ "model.layers.19.mlp.switch_mlp.down_proj": {
2220
+ "group_size": 64,
2221
+ "bits": 4
2222
+ },
2223
+ "model.layers.20.self_attn.q_proj": {
2224
+ "group_size": 64,
2225
+ "bits": 4
2226
+ },
2227
+ "model.layers.20.self_attn.k_proj": {
2228
+ "group_size": 64,
2229
+ "bits": 4
2230
+ },
2231
+ "model.layers.20.self_attn.v_proj": {
2232
+ "group_size": 32,
2233
+ "bits": 6
2234
+ },
2235
+ "model.layers.20.self_attn.o_proj": {
2236
+ "group_size": 64,
2237
+ "bits": 4
2238
+ },
2239
+ "model.layers.20.mlp.gate": {
2240
+ "group_size": 64,
2241
+ "bits": 4
2242
+ },
2243
+ "model.layers.20.mlp.switch_mlp.gate_proj": {
2244
+ "group_size": 64,
2245
+ "bits": 4
2246
+ },
2247
+ "model.layers.20.mlp.switch_mlp.up_proj": {
2248
+ "group_size": 64,
2249
+ "bits": 4
2250
+ },
2251
+ "model.layers.20.mlp.switch_mlp.down_proj": {
2252
+ "group_size": 32,
2253
+ "bits": 6
2254
+ },
2255
+ "model.layers.21.self_attn.q_proj": {
2256
+ "group_size": 64,
2257
+ "bits": 4
2258
+ },
2259
+ "model.layers.21.self_attn.k_proj": {
2260
+ "group_size": 64,
2261
+ "bits": 4
2262
+ },
2263
+ "model.layers.21.self_attn.v_proj": {
2264
+ "group_size": 64,
2265
+ "bits": 4
2266
+ },
2267
+ "model.layers.21.self_attn.o_proj": {
2268
+ "group_size": 64,
2269
+ "bits": 4
2270
+ },
2271
+ "model.layers.21.mlp.gate": {
2272
+ "group_size": 64,
2273
+ "bits": 4
2274
+ },
2275
+ "model.layers.21.mlp.switch_mlp.gate_proj": {
2276
+ "group_size": 64,
2277
+ "bits": 4
2278
+ },
2279
+ "model.layers.21.mlp.switch_mlp.up_proj": {
2280
+ "group_size": 64,
2281
+ "bits": 4
2282
+ },
2283
+ "model.layers.21.mlp.switch_mlp.down_proj": {
2284
+ "group_size": 64,
2285
+ "bits": 4
2286
+ },
2287
+ "model.layers.22.self_attn.q_proj": {
2288
+ "group_size": 64,
2289
+ "bits": 4
2290
+ },
2291
+ "model.layers.22.self_attn.k_proj": {
2292
+ "group_size": 64,
2293
+ "bits": 4
2294
+ },
2295
+ "model.layers.22.self_attn.v_proj": {
2296
+ "group_size": 64,
2297
+ "bits": 4
2298
+ },
2299
+ "model.layers.22.self_attn.o_proj": {
2300
+ "group_size": 64,
2301
+ "bits": 4
2302
+ },
2303
+ "model.layers.22.mlp.gate": {
2304
+ "group_size": 64,
2305
+ "bits": 4
2306
+ },
2307
+ "model.layers.22.mlp.switch_mlp.gate_proj": {
2308
+ "group_size": 64,
2309
+ "bits": 4
2310
+ },
2311
+ "model.layers.22.mlp.switch_mlp.up_proj": {
2312
+ "group_size": 64,
2313
+ "bits": 4
2314
+ },
2315
+ "model.layers.22.mlp.switch_mlp.down_proj": {
2316
+ "group_size": 64,
2317
+ "bits": 4
2318
+ },
2319
+ "model.layers.23.self_attn.q_proj": {
2320
+ "group_size": 64,
2321
+ "bits": 4
2322
+ },
2323
+ "model.layers.23.self_attn.k_proj": {
2324
+ "group_size": 64,
2325
+ "bits": 4
2326
+ },
2327
+ "model.layers.23.self_attn.v_proj": {
2328
+ "group_size": 32,
2329
+ "bits": 6
2330
+ },
2331
+ "model.layers.23.self_attn.o_proj": {
2332
+ "group_size": 64,
2333
+ "bits": 4
2334
+ },
2335
+ "model.layers.23.mlp.gate": {
2336
+ "group_size": 64,
2337
+ "bits": 4
2338
+ },
2339
+ "model.layers.23.mlp.switch_mlp.gate_proj": {
2340
+ "group_size": 64,
2341
+ "bits": 4
2342
+ },
2343
+ "model.layers.23.mlp.switch_mlp.up_proj": {
2344
+ "group_size": 64,
2345
+ "bits": 4
2346
+ },
2347
+ "model.layers.23.mlp.switch_mlp.down_proj": {
2348
+ "group_size": 32,
2349
+ "bits": 6
2350
+ },
2351
+ "model.layers.24.self_attn.q_proj": {
2352
+ "group_size": 64,
2353
+ "bits": 4
2354
+ },
2355
+ "model.layers.24.self_attn.k_proj": {
2356
+ "group_size": 64,
2357
+ "bits": 4
2358
+ },
2359
+ "model.layers.24.self_attn.v_proj": {
2360
+ "group_size": 64,
2361
+ "bits": 4
2362
+ },
2363
+ "model.layers.24.self_attn.o_proj": {
2364
+ "group_size": 64,
2365
+ "bits": 4
2366
+ },
2367
+ "model.layers.24.mlp.gate": {
2368
+ "group_size": 64,
2369
+ "bits": 4
2370
+ },
2371
+ "model.layers.24.mlp.switch_mlp.gate_proj": {
2372
+ "group_size": 64,
2373
+ "bits": 4
2374
+ },
2375
+ "model.layers.24.mlp.switch_mlp.up_proj": {
2376
+ "group_size": 64,
2377
+ "bits": 4
2378
+ },
2379
+ "model.layers.24.mlp.switch_mlp.down_proj": {
2380
+ "group_size": 64,
2381
+ "bits": 4
2382
+ },
2383
+ "model.layers.25.self_attn.q_proj": {
2384
+ "group_size": 64,
2385
+ "bits": 4
2386
+ },
2387
+ "model.layers.25.self_attn.k_proj": {
2388
+ "group_size": 64,
2389
+ "bits": 4
2390
+ },
2391
+ "model.layers.25.self_attn.v_proj": {
2392
+ "group_size": 64,
2393
+ "bits": 4
2394
+ },
2395
+ "model.layers.25.self_attn.o_proj": {
2396
+ "group_size": 64,
2397
+ "bits": 4
2398
+ },
2399
+ "model.layers.25.mlp.gate": {
2400
+ "group_size": 64,
2401
+ "bits": 4
2402
+ },
2403
+ "model.layers.25.mlp.switch_mlp.gate_proj": {
2404
+ "group_size": 64,
2405
+ "bits": 4
2406
+ },
2407
+ "model.layers.25.mlp.switch_mlp.up_proj": {
2408
+ "group_size": 64,
2409
+ "bits": 4
2410
+ },
2411
+ "model.layers.25.mlp.switch_mlp.down_proj": {
2412
+ "group_size": 64,
2413
+ "bits": 4
2414
+ },
2415
+ "model.layers.26.self_attn.q_proj": {
2416
+ "group_size": 64,
2417
+ "bits": 4
2418
+ },
2419
+ "model.layers.26.self_attn.k_proj": {
2420
+ "group_size": 64,
2421
+ "bits": 4
2422
+ },
2423
+ "model.layers.26.self_attn.v_proj": {
2424
+ "group_size": 32,
2425
+ "bits": 6
2426
+ },
2427
+ "model.layers.26.self_attn.o_proj": {
2428
+ "group_size": 64,
2429
+ "bits": 4
2430
+ },
2431
+ "model.layers.26.mlp.gate": {
2432
+ "group_size": 64,
2433
+ "bits": 4
2434
+ },
2435
+ "model.layers.26.mlp.switch_mlp.gate_proj": {
2436
+ "group_size": 64,
2437
+ "bits": 4
2438
+ },
2439
+ "model.layers.26.mlp.switch_mlp.up_proj": {
2440
+ "group_size": 64,
2441
+ "bits": 4
2442
+ },
2443
+ "model.layers.26.mlp.switch_mlp.down_proj": {
2444
+ "group_size": 32,
2445
+ "bits": 6
2446
+ },
2447
+ "model.layers.27.self_attn.q_proj": {
2448
+ "group_size": 64,
2449
+ "bits": 4
2450
+ },
2451
+ "model.layers.27.self_attn.k_proj": {
2452
+ "group_size": 64,
2453
+ "bits": 4
2454
+ },
2455
+ "model.layers.27.self_attn.v_proj": {
2456
+ "group_size": 64,
2457
+ "bits": 4
2458
+ },
2459
+ "model.layers.27.self_attn.o_proj": {
2460
+ "group_size": 64,
2461
+ "bits": 4
2462
+ },
2463
+ "model.layers.27.mlp.gate": {
2464
+ "group_size": 64,
2465
+ "bits": 4
2466
+ },
2467
+ "model.layers.27.mlp.switch_mlp.gate_proj": {
2468
+ "group_size": 64,
2469
+ "bits": 4
2470
+ },
2471
+ "model.layers.27.mlp.switch_mlp.up_proj": {
2472
+ "group_size": 64,
2473
+ "bits": 4
2474
+ },
2475
+ "model.layers.27.mlp.switch_mlp.down_proj": {
2476
+ "group_size": 64,
2477
+ "bits": 4
2478
+ },
2479
+ "model.layers.28.self_attn.q_proj": {
2480
+ "group_size": 64,
2481
+ "bits": 4
2482
+ },
2483
+ "model.layers.28.self_attn.k_proj": {
2484
+ "group_size": 64,
2485
+ "bits": 4
2486
+ },
2487
+ "model.layers.28.self_attn.v_proj": {
2488
+ "group_size": 64,
2489
+ "bits": 4
2490
+ },
2491
+ "model.layers.28.self_attn.o_proj": {
2492
+ "group_size": 64,
2493
+ "bits": 4
2494
+ },
2495
+ "model.layers.28.mlp.gate": {
2496
+ "group_size": 64,
2497
+ "bits": 4
2498
+ },
2499
+ "model.layers.28.mlp.switch_mlp.gate_proj": {
2500
+ "group_size": 64,
2501
+ "bits": 4
2502
+ },
2503
+ "model.layers.28.mlp.switch_mlp.up_proj": {
2504
+ "group_size": 64,
2505
+ "bits": 4
2506
+ },
2507
+ "model.layers.28.mlp.switch_mlp.down_proj": {
2508
+ "group_size": 64,
2509
+ "bits": 4
2510
+ },
2511
+ "model.layers.29.self_attn.q_proj": {
2512
+ "group_size": 64,
2513
+ "bits": 4
2514
+ },
2515
+ "model.layers.29.self_attn.k_proj": {
2516
+ "group_size": 64,
2517
+ "bits": 4
2518
+ },
2519
+ "model.layers.29.self_attn.v_proj": {
2520
+ "group_size": 32,
2521
+ "bits": 6
2522
+ },
2523
+ "model.layers.29.self_attn.o_proj": {
2524
+ "group_size": 64,
2525
+ "bits": 4
2526
+ },
2527
+ "model.layers.29.mlp.gate": {
2528
+ "group_size": 64,
2529
+ "bits": 4
2530
+ },
2531
+ "model.layers.29.mlp.switch_mlp.gate_proj": {
2532
+ "group_size": 64,
2533
+ "bits": 4
2534
+ },
2535
+ "model.layers.29.mlp.switch_mlp.up_proj": {
2536
+ "group_size": 64,
2537
+ "bits": 4
2538
+ },
2539
+ "model.layers.29.mlp.switch_mlp.down_proj": {
2540
+ "group_size": 32,
2541
+ "bits": 6
2542
+ },
2543
+ "model.layers.30.self_attn.q_proj": {
2544
+ "group_size": 64,
2545
+ "bits": 4
2546
+ },
2547
+ "model.layers.30.self_attn.k_proj": {
2548
+ "group_size": 64,
2549
+ "bits": 4
2550
+ },
2551
+ "model.layers.30.self_attn.v_proj": {
2552
+ "group_size": 64,
2553
+ "bits": 4
2554
+ },
2555
+ "model.layers.30.self_attn.o_proj": {
2556
+ "group_size": 64,
2557
+ "bits": 4
2558
+ },
2559
+ "model.layers.30.mlp.gate": {
2560
+ "group_size": 64,
2561
+ "bits": 4
2562
+ },
2563
+ "model.layers.30.mlp.switch_mlp.gate_proj": {
2564
+ "group_size": 64,
2565
+ "bits": 4
2566
+ },
2567
+ "model.layers.30.mlp.switch_mlp.up_proj": {
2568
+ "group_size": 64,
2569
+ "bits": 4
2570
+ },
2571
+ "model.layers.30.mlp.switch_mlp.down_proj": {
2572
+ "group_size": 64,
2573
+ "bits": 4
2574
+ },
2575
+ "model.layers.31.self_attn.q_proj": {
2576
+ "group_size": 64,
2577
+ "bits": 4
2578
+ },
2579
+ "model.layers.31.self_attn.k_proj": {
2580
+ "group_size": 64,
2581
+ "bits": 4
2582
+ },
2583
+ "model.layers.31.self_attn.v_proj": {
2584
+ "group_size": 64,
2585
+ "bits": 4
2586
+ },
2587
+ "model.layers.31.self_attn.o_proj": {
2588
+ "group_size": 64,
2589
+ "bits": 4
2590
+ },
2591
+ "model.layers.31.mlp.gate": {
2592
+ "group_size": 64,
2593
+ "bits": 4
2594
+ },
2595
+ "model.layers.31.mlp.switch_mlp.gate_proj": {
2596
+ "group_size": 64,
2597
+ "bits": 4
2598
+ },
2599
+ "model.layers.31.mlp.switch_mlp.up_proj": {
2600
+ "group_size": 64,
2601
+ "bits": 4
2602
+ },
2603
+ "model.layers.31.mlp.switch_mlp.down_proj": {
2604
+ "group_size": 64,
2605
+ "bits": 4
2606
+ },
2607
+ "model.layers.32.self_attn.q_proj": {
2608
+ "group_size": 64,
2609
+ "bits": 4
2610
+ },
2611
+ "model.layers.32.self_attn.k_proj": {
2612
+ "group_size": 64,
2613
+ "bits": 4
2614
+ },
2615
+ "model.layers.32.self_attn.v_proj": {
2616
+ "group_size": 32,
2617
+ "bits": 6
2618
+ },
2619
+ "model.layers.32.self_attn.o_proj": {
2620
+ "group_size": 64,
2621
+ "bits": 4
2622
+ },
2623
+ "model.layers.32.mlp.gate": {
2624
+ "group_size": 64,
2625
+ "bits": 4
2626
+ },
2627
+ "model.layers.32.mlp.switch_mlp.gate_proj": {
2628
+ "group_size": 64,
2629
+ "bits": 4
2630
+ },
2631
+ "model.layers.32.mlp.switch_mlp.up_proj": {
2632
+ "group_size": 64,
2633
+ "bits": 4
2634
+ },
2635
+ "model.layers.32.mlp.switch_mlp.down_proj": {
2636
+ "group_size": 32,
2637
+ "bits": 6
2638
+ },
2639
+ "model.layers.33.self_attn.q_proj": {
2640
+ "group_size": 64,
2641
+ "bits": 4
2642
+ },
2643
+ "model.layers.33.self_attn.k_proj": {
2644
+ "group_size": 64,
2645
+ "bits": 4
2646
+ },
2647
+ "model.layers.33.self_attn.v_proj": {
2648
+ "group_size": 64,
2649
+ "bits": 4
2650
+ },
2651
+ "model.layers.33.self_attn.o_proj": {
2652
+ "group_size": 64,
2653
+ "bits": 4
2654
+ },
2655
+ "model.layers.33.mlp.gate": {
2656
+ "group_size": 64,
2657
+ "bits": 4
2658
+ },
2659
+ "model.layers.33.mlp.switch_mlp.gate_proj": {
2660
+ "group_size": 64,
2661
+ "bits": 4
2662
+ },
2663
+ "model.layers.33.mlp.switch_mlp.up_proj": {
2664
+ "group_size": 64,
2665
+ "bits": 4
2666
+ },
2667
+ "model.layers.33.mlp.switch_mlp.down_proj": {
2668
+ "group_size": 64,
2669
+ "bits": 4
2670
+ },
2671
+ "model.layers.34.self_attn.q_proj": {
2672
+ "group_size": 64,
2673
+ "bits": 4
2674
+ },
2675
+ "model.layers.34.self_attn.k_proj": {
2676
+ "group_size": 64,
2677
+ "bits": 4
2678
+ },
2679
+ "model.layers.34.self_attn.v_proj": {
2680
+ "group_size": 64,
2681
+ "bits": 4
2682
+ },
2683
+ "model.layers.34.self_attn.o_proj": {
2684
+ "group_size": 64,
2685
+ "bits": 4
2686
+ },
2687
+ "model.layers.34.mlp.gate": {
2688
+ "group_size": 64,
2689
+ "bits": 4
2690
+ },
2691
+ "model.layers.34.mlp.switch_mlp.gate_proj": {
2692
+ "group_size": 64,
2693
+ "bits": 4
2694
+ },
2695
+ "model.layers.34.mlp.switch_mlp.up_proj": {
2696
+ "group_size": 64,
2697
+ "bits": 4
2698
+ },
2699
+ "model.layers.34.mlp.switch_mlp.down_proj": {
2700
+ "group_size": 64,
2701
+ "bits": 4
2702
+ },
2703
+ "model.layers.35.self_attn.q_proj": {
2704
+ "group_size": 64,
2705
+ "bits": 4
2706
+ },
2707
+ "model.layers.35.self_attn.k_proj": {
2708
+ "group_size": 64,
2709
+ "bits": 4
2710
+ },
2711
+ "model.layers.35.self_attn.v_proj": {
2712
+ "group_size": 32,
2713
+ "bits": 6
2714
+ },
2715
+ "model.layers.35.self_attn.o_proj": {
2716
+ "group_size": 64,
2717
+ "bits": 4
2718
+ },
2719
+ "model.layers.35.mlp.gate": {
2720
+ "group_size": 64,
2721
+ "bits": 4
2722
+ },
2723
+ "model.layers.35.mlp.switch_mlp.gate_proj": {
2724
+ "group_size": 64,
2725
+ "bits": 4
2726
+ },
2727
+ "model.layers.35.mlp.switch_mlp.up_proj": {
2728
+ "group_size": 64,
2729
+ "bits": 4
2730
+ },
2731
+ "model.layers.35.mlp.switch_mlp.down_proj": {
2732
+ "group_size": 32,
2733
+ "bits": 6
2734
+ },
2735
+ "model.layers.36.self_attn.q_proj": {
2736
+ "group_size": 64,
2737
+ "bits": 4
2738
+ },
2739
+ "model.layers.36.self_attn.k_proj": {
2740
+ "group_size": 64,
2741
+ "bits": 4
2742
+ },
2743
+ "model.layers.36.self_attn.v_proj": {
2744
+ "group_size": 64,
2745
+ "bits": 4
2746
+ },
2747
+ "model.layers.36.self_attn.o_proj": {
2748
+ "group_size": 64,
2749
+ "bits": 4
2750
+ },
2751
+ "model.layers.36.mlp.gate": {
2752
+ "group_size": 64,
2753
+ "bits": 4
2754
+ },
2755
+ "model.layers.36.mlp.switch_mlp.gate_proj": {
2756
+ "group_size": 64,
2757
+ "bits": 4
2758
+ },
2759
+ "model.layers.36.mlp.switch_mlp.up_proj": {
2760
+ "group_size": 64,
2761
+ "bits": 4
2762
+ },
2763
+ "model.layers.36.mlp.switch_mlp.down_proj": {
2764
+ "group_size": 64,
2765
+ "bits": 4
2766
+ },
2767
+ "model.layers.37.self_attn.q_proj": {
2768
+ "group_size": 64,
2769
+ "bits": 4
2770
+ },
2771
+ "model.layers.37.self_attn.k_proj": {
2772
+ "group_size": 64,
2773
+ "bits": 4
2774
+ },
2775
+ "model.layers.37.self_attn.v_proj": {
2776
+ "group_size": 64,
2777
+ "bits": 4
2778
+ },
2779
+ "model.layers.37.self_attn.o_proj": {
2780
+ "group_size": 64,
2781
+ "bits": 4
2782
+ },
2783
+ "model.layers.37.mlp.gate": {
2784
+ "group_size": 64,
2785
+ "bits": 4
2786
+ },
2787
+ "model.layers.37.mlp.switch_mlp.gate_proj": {
2788
+ "group_size": 64,
2789
+ "bits": 4
2790
+ },
2791
+ "model.layers.37.mlp.switch_mlp.up_proj": {
2792
+ "group_size": 64,
2793
+ "bits": 4
2794
+ },
2795
+ "model.layers.37.mlp.switch_mlp.down_proj": {
2796
+ "group_size": 64,
2797
+ "bits": 4
2798
+ },
2799
+ "model.layers.38.self_attn.q_proj": {
2800
+ "group_size": 64,
2801
+ "bits": 4
2802
+ },
2803
+ "model.layers.38.self_attn.k_proj": {
2804
+ "group_size": 64,
2805
+ "bits": 4
2806
+ },
2807
+ "model.layers.38.self_attn.v_proj": {
2808
+ "group_size": 32,
2809
+ "bits": 6
2810
+ },
2811
+ "model.layers.38.self_attn.o_proj": {
2812
+ "group_size": 64,
2813
+ "bits": 4
2814
+ },
2815
+ "model.layers.38.mlp.gate": {
2816
+ "group_size": 64,
2817
+ "bits": 4
2818
+ },
2819
+ "model.layers.38.mlp.switch_mlp.gate_proj": {
2820
+ "group_size": 64,
2821
+ "bits": 4
2822
+ },
2823
+ "model.layers.38.mlp.switch_mlp.up_proj": {
2824
+ "group_size": 64,
2825
+ "bits": 4
2826
+ },
2827
+ "model.layers.38.mlp.switch_mlp.down_proj": {
2828
+ "group_size": 32,
2829
+ "bits": 6
2830
+ },
2831
+ "model.layers.39.self_attn.q_proj": {
2832
+ "group_size": 64,
2833
+ "bits": 4
2834
+ },
2835
+ "model.layers.39.self_attn.k_proj": {
2836
+ "group_size": 64,
2837
+ "bits": 4
2838
+ },
2839
+ "model.layers.39.self_attn.v_proj": {
2840
+ "group_size": 64,
2841
+ "bits": 4
2842
+ },
2843
+ "model.layers.39.self_attn.o_proj": {
2844
+ "group_size": 64,
2845
+ "bits": 4
2846
+ },
2847
+ "model.layers.39.mlp.gate": {
2848
+ "group_size": 64,
2849
+ "bits": 4
2850
+ },
2851
+ "model.layers.39.mlp.switch_mlp.gate_proj": {
2852
+ "group_size": 64,
2853
+ "bits": 4
2854
+ },
2855
+ "model.layers.39.mlp.switch_mlp.up_proj": {
2856
+ "group_size": 64,
2857
+ "bits": 4
2858
+ },
2859
+ "model.layers.39.mlp.switch_mlp.down_proj": {
2860
+ "group_size": 64,
2861
+ "bits": 4
2862
+ },
2863
+ "model.layers.40.self_attn.q_proj": {
2864
+ "group_size": 64,
2865
+ "bits": 4
2866
+ },
2867
+ "model.layers.40.self_attn.k_proj": {
2868
+ "group_size": 64,
2869
+ "bits": 4
2870
+ },
2871
+ "model.layers.40.self_attn.v_proj": {
2872
+ "group_size": 64,
2873
+ "bits": 4
2874
+ },
2875
+ "model.layers.40.self_attn.o_proj": {
2876
+ "group_size": 64,
2877
+ "bits": 4
2878
+ },
2879
+ "model.layers.40.mlp.gate": {
2880
+ "group_size": 64,
2881
+ "bits": 4
2882
+ },
2883
+ "model.layers.40.mlp.switch_mlp.gate_proj": {
2884
+ "group_size": 64,
2885
+ "bits": 4
2886
+ },
2887
+ "model.layers.40.mlp.switch_mlp.up_proj": {
2888
+ "group_size": 64,
2889
+ "bits": 4
2890
+ },
2891
+ "model.layers.40.mlp.switch_mlp.down_proj": {
2892
+ "group_size": 64,
2893
+ "bits": 4
2894
+ },
2895
+ "model.layers.41.self_attn.q_proj": {
2896
+ "group_size": 64,
2897
+ "bits": 4
2898
+ },
2899
+ "model.layers.41.self_attn.k_proj": {
2900
+ "group_size": 64,
2901
+ "bits": 4
2902
+ },
2903
+ "model.layers.41.self_attn.v_proj": {
2904
+ "group_size": 32,
2905
+ "bits": 6
2906
+ },
2907
+ "model.layers.41.self_attn.o_proj": {
2908
+ "group_size": 64,
2909
+ "bits": 4
2910
+ },
2911
+ "model.layers.41.mlp.gate": {
2912
+ "group_size": 64,
2913
+ "bits": 4
2914
+ },
2915
+ "model.layers.41.mlp.switch_mlp.gate_proj": {
2916
+ "group_size": 64,
2917
+ "bits": 4
2918
+ },
2919
+ "model.layers.41.mlp.switch_mlp.up_proj": {
2920
+ "group_size": 64,
2921
+ "bits": 4
2922
+ },
2923
+ "model.layers.41.mlp.switch_mlp.down_proj": {
2924
+ "group_size": 32,
2925
+ "bits": 6
2926
+ },
2927
+ "model.layers.42.self_attn.q_proj": {
2928
+ "group_size": 64,
2929
+ "bits": 4
2930
+ },
2931
+ "model.layers.42.self_attn.k_proj": {
2932
+ "group_size": 64,
2933
+ "bits": 4
2934
+ },
2935
+ "model.layers.42.self_attn.v_proj": {
2936
+ "group_size": 32,
2937
+ "bits": 6
2938
+ },
2939
+ "model.layers.42.self_attn.o_proj": {
2940
+ "group_size": 64,
2941
+ "bits": 4
2942
+ },
2943
+ "model.layers.42.mlp.gate": {
2944
+ "group_size": 64,
2945
+ "bits": 4
2946
+ },
2947
+ "model.layers.42.mlp.switch_mlp.gate_proj": {
2948
+ "group_size": 64,
2949
+ "bits": 4
2950
+ },
2951
+ "model.layers.42.mlp.switch_mlp.up_proj": {
2952
+ "group_size": 64,
2953
+ "bits": 4
2954
+ },
2955
+ "model.layers.42.mlp.switch_mlp.down_proj": {
2956
+ "group_size": 32,
2957
+ "bits": 6
2958
+ },
2959
+ "model.layers.43.self_attn.q_proj": {
2960
+ "group_size": 64,
2961
+ "bits": 4
2962
+ },
2963
+ "model.layers.43.self_attn.k_proj": {
2964
+ "group_size": 64,
2965
+ "bits": 4
2966
+ },
2967
+ "model.layers.43.self_attn.v_proj": {
2968
+ "group_size": 32,
2969
+ "bits": 6
2970
+ },
2971
+ "model.layers.43.self_attn.o_proj": {
2972
+ "group_size": 64,
2973
+ "bits": 4
2974
+ },
2975
+ "model.layers.43.mlp.gate": {
2976
+ "group_size": 64,
2977
+ "bits": 4
2978
+ },
2979
+ "model.layers.43.mlp.switch_mlp.gate_proj": {
2980
+ "group_size": 64,
2981
+ "bits": 4
2982
+ },
2983
+ "model.layers.43.mlp.switch_mlp.up_proj": {
2984
+ "group_size": 64,
2985
+ "bits": 4
2986
+ },
2987
+ "model.layers.43.mlp.switch_mlp.down_proj": {
2988
+ "group_size": 32,
2989
+ "bits": 6
2990
+ },
2991
+ "model.layers.44.self_attn.q_proj": {
2992
+ "group_size": 64,
2993
+ "bits": 4
2994
+ },
2995
+ "model.layers.44.self_attn.k_proj": {
2996
+ "group_size": 64,
2997
+ "bits": 4
2998
+ },
2999
+ "model.layers.44.self_attn.v_proj": {
3000
+ "group_size": 32,
3001
+ "bits": 6
3002
+ },
3003
+ "model.layers.44.self_attn.o_proj": {
3004
+ "group_size": 64,
3005
+ "bits": 4
3006
+ },
3007
+ "model.layers.44.mlp.gate": {
3008
+ "group_size": 64,
3009
+ "bits": 4
3010
+ },
3011
+ "model.layers.44.mlp.switch_mlp.gate_proj": {
3012
+ "group_size": 64,
3013
+ "bits": 4
3014
+ },
3015
+ "model.layers.44.mlp.switch_mlp.up_proj": {
3016
+ "group_size": 64,
3017
+ "bits": 4
3018
+ },
3019
+ "model.layers.44.mlp.switch_mlp.down_proj": {
3020
+ "group_size": 32,
3021
+ "bits": 6
3022
+ },
3023
+ "model.layers.45.self_attn.q_proj": {
3024
+ "group_size": 64,
3025
+ "bits": 4
3026
+ },
3027
+ "model.layers.45.self_attn.k_proj": {
3028
+ "group_size": 64,
3029
+ "bits": 4
3030
+ },
3031
+ "model.layers.45.self_attn.v_proj": {
3032
+ "group_size": 32,
3033
+ "bits": 6
3034
+ },
3035
+ "model.layers.45.self_attn.o_proj": {
3036
+ "group_size": 64,
3037
+ "bits": 4
3038
+ },
3039
+ "model.layers.45.mlp.gate": {
3040
+ "group_size": 64,
3041
+ "bits": 4
3042
+ },
3043
+ "model.layers.45.mlp.switch_mlp.gate_proj": {
3044
+ "group_size": 64,
3045
+ "bits": 4
3046
+ },
3047
+ "model.layers.45.mlp.switch_mlp.up_proj": {
3048
+ "group_size": 64,
3049
+ "bits": 4
3050
+ },
3051
+ "model.layers.45.mlp.switch_mlp.down_proj": {
3052
+ "group_size": 32,
3053
+ "bits": 6
3054
+ },
3055
+ "model.layers.46.self_attn.q_proj": {
3056
+ "group_size": 64,
3057
+ "bits": 4
3058
+ },
3059
+ "model.layers.46.self_attn.k_proj": {
3060
+ "group_size": 64,
3061
+ "bits": 4
3062
+ },
3063
+ "model.layers.46.self_attn.v_proj": {
3064
+ "group_size": 32,
3065
+ "bits": 6
3066
+ },
3067
+ "model.layers.46.self_attn.o_proj": {
3068
+ "group_size": 64,
3069
+ "bits": 4
3070
+ },
3071
+ "model.layers.46.mlp.gate": {
3072
+ "group_size": 64,
3073
+ "bits": 4
3074
+ },
3075
+ "model.layers.46.mlp.switch_mlp.gate_proj": {
3076
+ "group_size": 64,
3077
+ "bits": 4
3078
+ },
3079
+ "model.layers.46.mlp.switch_mlp.up_proj": {
3080
+ "group_size": 64,
3081
+ "bits": 4
3082
+ },
3083
+ "model.layers.46.mlp.switch_mlp.down_proj": {
3084
+ "group_size": 32,
3085
+ "bits": 6
3086
+ },
3087
+ "model.layers.47.self_attn.q_proj": {
3088
+ "group_size": 64,
3089
+ "bits": 4
3090
+ },
3091
+ "model.layers.47.self_attn.k_proj": {
3092
+ "group_size": 64,
3093
+ "bits": 4
3094
+ },
3095
+ "model.layers.47.self_attn.v_proj": {
3096
+ "group_size": 32,
3097
+ "bits": 6
3098
+ },
3099
+ "model.layers.47.self_attn.o_proj": {
3100
+ "group_size": 64,
3101
+ "bits": 4
3102
+ },
3103
+ "model.layers.47.mlp.gate": {
3104
+ "group_size": 64,
3105
+ "bits": 4
3106
+ },
3107
+ "model.layers.47.mlp.switch_mlp.gate_proj": {
3108
+ "group_size": 64,
3109
+ "bits": 4
3110
+ },
3111
+ "model.layers.47.mlp.switch_mlp.up_proj": {
3112
+ "group_size": 64,
3113
+ "bits": 4
3114
+ },
3115
+ "model.layers.47.mlp.switch_mlp.down_proj": {
3116
+ "group_size": 32,
3117
+ "bits": 6
3118
+ },
3119
+ "lm_head": {
3120
+ "group_size": 32,
3121
+ "bits": 8
3122
+ }
3123
+ },
3124
+ "rms_norm_eps": 1e-06,
3125
+ "rope_scaling": null,
3126
+ "rope_theta": 10000000,
3127
+ "router_aux_loss_coef": 0.0,
3128
+ "shared_expert_intermediate_size": 0,
3129
+ "sliding_window": null,
3130
+ "tie_word_embeddings": false,
3131
+ "torch_dtype": "bfloat16",
3132
+ "transformers_version": "4.54.1",
3133
+ "unsloth_fixed": true,
3134
+ "use_cache": true,
3135
+ "use_qk_norm": true,
3136
+ "use_sliding_window": false,
3137
+ "vocab_size": 151936
3138
+ }
generation_config.json ADDED
@@ -0,0 +1,12 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "pad_token_id": 151643,
3
+ "do_sample": true,
4
+ "eos_token_id": [
5
+ 151645,
6
+ 151643
7
+ ],
8
+ "repetition_penalty": 1.05,
9
+ "temperature": 0.7,
10
+ "top_p": 0.8,
11
+ "top_k": 20
12
+ }
merges.txt ADDED
The diff for this file is too large to render. See raw diff
 
model-00001-of-00004.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:57a0225b8bbe916570a8849c38eaa7fc20b0414f528a8b1f81ad34c967b451ac
3
+ size 5299740222
model-00002-of-00004.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d23e1d0415c0747fe9cb502ad4c687ece7a314701cdb991073fe459cb54528f2
3
+ size 5273751967
model-00003-of-00004.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5b3d34dc0d89773bdb87ca7b249dceb89345816ee94f7772d4c6e514bcc0c340
3
+ size 5347441648
model-00004-of-00004.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8b26dbea76f5f12408512cf823d7f6c2f05684a879e37ae78c4e32d33d77bdfa
3
+ size 3235262240
model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
qwen3coder_tool_parser.py ADDED
@@ -0,0 +1,675 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # SPDX-License-Identifier: Apache-2.0
2
+
3
+ import json
4
+ import re
5
+ import uuid
6
+ from collections.abc import Sequence
7
+ from typing import Union, Optional, Any, List, Dict
8
+ from enum import Enum
9
+
10
+ from vllm.entrypoints.openai.protocol import (
11
+ ChatCompletionRequest,
12
+ ChatCompletionToolsParam,
13
+ DeltaMessage,
14
+ DeltaToolCall,
15
+ DeltaFunctionCall,
16
+ ExtractedToolCallInformation,
17
+ FunctionCall,
18
+ ToolCall,
19
+ )
20
+ from vllm.entrypoints.openai.tool_parsers.abstract_tool_parser import (
21
+ ToolParser,
22
+ ToolParserManager,
23
+ )
24
+ from vllm.logger import init_logger
25
+ from vllm.transformers_utils.tokenizer import AnyTokenizer
26
+
27
+ logger = init_logger(__name__)
28
+
29
+
30
+ @ToolParserManager.register_module("qwen3_xml")
31
+ class Qwen3XMLToolParser(ToolParser):
32
+ def __init__(self, tokenizer: AnyTokenizer):
33
+ super().__init__(tokenizer)
34
+
35
+ self.current_tool_name_sent: bool = False
36
+ self.prev_tool_call_arr: list[dict] = []
37
+ self.current_tool_id: int = -1
38
+ self.streamed_args_for_tool: list[str] = []
39
+
40
+ # Sentinel tokens for streaming mode
41
+ self.tool_call_start_token: str = "<tool_call>"
42
+ self.tool_call_end_token: str = "</tool_call>"
43
+ self.tool_call_prefix: str = "<function="
44
+ self.function_end_token: str = "</function>"
45
+ self.parameter_prefix: str = "<parameter="
46
+ self.parameter_end_token: str = "</parameter>"
47
+ self.is_tool_call_started: bool = False
48
+ self.failed_count: int = 0
49
+
50
+ # Enhanced streaming state - reset for each new message
51
+ self._reset_streaming_state()
52
+
53
+ # Regex patterns
54
+ self.tool_call_complete_regex = re.compile(
55
+ r"<tool_call>(.*?)</tool_call>", re.DOTALL
56
+ )
57
+ self.tool_call_regex = re.compile(
58
+ r"<tool_call>(.*?)</tool_call>|<tool_call>(.*?)$", re.DOTALL
59
+ )
60
+ self.tool_call_function_regex = re.compile(
61
+ r"<function=(.*?)</function>|<function=(.*)$", re.DOTALL
62
+ )
63
+ self.tool_call_parameter_regex = re.compile(
64
+ r"<parameter=(.*?)</parameter>|<parameter=(.*?)$", re.DOTALL
65
+ )
66
+
67
+ if not self.model_tokenizer:
68
+ raise ValueError(
69
+ "The model tokenizer must be passed to the ToolParser "
70
+ "constructor during construction."
71
+ )
72
+
73
+ self.tool_call_start_token_id = self.vocab.get(self.tool_call_start_token)
74
+ self.tool_call_end_token_id = self.vocab.get(self.tool_call_end_token)
75
+
76
+ if self.tool_call_start_token_id is None or self.tool_call_end_token_id is None:
77
+ raise RuntimeError(
78
+ "Qwen3 XML Tool parser could not locate tool call start/end "
79
+ "tokens in the tokenizer!"
80
+ )
81
+
82
+ logger.info(f"vLLM Successfully import tool parser {self.__class__.__name__} !")
83
+
84
+ def _generate_tool_call_id(self) -> str:
85
+ """Generate a unique tool call ID."""
86
+ return f"call_{uuid.uuid4().hex[:24]}"
87
+
88
+ def _reset_streaming_state(self):
89
+ """Reset all streaming state."""
90
+ self.current_tool_index = 0
91
+ self.is_tool_call_started = False
92
+ self.header_sent = False
93
+ self.current_tool_id = None
94
+ self.current_function_name = None
95
+ self.current_param_name = None
96
+ self.current_param_value = ""
97
+ self.param_count = 0
98
+ self.in_param = False
99
+ self.in_function = False
100
+ self.accumulated_text = ""
101
+ self.json_started = False
102
+ self.json_closed = False
103
+
104
+ def _parse_xml_function_call(
105
+ self, function_call_str: str, tools: Optional[list[ChatCompletionToolsParam]]
106
+ ) -> Optional[ToolCall]:
107
+ def get_arguments_config(func_name: str) -> dict:
108
+ if tools is None:
109
+ return {}
110
+ for config in tools:
111
+ if not hasattr(config, "type") or not (
112
+ hasattr(config, "function") and hasattr(config.function, "name")
113
+ ):
114
+ continue
115
+ if config.type == "function" and config.function.name == func_name:
116
+ if not hasattr(config.function, "parameters"):
117
+ return {}
118
+ params = config.function.parameters
119
+ if isinstance(params, dict) and "properties" in params:
120
+ return params["properties"]
121
+ elif isinstance(params, dict):
122
+ return params
123
+ else:
124
+ return {}
125
+ logger.warning(f"Tool '{func_name}' is not defined in the tools list.")
126
+ return {}
127
+
128
+ def convert_param_value(
129
+ param_value: str, param_name: str, param_config: dict, func_name: str
130
+ ) -> Any:
131
+ # Handle null value for any type
132
+ if param_value.lower() == "null":
133
+ return None
134
+
135
+ if param_name not in param_config:
136
+ if param_config != {}:
137
+ logger.warning(
138
+ f"Parsed parameter '{param_name}' is not defined in the tool "
139
+ f"parameters for tool '{func_name}', directly returning the string value."
140
+ )
141
+ return param_value
142
+
143
+ if (
144
+ isinstance(param_config[param_name], dict)
145
+ and "type" in param_config[param_name]
146
+ ):
147
+ param_type = str(param_config[param_name]["type"]).strip().lower()
148
+ else:
149
+ param_type = "string"
150
+ if param_type in ["string", "str", "text", "varchar", "char", "enum"]:
151
+ return param_value
152
+ elif (
153
+ param_type.startswith("int")
154
+ or param_type.startswith("uint")
155
+ or param_type.startswith("long")
156
+ or param_type.startswith("short")
157
+ or param_type.startswith("unsigned")
158
+ ):
159
+ try:
160
+ param_value = int(param_value)
161
+ except:
162
+ logger.warning(
163
+ f"Parsed value '{param_value}' of parameter '{param_name}' is not an integer in tool "
164
+ f"'{func_name}', degenerating to string."
165
+ )
166
+ return param_value
167
+ elif param_type.startswith("num") or param_type.startswith("float"):
168
+ try:
169
+ float_param_value = float(param_value)
170
+ param_value = float_param_value if float_param_value - int(float_param_value) != 0 else int(float_param_value)
171
+ except:
172
+ logger.warning(
173
+ f"Parsed value '{param_value}' of parameter '{param_name}' is not a float in tool "
174
+ f"'{func_name}', degenerating to string."
175
+ )
176
+ return param_value
177
+ elif param_type in ["boolean", "bool", "binary"]:
178
+ param_value = param_value.lower()
179
+ if param_value not in ["true", "false"]:
180
+ logger.warning(
181
+ f"Parsed value '{param_value}' of parameter '{param_name}' is not a boolean (`true` of `false`) in tool '{func_name}', degenerating to false."
182
+ )
183
+ return param_value == "true"
184
+ else:
185
+ if param_type == "object" or param_type.startswith("dict"):
186
+ try:
187
+ param_value = json.loads(param_value)
188
+ return param_value
189
+ except:
190
+ logger.warning(
191
+ f"Parsed value '{param_value}' of parameter '{param_name}' is not a valid JSON object in tool "
192
+ f"'{func_name}', will try other methods to parse it."
193
+ )
194
+ try:
195
+ param_value = eval(param_value)
196
+ except:
197
+ logger.warning(
198
+ f"Parsed value '{param_value}' of parameter '{param_name}' cannot be converted via Python `eval()` in tool '{func_name}', degenerating to string."
199
+ )
200
+ return param_value
201
+
202
+ # Extract function name
203
+ end_index = function_call_str.index(">")
204
+ function_name = function_call_str[:end_index]
205
+ param_config = get_arguments_config(function_name)
206
+ parameters = function_call_str[end_index + 1 :]
207
+ param_dict = {}
208
+ for match in self.tool_call_parameter_regex.findall(parameters):
209
+ match_text = match[0] if match[0] else match[1]
210
+ idx = match_text.index(">")
211
+ param_name = match_text[:idx]
212
+ param_value = str(match_text[idx + 1 :])
213
+ # Remove prefix and trailing \n
214
+ if param_value.startswith("\n"):
215
+ param_value = param_value[1:]
216
+ if param_value.endswith("\n"):
217
+ param_value = param_value[:-1]
218
+
219
+ param_dict[param_name] = convert_param_value(
220
+ param_value, param_name, param_config, function_name
221
+ )
222
+ return ToolCall(
223
+ type="function",
224
+ function=FunctionCall(
225
+ name=function_name, arguments=json.dumps(param_dict, ensure_ascii=False)
226
+ ),
227
+ )
228
+
229
+ def _get_function_calls(self, model_output: str) -> List[str]:
230
+ # Find all tool calls
231
+ matched_ranges = self.tool_call_regex.findall(model_output)
232
+ raw_tool_calls = [
233
+ match[0] if match[0] else match[1] for match in matched_ranges
234
+ ]
235
+
236
+ # Back-off strategy if no tool_call tags found
237
+ if len(raw_tool_calls) == 0:
238
+ raw_tool_calls = [model_output]
239
+
240
+ raw_function_calls = []
241
+ for tool_call in raw_tool_calls:
242
+ raw_function_calls.extend(self.tool_call_function_regex.findall(tool_call))
243
+
244
+ function_calls = [
245
+ match[0] if match[0] else match[1] for match in raw_function_calls
246
+ ]
247
+ return function_calls
248
+
249
+ def extract_tool_calls(
250
+ self,
251
+ model_output: str,
252
+ request: ChatCompletionRequest,
253
+ ) -> ExtractedToolCallInformation:
254
+ # Quick check to avoid unnecessary processing
255
+ if self.tool_call_prefix not in model_output:
256
+ return ExtractedToolCallInformation(
257
+ tools_called=False, tool_calls=[], content=model_output
258
+ )
259
+
260
+ try:
261
+ function_calls = self._get_function_calls(model_output)
262
+ if len(function_calls) == 0:
263
+ return ExtractedToolCallInformation(
264
+ tools_called=False, tool_calls=[], content=model_output
265
+ )
266
+
267
+ tool_calls = [
268
+ self._parse_xml_function_call(function_call_str, request.tools)
269
+ for function_call_str in function_calls
270
+ ]
271
+
272
+ # Populate prev_tool_call_arr for serving layer to set finish_reason
273
+ self.prev_tool_call_arr.clear() # Clear previous calls
274
+ for tool_call in tool_calls:
275
+ if tool_call:
276
+ self.prev_tool_call_arr.append(
277
+ {
278
+ "name": tool_call.function.name,
279
+ "arguments": tool_call.function.arguments,
280
+ }
281
+ )
282
+
283
+ # Extract content before tool calls
284
+ content_index = model_output.find(self.tool_call_start_token)
285
+ content_index = (
286
+ content_index
287
+ if content_index >= 0
288
+ else model_output.find(self.tool_call_prefix)
289
+ )
290
+ content = model_output[:content_index] # .rstrip()
291
+
292
+ return ExtractedToolCallInformation(
293
+ tools_called=(len(tool_calls) > 0),
294
+ tool_calls=tool_calls,
295
+ content=content if content else None,
296
+ )
297
+
298
+ except Exception:
299
+ logger.exception("Error in extracting tool call from response.")
300
+ return ExtractedToolCallInformation(
301
+ tools_called=False, tool_calls=[], content=model_output
302
+ )
303
+
304
+ def extract_tool_calls_streaming(
305
+ self,
306
+ previous_text: str,
307
+ current_text: str,
308
+ delta_text: str,
309
+ previous_token_ids: Sequence[int],
310
+ current_token_ids: Sequence[int],
311
+ delta_token_ids: Sequence[int],
312
+ request: ChatCompletionRequest,
313
+ ) -> Union[DeltaMessage, None]:
314
+ # If no delta text, return None unless it's an EOS token after tool calls
315
+ if not delta_text:
316
+ # Check if this is an EOS token after all tool calls are complete
317
+ # We check for tool calls in the text even if is_tool_call_started is False
318
+ # because it might have been reset after processing all tools
319
+ if delta_token_ids and self.tool_call_end_token_id not in delta_token_ids:
320
+ # Count complete tool calls
321
+ complete_calls = len(
322
+ self.tool_call_complete_regex.findall(current_text)
323
+ )
324
+
325
+ # If we have completed tool calls and populated prev_tool_call_arr
326
+ if complete_calls > 0 and len(self.prev_tool_call_arr) > 0:
327
+ # Check if all tool calls are closed
328
+ open_calls = current_text.count(
329
+ self.tool_call_start_token
330
+ ) - current_text.count(self.tool_call_end_token)
331
+ if open_calls == 0:
332
+ # Return empty delta message to allow finish_reason processing
333
+ return DeltaMessage(content="")
334
+ elif not self.is_tool_call_started and current_text:
335
+ # This is a regular content response that's now complete
336
+ return DeltaMessage(content="")
337
+ return None
338
+
339
+ # Check if this is the first call (reset state if needed)
340
+ if not previous_text:
341
+ self._reset_streaming_state()
342
+
343
+ # Update accumulated text
344
+ self.accumulated_text = current_text
345
+
346
+ # Check if we need to advance to next tool
347
+ if self.json_closed and not self.in_function:
348
+ # Check if this tool call has ended
349
+ tool_ends = current_text.count(self.tool_call_end_token)
350
+ if tool_ends > self.current_tool_index:
351
+ # This tool has ended, advance to next
352
+ self.current_tool_index += 1
353
+ self.header_sent = False
354
+ self.param_count = 0
355
+ self.json_started = False
356
+ self.json_closed = False
357
+
358
+ # Check if there are more tool calls
359
+ tool_starts = current_text.count(self.tool_call_start_token)
360
+ if self.current_tool_index >= tool_starts:
361
+ # No more tool calls
362
+ self.is_tool_call_started = False
363
+ # Continue processing next tool
364
+ return None
365
+
366
+ # Handle normal content before tool calls
367
+ if not self.is_tool_call_started:
368
+ # Check if tool call is starting
369
+ if (
370
+ self.tool_call_start_token_id in delta_token_ids
371
+ or self.tool_call_start_token in delta_text
372
+ ):
373
+ self.is_tool_call_started = True
374
+ # Return any content before the tool call
375
+ if self.tool_call_start_token in delta_text:
376
+ content_before = delta_text[
377
+ : delta_text.index(self.tool_call_start_token)
378
+ ]
379
+ if content_before:
380
+ return DeltaMessage(content=content_before)
381
+ return None
382
+ else:
383
+ # Check if we're between tool calls - skip whitespace
384
+ if current_text.rstrip().endswith(self.tool_call_end_token):
385
+ # We just ended a tool call, skip whitespace
386
+ if delta_text.strip() == "":
387
+ return None
388
+ # Normal content, no tool call
389
+ return DeltaMessage(content=delta_text)
390
+
391
+ # Check if we're between tool calls (waiting for next one)
392
+ # Count tool calls we've seen vs processed
393
+ tool_starts_count = current_text.count(self.tool_call_start_token)
394
+ if self.current_tool_index >= tool_starts_count:
395
+ # We're past all tool calls, shouldn't be here
396
+ return None
397
+
398
+ # We're in a tool call, find the current tool call portion
399
+ # Need to find the correct tool call based on current_tool_index
400
+ tool_starts = []
401
+ idx = 0
402
+ while True:
403
+ idx = current_text.find(self.tool_call_start_token, idx)
404
+ if idx == -1:
405
+ break
406
+ tool_starts.append(idx)
407
+ idx += len(self.tool_call_start_token)
408
+
409
+ if self.current_tool_index >= len(tool_starts):
410
+ # No more tool calls to process yet
411
+ return None
412
+
413
+ tool_start_idx = tool_starts[self.current_tool_index]
414
+ # Find where this tool call ends (or current position if not ended yet)
415
+ tool_end_idx = current_text.find(self.tool_call_end_token, tool_start_idx)
416
+ if tool_end_idx == -1:
417
+ tool_text = current_text[tool_start_idx:]
418
+ else:
419
+ tool_text = current_text[
420
+ tool_start_idx : tool_end_idx + len(self.tool_call_end_token)
421
+ ]
422
+
423
+ # Looking for function header
424
+ if not self.header_sent:
425
+ if self.tool_call_prefix in tool_text:
426
+ func_start = tool_text.find(self.tool_call_prefix) + len(
427
+ self.tool_call_prefix
428
+ )
429
+ func_end = tool_text.find(">", func_start)
430
+
431
+ if func_end != -1:
432
+ # Found complete function name
433
+ self.current_function_name = tool_text[func_start:func_end]
434
+ self.current_tool_id = self._generate_tool_call_id()
435
+ self.header_sent = True
436
+ self.in_function = True
437
+
438
+ # IMPORTANT: Add to prev_tool_call_arr immediately when we detect a tool call
439
+ # This ensures finish_reason="tool_calls" even if parsing isn't complete
440
+ already_added = any(
441
+ tool.get("name") == self.current_function_name
442
+ for tool in self.prev_tool_call_arr
443
+ )
444
+ if not already_added:
445
+ self.prev_tool_call_arr.append(
446
+ {
447
+ "name": self.current_function_name,
448
+ "arguments": "{}", # Placeholder, will be updated later
449
+ }
450
+ )
451
+
452
+ # Send header with function info
453
+ return DeltaMessage(
454
+ tool_calls=[
455
+ DeltaToolCall(
456
+ index=self.current_tool_index,
457
+ id=self.current_tool_id,
458
+ function=DeltaFunctionCall(
459
+ name=self.current_function_name, arguments=""
460
+ ),
461
+ type="function",
462
+ )
463
+ ]
464
+ )
465
+ return None
466
+
467
+ # We've sent header, now handle function body
468
+ if self.in_function:
469
+ # Send opening brace if not sent yet
470
+ if not self.json_started and not self.parameter_prefix in delta_text:
471
+ self.json_started = True
472
+ return DeltaMessage(
473
+ tool_calls=[
474
+ DeltaToolCall(
475
+ index=self.current_tool_index,
476
+ function=DeltaFunctionCall(arguments="{"),
477
+ )
478
+ ]
479
+ )
480
+
481
+ # Make sure json_started is set if we're processing parameters
482
+ if not self.json_started:
483
+ self.json_started = True
484
+
485
+ # Check for function end in accumulated text
486
+ if not self.json_closed and self.function_end_token in tool_text:
487
+ # Close JSON
488
+ self.json_closed = True
489
+
490
+ # Extract the complete tool call to update prev_tool_call_arr with final arguments
491
+ # Find the function content
492
+ func_start = tool_text.find(self.tool_call_prefix) + len(
493
+ self.tool_call_prefix
494
+ )
495
+ func_content_end = tool_text.find(self.function_end_token, func_start)
496
+ if func_content_end != -1:
497
+ func_content = tool_text[func_start:func_content_end]
498
+ # Parse to get the complete arguments
499
+ try:
500
+ parsed_tool = self._parse_xml_function_call(
501
+ func_content, request.tools if request else None
502
+ )
503
+ if parsed_tool:
504
+ # Update existing entry in prev_tool_call_arr with complete arguments
505
+ for i, tool in enumerate(self.prev_tool_call_arr):
506
+ if tool.get("name") == parsed_tool.function.name:
507
+ self.prev_tool_call_arr[i]["arguments"] = (
508
+ parsed_tool.function.arguments
509
+ )
510
+ break
511
+ except Exception:
512
+ pass # Ignore parsing errors during streaming
513
+
514
+ result = DeltaMessage(
515
+ tool_calls=[
516
+ DeltaToolCall(
517
+ index=self.current_tool_index,
518
+ function=DeltaFunctionCall(arguments="}"),
519
+ )
520
+ ]
521
+ )
522
+
523
+ # Reset state for next tool
524
+ self.in_function = False
525
+ self.json_closed = True
526
+
527
+ return result
528
+
529
+ # Look for parameters
530
+ # Count how many complete parameters we have processed
531
+ complete_params = tool_text.count(self.parameter_end_token)
532
+
533
+ # Check if we should start a new parameter
534
+ if not self.in_param and self.param_count < complete_params:
535
+ # Find the unprocessed parameter
536
+ # Count parameter starts
537
+ param_starts = []
538
+ idx = 0
539
+ while True:
540
+ idx = tool_text.find(self.parameter_prefix, idx)
541
+ if idx == -1:
542
+ break
543
+ param_starts.append(idx)
544
+ idx += len(self.parameter_prefix)
545
+
546
+ if len(param_starts) > self.param_count:
547
+ # Process the next parameter
548
+ param_idx = param_starts[self.param_count]
549
+ param_start = param_idx + len(self.parameter_prefix)
550
+ remaining = tool_text[param_start:]
551
+
552
+ if ">" in remaining:
553
+ # We have the complete parameter name
554
+ name_end = remaining.find(">")
555
+ self.current_param_name = remaining[:name_end]
556
+
557
+ # Find the parameter value
558
+ value_start = param_start + name_end + 1
559
+ value_text = tool_text[value_start:]
560
+ if value_text.startswith("\n"):
561
+ value_text = value_text[1:]
562
+
563
+ # Find where this parameter ends
564
+ param_end_idx = value_text.find(self.parameter_end_token)
565
+ if param_end_idx != -1:
566
+ # Complete parameter found
567
+ param_value = value_text[:param_end_idx]
568
+ if param_value.endswith("\n"):
569
+ param_value = param_value[:-1]
570
+
571
+ # Build complete JSON fragment for this parameter
572
+ if self.param_count == 0:
573
+ json_fragment = (
574
+ '"'
575
+ + self.current_param_name
576
+ + '": "'
577
+ + json.dumps(param_value)[1:-1]
578
+ + '"'
579
+ )
580
+ else:
581
+ json_fragment = (
582
+ ', "'
583
+ + self.current_param_name
584
+ + '": "'
585
+ + json.dumps(param_value)[1:-1]
586
+ + '"'
587
+ )
588
+
589
+ self.param_count += 1
590
+
591
+ return DeltaMessage(
592
+ tool_calls=[
593
+ DeltaToolCall(
594
+ index=self.current_tool_index,
595
+ function=DeltaFunctionCall(
596
+ arguments=json_fragment
597
+ ),
598
+ )
599
+ ]
600
+ )
601
+
602
+ # Continue parameter value
603
+ if self.in_param:
604
+ if self.parameter_end_token in delta_text:
605
+ # End of parameter
606
+ end_idx = delta_text.find(self.parameter_end_token)
607
+ value_chunk = delta_text[:end_idx]
608
+
609
+ # Skip past > if at start
610
+ if not self.current_param_value and ">" in value_chunk:
611
+ gt_idx = value_chunk.find(">")
612
+ value_chunk = value_chunk[gt_idx + 1 :]
613
+
614
+ if not self.current_param_value and value_chunk.startswith("\n"):
615
+ value_chunk = value_chunk[1:]
616
+
617
+ # Calculate incremental JSON
618
+ full_value = self.current_param_value + value_chunk
619
+ prev_escaped = (
620
+ json.dumps(self.current_param_value)[1:-1]
621
+ if self.current_param_value
622
+ else ""
623
+ )
624
+ full_escaped = json.dumps(full_value)[1:-1]
625
+ delta_escaped = full_escaped[len(prev_escaped) :]
626
+
627
+ self.in_param = False
628
+ self.current_param_value = ""
629
+
630
+ return DeltaMessage(
631
+ tool_calls=[
632
+ DeltaToolCall(
633
+ index=self.current_tool_index,
634
+ function=DeltaFunctionCall(
635
+ arguments=delta_escaped + '"'
636
+ ),
637
+ )
638
+ ]
639
+ )
640
+ else:
641
+ # Continue accumulating value
642
+ value_chunk = delta_text
643
+
644
+ # Handle first chunk after param name
645
+ if not self.current_param_value and ">" in value_chunk:
646
+ gt_idx = value_chunk.find(">")
647
+ value_chunk = value_chunk[gt_idx + 1 :]
648
+
649
+ if not self.current_param_value and value_chunk.startswith("\n"):
650
+ value_chunk = value_chunk[1:]
651
+
652
+ if value_chunk:
653
+ # Stream the escaped delta
654
+ prev_escaped = (
655
+ json.dumps(self.current_param_value)[1:-1]
656
+ if self.current_param_value
657
+ else ""
658
+ )
659
+ self.current_param_value += value_chunk
660
+ full_escaped = json.dumps(self.current_param_value)[1:-1]
661
+ delta_escaped = full_escaped[len(prev_escaped) :]
662
+
663
+ if delta_escaped:
664
+ return DeltaMessage(
665
+ tool_calls=[
666
+ DeltaToolCall(
667
+ index=self.current_tool_index,
668
+ function=DeltaFunctionCall(
669
+ arguments=delta_escaped
670
+ ),
671
+ )
672
+ ]
673
+ )
674
+
675
+ return None
special_tokens_map.json ADDED
@@ -0,0 +1,31 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "additional_special_tokens": [
3
+ "<|im_start|>",
4
+ "<|im_end|>",
5
+ "<|object_ref_start|>",
6
+ "<|object_ref_end|>",
7
+ "<|box_start|>",
8
+ "<|box_end|>",
9
+ "<|quad_start|>",
10
+ "<|quad_end|>",
11
+ "<|vision_start|>",
12
+ "<|vision_end|>",
13
+ "<|vision_pad|>",
14
+ "<|image_pad|>",
15
+ "<|video_pad|>"
16
+ ],
17
+ "eos_token": {
18
+ "content": "<|im_end|>",
19
+ "lstrip": false,
20
+ "normalized": false,
21
+ "rstrip": false,
22
+ "single_word": false
23
+ },
24
+ "pad_token": {
25
+ "content": "<|vision_pad|>",
26
+ "lstrip": false,
27
+ "normalized": false,
28
+ "rstrip": false,
29
+ "single_word": false
30
+ }
31
+ }
tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:aeb13307a71acd8fe81861d94ad54ab689df773318809eed3cbe794b4492dae4
3
+ size 11422654
tokenizer_config.json ADDED
@@ -0,0 +1,240 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_bos_token": false,
3
+ "add_prefix_space": false,
4
+ "added_tokens_decoder": {
5
+ "151643": {
6
+ "content": "<|endoftext|>",
7
+ "lstrip": false,
8
+ "normalized": false,
9
+ "rstrip": false,
10
+ "single_word": false,
11
+ "special": true
12
+ },
13
+ "151644": {
14
+ "content": "<|im_start|>",
15
+ "lstrip": false,
16
+ "normalized": false,
17
+ "rstrip": false,
18
+ "single_word": false,
19
+ "special": true
20
+ },
21
+ "151645": {
22
+ "content": "<|im_end|>",
23
+ "lstrip": false,
24
+ "normalized": false,
25
+ "rstrip": false,
26
+ "single_word": false,
27
+ "special": true
28
+ },
29
+ "151646": {
30
+ "content": "<|object_ref_start|>",
31
+ "lstrip": false,
32
+ "normalized": false,
33
+ "rstrip": false,
34
+ "single_word": false,
35
+ "special": true
36
+ },
37
+ "151647": {
38
+ "content": "<|object_ref_end|>",
39
+ "lstrip": false,
40
+ "normalized": false,
41
+ "rstrip": false,
42
+ "single_word": false,
43
+ "special": true
44
+ },
45
+ "151648": {
46
+ "content": "<|box_start|>",
47
+ "lstrip": false,
48
+ "normalized": false,
49
+ "rstrip": false,
50
+ "single_word": false,
51
+ "special": true
52
+ },
53
+ "151649": {
54
+ "content": "<|box_end|>",
55
+ "lstrip": false,
56
+ "normalized": false,
57
+ "rstrip": false,
58
+ "single_word": false,
59
+ "special": true
60
+ },
61
+ "151650": {
62
+ "content": "<|quad_start|>",
63
+ "lstrip": false,
64
+ "normalized": false,
65
+ "rstrip": false,
66
+ "single_word": false,
67
+ "special": true
68
+ },
69
+ "151651": {
70
+ "content": "<|quad_end|>",
71
+ "lstrip": false,
72
+ "normalized": false,
73
+ "rstrip": false,
74
+ "single_word": false,
75
+ "special": true
76
+ },
77
+ "151652": {
78
+ "content": "<|vision_start|>",
79
+ "lstrip": false,
80
+ "normalized": false,
81
+ "rstrip": false,
82
+ "single_word": false,
83
+ "special": true
84
+ },
85
+ "151653": {
86
+ "content": "<|vision_end|>",
87
+ "lstrip": false,
88
+ "normalized": false,
89
+ "rstrip": false,
90
+ "single_word": false,
91
+ "special": true
92
+ },
93
+ "151654": {
94
+ "content": "<|vision_pad|>",
95
+ "lstrip": false,
96
+ "normalized": false,
97
+ "rstrip": false,
98
+ "single_word": false,
99
+ "special": true
100
+ },
101
+ "151655": {
102
+ "content": "<|image_pad|>",
103
+ "lstrip": false,
104
+ "normalized": false,
105
+ "rstrip": false,
106
+ "single_word": false,
107
+ "special": true
108
+ },
109
+ "151656": {
110
+ "content": "<|video_pad|>",
111
+ "lstrip": false,
112
+ "normalized": false,
113
+ "rstrip": false,
114
+ "single_word": false,
115
+ "special": true
116
+ },
117
+ "151657": {
118
+ "content": "<tool_call>",
119
+ "lstrip": false,
120
+ "normalized": false,
121
+ "rstrip": false,
122
+ "single_word": false,
123
+ "special": false
124
+ },
125
+ "151658": {
126
+ "content": "</tool_call>",
127
+ "lstrip": false,
128
+ "normalized": false,
129
+ "rstrip": false,
130
+ "single_word": false,
131
+ "special": false
132
+ },
133
+ "151659": {
134
+ "content": "<|fim_prefix|>",
135
+ "lstrip": false,
136
+ "normalized": false,
137
+ "rstrip": false,
138
+ "single_word": false,
139
+ "special": false
140
+ },
141
+ "151660": {
142
+ "content": "<|fim_middle|>",
143
+ "lstrip": false,
144
+ "normalized": false,
145
+ "rstrip": false,
146
+ "single_word": false,
147
+ "special": false
148
+ },
149
+ "151661": {
150
+ "content": "<|fim_suffix|>",
151
+ "lstrip": false,
152
+ "normalized": false,
153
+ "rstrip": false,
154
+ "single_word": false,
155
+ "special": false
156
+ },
157
+ "151662": {
158
+ "content": "<|fim_pad|>",
159
+ "lstrip": false,
160
+ "normalized": false,
161
+ "rstrip": false,
162
+ "single_word": false,
163
+ "special": false
164
+ },
165
+ "151663": {
166
+ "content": "<|repo_name|>",
167
+ "lstrip": false,
168
+ "normalized": false,
169
+ "rstrip": false,
170
+ "single_word": false,
171
+ "special": false
172
+ },
173
+ "151664": {
174
+ "content": "<|file_sep|>",
175
+ "lstrip": false,
176
+ "normalized": false,
177
+ "rstrip": false,
178
+ "single_word": false,
179
+ "special": false
180
+ },
181
+ "151665": {
182
+ "content": "<tool_response>",
183
+ "lstrip": false,
184
+ "normalized": false,
185
+ "rstrip": false,
186
+ "single_word": false,
187
+ "special": false
188
+ },
189
+ "151666": {
190
+ "content": "</tool_response>",
191
+ "lstrip": false,
192
+ "normalized": false,
193
+ "rstrip": false,
194
+ "single_word": false,
195
+ "special": false
196
+ },
197
+ "151667": {
198
+ "content": "<think>",
199
+ "lstrip": false,
200
+ "normalized": false,
201
+ "rstrip": false,
202
+ "single_word": false,
203
+ "special": false
204
+ },
205
+ "151668": {
206
+ "content": "</think>",
207
+ "lstrip": false,
208
+ "normalized": false,
209
+ "rstrip": false,
210
+ "single_word": false,
211
+ "special": false
212
+ }
213
+ },
214
+ "additional_special_tokens": [
215
+ "<|im_start|>",
216
+ "<|im_end|>",
217
+ "<|object_ref_start|>",
218
+ "<|object_ref_end|>",
219
+ "<|box_start|>",
220
+ "<|box_end|>",
221
+ "<|quad_start|>",
222
+ "<|quad_end|>",
223
+ "<|vision_start|>",
224
+ "<|vision_end|>",
225
+ "<|vision_pad|>",
226
+ "<|image_pad|>",
227
+ "<|video_pad|>"
228
+ ],
229
+ "bos_token": null,
230
+ "clean_up_tokenization_spaces": false,
231
+ "eos_token": "<|im_end|>",
232
+ "errors": "replace",
233
+ "extra_special_tokens": {},
234
+ "model_max_length": 262144,
235
+ "pad_token": "<|vision_pad|>",
236
+ "padding_side": "left",
237
+ "split_special_tokens": false,
238
+ "tokenizer_class": "Qwen2Tokenizer",
239
+ "unk_token": null
240
+ }
vocab.json ADDED
The diff for this file is too large to render. See raw diff