Hugging Face's logo

nightmedia
/
SR-Scientist-30B-qx86-hi-mlx

Text Generation
MLX
Safetensors
qwen3_moe
conversational
8-bit precision
Model card Files
xet
Community
nightmedia commited on
Commit
08f1016
·
verified ·
1 Parent(s): ecfc725

Add files using upload-large-folder tool

Browse files
Files changed (18) hide show
  1. .gitattributes +1 -0
  2. README.md +38 -0
  3. added_tokens.json +28 -0
  4. chat_template.jinja +131 -0
  5. config.json +3139 -0
  6. generation_config.json +13 -0
  7. merges.txt +0 -0
  8. model-00001-of-00006.safetensors +3 -0
  9. model-00002-of-00006.safetensors +3 -0
  10. model-00003-of-00006.safetensors +3 -0
  11. model-00004-of-00006.safetensors +3 -0
  12. model-00005-of-00006.safetensors +3 -0
  13. model-00006-of-00006.safetensors +3 -0
  14. model.safetensors.index.json +0 -0
  15. special_tokens_map.json +31 -0
  16. tokenizer.json +3 -0
  17. tokenizer_config.json +239 -0
  18. 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
+ license: apache-2.0
3
+ datasets:
4
+ - GAIR/SR-Scientist
5
+ base_model: GAIR/SR-Scientist-30B
6
+ tags:
7
+ - mlx
8
+ pipeline_tag: text-generation
9
+ library_name: mlx
10
+ ---
11
+
12
+ # SR-Scientist-30B-qx86-hi-mlx
13
+
14
+ This model [SR-Scientist-30B-qx86-hi-mlx](https://huggingface.co/SR-Scientist-30B-qx86-hi-mlx) was
15
+ converted to MLX format from [GAIR/SR-Scientist-30B](https://huggingface.co/GAIR/SR-Scientist-30B)
16
+ using mlx-lm version **0.28.2**.
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("SR-Scientist-30B-qx86-hi-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 +1,131 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {% macro render_item_list(item_list, tag_name='required') %}
2
+ {%- if item_list is defined and item_list is iterable and item_list | length > 0 %}
3
+ {%- if tag_name %}{{- '\n<' ~ tag_name ~ '>' -}}{% endif %}
4
+ {{- '[' }}
5
+ {%- for item in item_list -%}
6
+ {%- if loop.index > 1 %}{{- ", "}}{% endif -%}
7
+ {%- if item is string -%}
8
+ {{ "`" ~ item ~ "`" }}
9
+ {%- else -%}
10
+ {{ item }}
11
+ {%- endif -%}
12
+ {%- endfor -%}
13
+ {{- ']' }}
14
+ {%- if tag_name %}{{- '</' ~ tag_name ~ '>' -}}{% endif %}
15
+ {%- endif %}
16
+ {% endmacro %}
17
+
18
+ {%- if messages[0]["role"] == "system" %}
19
+ {%- set system_message = messages[0]["content"] %}
20
+ {%- set loop_messages = messages[1:] %}
21
+ {%- else %}
22
+ {%- set loop_messages = messages %}
23
+ {%- endif %}
24
+
25
+ {%- if not tools is defined %}
26
+ {%- set tools = [] %}
27
+ {%- endif %}
28
+
29
+ {%- if system_message is defined %}
30
+ {{- "<|im_start|>system\n" + system_message }}
31
+ {%- else %}
32
+ {%- if tools is iterable and tools | length > 0 %}
33
+ {{- "<|im_start|>system\nYou are Qwen, a helpful AI assistant that can interact with a computer to solve tasks." }}
34
+ {%- endif %}
35
+ {%- endif %}
36
+ {%- if tools is iterable and tools | length > 0 %}
37
+ {{- "\n\nYou have access to the following functions:\n\n" }}
38
+ {{- "<tools>" }}
39
+ {%- for tool in tools %}
40
+ {%- if tool.function is defined %}
41
+ {%- set tool = tool.function %}
42
+ {%- endif %}
43
+ {{- "\n<function>\n<name>" ~ tool.name ~ "</name>" }}
44
+ {{- '\n<description>' ~ (tool.description | trim) ~ '</description>' }}
45
+ {{- '\n<parameters>' }}
46
+ {%- for param_name, param_fields in tool.parameters.properties|items %}
47
+ {{- '\n<parameter>' }}
48
+ {{- '\n<name>' ~ param_name ~ '</name>' }}
49
+ {%- if param_fields.type is defined %}
50
+ {{- '\n<type>' ~ (param_fields.type | string) ~ '</type>' }}
51
+ {%- endif %}
52
+ {%- if param_fields.description is defined %}
53
+ {{- '\n<description>' ~ (param_fields.description | trim) ~ '</description>' }}
54
+ {%- endif %}
55
+ {{- render_item_list(param_fields.enum, 'enum') }}
56
+ {%- set handled_keys = ['type', 'description', 'enum', 'required'] %}
57
+ {%- for json_key in param_fields.keys() | reject("in", handled_keys) %}
58
+ {%- set normed_json_key = json_key | replace("-", "_") | replace(" ", "_") | replace("$", "") %}
59
+ {%- if param_fields[json_key] is mapping %}
60
+ {{- '\n<' ~ normed_json_key ~ '>' ~ (param_fields[json_key] | tojson | safe) ~ '</' ~ normed_json_key ~ '>' }}
61
+ {%- else %}
62
+ {{-'\n<' ~ normed_json_key ~ '>' ~ (param_fields[json_key] | string) ~ '</' ~ normed_json_key ~ '>' }}
63
+ {%- endif %}
64
+ {%- endfor %}
65
+ {{- render_item_list(param_fields.required, 'required') }}
66
+ {{- '\n</parameter>' }}
67
+ {%- endfor %}
68
+ {{- render_item_list(tool.parameters.required, 'required') }}
69
+ {{- '\n</parameters>' }}
70
+ {%- if tool.return is defined %}
71
+ {%- if tool.return is mapping %}
72
+ {{- '\n<return>' ~ (tool.return | tojson | safe) ~ '</return>' }}
73
+ {%- else %}
74
+ {{- '\n<return>' ~ (tool.return | string) ~ '</return>' }}
75
+ {%- endif %}
76
+ {%- endif %}
77
+ {{- '\n</function>' }}
78
+ {%- endfor %}
79
+ {{- "\n</tools>" }}
80
+ {{- '\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>' }}
81
+ {%- endif %}
82
+ {%- if system_message is defined %}
83
+ {{- '<|im_end|>\n' }}
84
+ {%- else %}
85
+ {%- if tools is iterable and tools | length > 0 %}
86
+ {{- '<|im_end|>\n' }}
87
+ {%- endif %}
88
+ {%- endif %}
89
+ {%- for message in loop_messages %}
90
+ {%- if message.role == "assistant" and message.tool_calls is defined and message.tool_calls is iterable and message.tool_calls | length > 0 %}
91
+ {{- '<|im_start|>' + message.role }}
92
+ {%- if message.content is defined and message.content is string and message.content | trim | length > 0 %}
93
+ {{- '\n' + message.content | trim + '\n' }}
94
+ {%- endif %}
95
+ {%- for tool_call in message.tool_calls %}
96
+ {%- if tool_call.function is defined %}
97
+ {%- set tool_call = tool_call.function %}
98
+ {%- endif %}
99
+ {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
100
+ {%- if tool_call.arguments is defined %}
101
+ {%- for args_name, args_value in tool_call.arguments|items %}
102
+ {{- '<parameter=' + args_name + '>\n' }}
103
+ {%- set args_value = args_value if args_value is string else args_value | string %}
104
+ {{- args_value }}
105
+ {{- '\n</parameter>\n' }}
106
+ {%- endfor %}
107
+ {%- endif %}
108
+ {{- '</function>\n</tool_call>' }}
109
+ {%- endfor %}
110
+ {{- '<|im_end|>\n' }}
111
+ {%- elif message.role == "user" or message.role == "system" or message.role == "assistant" %}
112
+ {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
113
+ {%- elif message.role == "tool" %}
114
+ {%- if loop.previtem and loop.previtem.role != "tool" %}
115
+ {{- '<|im_start|>user\n' }}
116
+ {%- endif %}
117
+ {{- '<tool_response>\n' }}
118
+ {{- message.content }}
119
+ {{- '\n</tool_response>\n' }}
120
+ {%- if not loop.last and loop.nextitem.role != "tool" %}
121
+ {{- '<|im_end|>\n' }}
122
+ {%- elif loop.last %}
123
+ {{- '<|im_end|>\n' }}
124
+ {%- endif %}
125
+ {%- else %}
126
+ {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>\n' }}
127
+ {%- endif %}
128
+ {%- endfor %}
129
+ {%- if add_generation_prompt %}
130
+ {{- '<|im_start|>assistant\n' }}
131
+ {%- endif %}
config.json ADDED
@@ -0,0 +1,3139 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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": 151643,
27
+ "qkv_bias": false,
28
+ "quantization": {
29
+ "group_size": 32,
30
+ "bits": 8,
31
+ "mode": "affine",
32
+ "model.embed_tokens": {
33
+ "group_size": 32,
34
+ "bits": 6
35
+ },
36
+ "model.layers.0.self_attn.q_proj": {
37
+ "group_size": 32,
38
+ "bits": 8
39
+ },
40
+ "model.layers.0.self_attn.k_proj": {
41
+ "group_size": 32,
42
+ "bits": 8
43
+ },
44
+ "model.layers.0.self_attn.v_proj": {
45
+ "group_size": 32,
46
+ "bits": 8
47
+ },
48
+ "model.layers.0.self_attn.o_proj": {
49
+ "group_size": 32,
50
+ "bits": 8
51
+ },
52
+ "model.layers.0.mlp.gate": {
53
+ "group_size": 32,
54
+ "bits": 8
55
+ },
56
+ "model.layers.0.mlp.switch_mlp.gate_proj": {
57
+ "group_size": 32,
58
+ "bits": 8
59
+ },
60
+ "model.layers.0.mlp.switch_mlp.up_proj": {
61
+ "group_size": 32,
62
+ "bits": 8
63
+ },
64
+ "model.layers.0.mlp.switch_mlp.down_proj": {
65
+ "group_size": 32,
66
+ "bits": 8
67
+ },
68
+ "model.layers.1.self_attn.q_proj": {
69
+ "group_size": 32,
70
+ "bits": 6
71
+ },
72
+ "model.layers.1.self_attn.k_proj": {
73
+ "group_size": 32,
74
+ "bits": 6
75
+ },
76
+ "model.layers.1.self_attn.v_proj": {
77
+ "group_size": 32,
78
+ "bits": 8
79
+ },
80
+ "model.layers.1.self_attn.o_proj": {
81
+ "group_size": 32,
82
+ "bits": 6
83
+ },
84
+ "model.layers.1.mlp.gate": {
85
+ "group_size": 32,
86
+ "bits": 6
87
+ },
88
+ "model.layers.1.mlp.switch_mlp.gate_proj": {
89
+ "group_size": 32,
90
+ "bits": 6
91
+ },
92
+ "model.layers.1.mlp.switch_mlp.up_proj": {
93
+ "group_size": 32,
94
+ "bits": 6
95
+ },
96
+ "model.layers.1.mlp.switch_mlp.down_proj": {
97
+ "group_size": 32,
98
+ "bits": 8
99
+ },
100
+ "model.layers.2.self_attn.q_proj": {
101
+ "group_size": 32,
102
+ "bits": 6
103
+ },
104
+ "model.layers.2.self_attn.k_proj": {
105
+ "group_size": 32,
106
+ "bits": 6
107
+ },
108
+ "model.layers.2.self_attn.v_proj": {
109
+ "group_size": 32,
110
+ "bits": 8
111
+ },
112
+ "model.layers.2.self_attn.o_proj": {
113
+ "group_size": 32,
114
+ "bits": 6
115
+ },
116
+ "model.layers.2.mlp.gate": {
117
+ "group_size": 32,
118
+ "bits": 6
119
+ },
120
+ "model.layers.2.mlp.switch_mlp.gate_proj": {
121
+ "group_size": 32,
122
+ "bits": 6
123
+ },
124
+ "model.layers.2.mlp.switch_mlp.up_proj": {
125
+ "group_size": 32,
126
+ "bits": 6
127
+ },
128
+ "model.layers.2.mlp.switch_mlp.down_proj": {
129
+ "group_size": 32,
130
+ "bits": 8
131
+ },
132
+ "model.layers.3.self_attn.q_proj": {
133
+ "group_size": 32,
134
+ "bits": 6
135
+ },
136
+ "model.layers.3.self_attn.k_proj": {
137
+ "group_size": 32,
138
+ "bits": 6
139
+ },
140
+ "model.layers.3.self_attn.v_proj": {
141
+ "group_size": 32,
142
+ "bits": 8
143
+ },
144
+ "model.layers.3.self_attn.o_proj": {
145
+ "group_size": 32,
146
+ "bits": 6
147
+ },
148
+ "model.layers.3.mlp.gate": {
149
+ "group_size": 32,
150
+ "bits": 6
151
+ },
152
+ "model.layers.3.mlp.switch_mlp.gate_proj": {
153
+ "group_size": 32,
154
+ "bits": 6
155
+ },
156
+ "model.layers.3.mlp.switch_mlp.up_proj": {
157
+ "group_size": 32,
158
+ "bits": 6
159
+ },
160
+ "model.layers.3.mlp.switch_mlp.down_proj": {
161
+ "group_size": 32,
162
+ "bits": 8
163
+ },
164
+ "model.layers.4.self_attn.q_proj": {
165
+ "group_size": 32,
166
+ "bits": 6
167
+ },
168
+ "model.layers.4.self_attn.k_proj": {
169
+ "group_size": 32,
170
+ "bits": 6
171
+ },
172
+ "model.layers.4.self_attn.v_proj": {
173
+ "group_size": 32,
174
+ "bits": 8
175
+ },
176
+ "model.layers.4.self_attn.o_proj": {
177
+ "group_size": 32,
178
+ "bits": 6
179
+ },
180
+ "model.layers.4.mlp.gate": {
181
+ "group_size": 32,
182
+ "bits": 6
183
+ },
184
+ "model.layers.4.mlp.switch_mlp.gate_proj": {
185
+ "group_size": 32,
186
+ "bits": 6
187
+ },
188
+ "model.layers.4.mlp.switch_mlp.up_proj": {
189
+ "group_size": 32,
190
+ "bits": 6
191
+ },
192
+ "model.layers.4.mlp.switch_mlp.down_proj": {
193
+ "group_size": 32,
194
+ "bits": 8
195
+ },
196
+ "model.layers.5.self_attn.q_proj": {
197
+ "group_size": 32,
198
+ "bits": 6
199
+ },
200
+ "model.layers.5.self_attn.k_proj": {
201
+ "group_size": 32,
202
+ "bits": 6
203
+ },
204
+ "model.layers.5.self_attn.v_proj": {
205
+ "group_size": 32,
206
+ "bits": 8
207
+ },
208
+ "model.layers.5.self_attn.o_proj": {
209
+ "group_size": 32,
210
+ "bits": 6
211
+ },
212
+ "model.layers.5.mlp.gate": {
213
+ "group_size": 32,
214
+ "bits": 6
215
+ },
216
+ "model.layers.5.mlp.switch_mlp.gate_proj": {
217
+ "group_size": 32,
218
+ "bits": 6
219
+ },
220
+ "model.layers.5.mlp.switch_mlp.up_proj": {
221
+ "group_size": 32,
222
+ "bits": 6
223
+ },
224
+ "model.layers.5.mlp.switch_mlp.down_proj": {
225
+ "group_size": 32,
226
+ "bits": 8
227
+ },
228
+ "model.layers.6.self_attn.q_proj": {
229
+ "group_size": 32,
230
+ "bits": 6
231
+ },
232
+ "model.layers.6.self_attn.k_proj": {
233
+ "group_size": 32,
234
+ "bits": 6
235
+ },
236
+ "model.layers.6.self_attn.v_proj": {
237
+ "group_size": 32,
238
+ "bits": 8
239
+ },
240
+ "model.layers.6.self_attn.o_proj": {
241
+ "group_size": 32,
242
+ "bits": 6
243
+ },
244
+ "model.layers.6.mlp.gate": {
245
+ "group_size": 32,
246
+ "bits": 6
247
+ },
248
+ "model.layers.6.mlp.switch_mlp.gate_proj": {
249
+ "group_size": 32,
250
+ "bits": 6
251
+ },
252
+ "model.layers.6.mlp.switch_mlp.up_proj": {
253
+ "group_size": 32,
254
+ "bits": 6
255
+ },
256
+ "model.layers.6.mlp.switch_mlp.down_proj": {
257
+ "group_size": 32,
258
+ "bits": 6
259
+ },
260
+ "model.layers.7.self_attn.q_proj": {
261
+ "group_size": 32,
262
+ "bits": 6
263
+ },
264
+ "model.layers.7.self_attn.k_proj": {
265
+ "group_size": 32,
266
+ "bits": 6
267
+ },
268
+ "model.layers.7.self_attn.v_proj": {
269
+ "group_size": 32,
270
+ "bits": 8
271
+ },
272
+ "model.layers.7.self_attn.o_proj": {
273
+ "group_size": 32,
274
+ "bits": 6
275
+ },
276
+ "model.layers.7.mlp.gate": {
277
+ "group_size": 32,
278
+ "bits": 6
279
+ },
280
+ "model.layers.7.mlp.switch_mlp.gate_proj": {
281
+ "group_size": 32,
282
+ "bits": 6
283
+ },
284
+ "model.layers.7.mlp.switch_mlp.up_proj": {
285
+ "group_size": 32,
286
+ "bits": 6
287
+ },
288
+ "model.layers.7.mlp.switch_mlp.down_proj": {
289
+ "group_size": 32,
290
+ "bits": 6
291
+ },
292
+ "model.layers.8.self_attn.q_proj": {
293
+ "group_size": 32,
294
+ "bits": 6
295
+ },
296
+ "model.layers.8.self_attn.k_proj": {
297
+ "group_size": 32,
298
+ "bits": 6
299
+ },
300
+ "model.layers.8.self_attn.v_proj": {
301
+ "group_size": 32,
302
+ "bits": 8
303
+ },
304
+ "model.layers.8.self_attn.o_proj": {
305
+ "group_size": 32,
306
+ "bits": 6
307
+ },
308
+ "model.layers.8.mlp.gate": {
309
+ "group_size": 32,
310
+ "bits": 6
311
+ },
312
+ "model.layers.8.mlp.switch_mlp.gate_proj": {
313
+ "group_size": 32,
314
+ "bits": 6
315
+ },
316
+ "model.layers.8.mlp.switch_mlp.up_proj": {
317
+ "group_size": 32,
318
+ "bits": 6
319
+ },
320
+ "model.layers.8.mlp.switch_mlp.down_proj": {
321
+ "group_size": 32,
322
+ "bits": 8
323
+ },
324
+ "model.layers.9.self_attn.q_proj": {
325
+ "group_size": 32,
326
+ "bits": 6
327
+ },
328
+ "model.layers.9.self_attn.k_proj": {
329
+ "group_size": 32,
330
+ "bits": 6
331
+ },
332
+ "model.layers.9.self_attn.v_proj": {
333
+ "group_size": 32,
334
+ "bits": 8
335
+ },
336
+ "model.layers.9.self_attn.o_proj": {
337
+ "group_size": 32,
338
+ "bits": 6
339
+ },
340
+ "model.layers.9.mlp.gate": {
341
+ "group_size": 32,
342
+ "bits": 6
343
+ },
344
+ "model.layers.9.mlp.switch_mlp.gate_proj": {
345
+ "group_size": 32,
346
+ "bits": 6
347
+ },
348
+ "model.layers.9.mlp.switch_mlp.up_proj": {
349
+ "group_size": 32,
350
+ "bits": 6
351
+ },
352
+ "model.layers.9.mlp.switch_mlp.down_proj": {
353
+ "group_size": 32,
354
+ "bits": 6
355
+ },
356
+ "model.layers.10.self_attn.q_proj": {
357
+ "group_size": 32,
358
+ "bits": 6
359
+ },
360
+ "model.layers.10.self_attn.k_proj": {
361
+ "group_size": 32,
362
+ "bits": 6
363
+ },
364
+ "model.layers.10.self_attn.v_proj": {
365
+ "group_size": 32,
366
+ "bits": 8
367
+ },
368
+ "model.layers.10.self_attn.o_proj": {
369
+ "group_size": 32,
370
+ "bits": 6
371
+ },
372
+ "model.layers.10.mlp.gate": {
373
+ "group_size": 32,
374
+ "bits": 6
375
+ },
376
+ "model.layers.10.mlp.switch_mlp.gate_proj": {
377
+ "group_size": 32,
378
+ "bits": 6
379
+ },
380
+ "model.layers.10.mlp.switch_mlp.up_proj": {
381
+ "group_size": 32,
382
+ "bits": 6
383
+ },
384
+ "model.layers.10.mlp.switch_mlp.down_proj": {
385
+ "group_size": 32,
386
+ "bits": 6
387
+ },
388
+ "model.layers.11.self_attn.q_proj": {
389
+ "group_size": 32,
390
+ "bits": 6
391
+ },
392
+ "model.layers.11.self_attn.k_proj": {
393
+ "group_size": 32,
394
+ "bits": 6
395
+ },
396
+ "model.layers.11.self_attn.v_proj": {
397
+ "group_size": 32,
398
+ "bits": 8
399
+ },
400
+ "model.layers.11.self_attn.o_proj": {
401
+ "group_size": 32,
402
+ "bits": 6
403
+ },
404
+ "model.layers.11.mlp.gate": {
405
+ "group_size": 32,
406
+ "bits": 6
407
+ },
408
+ "model.layers.11.mlp.switch_mlp.gate_proj": {
409
+ "group_size": 32,
410
+ "bits": 6
411
+ },
412
+ "model.layers.11.mlp.switch_mlp.up_proj": {
413
+ "group_size": 32,
414
+ "bits": 6
415
+ },
416
+ "model.layers.11.mlp.switch_mlp.down_proj": {
417
+ "group_size": 32,
418
+ "bits": 8
419
+ },
420
+ "model.layers.12.self_attn.q_proj": {
421
+ "group_size": 32,
422
+ "bits": 6
423
+ },
424
+ "model.layers.12.self_attn.k_proj": {
425
+ "group_size": 32,
426
+ "bits": 6
427
+ },
428
+ "model.layers.12.self_attn.v_proj": {
429
+ "group_size": 32,
430
+ "bits": 8
431
+ },
432
+ "model.layers.12.self_attn.o_proj": {
433
+ "group_size": 32,
434
+ "bits": 6
435
+ },
436
+ "model.layers.12.mlp.gate": {
437
+ "group_size": 32,
438
+ "bits": 6
439
+ },
440
+ "model.layers.12.mlp.switch_mlp.gate_proj": {
441
+ "group_size": 32,
442
+ "bits": 6
443
+ },
444
+ "model.layers.12.mlp.switch_mlp.up_proj": {
445
+ "group_size": 32,
446
+ "bits": 6
447
+ },
448
+ "model.layers.12.mlp.switch_mlp.down_proj": {
449
+ "group_size": 32,
450
+ "bits": 6
451
+ },
452
+ "model.layers.13.self_attn.q_proj": {
453
+ "group_size": 32,
454
+ "bits": 6
455
+ },
456
+ "model.layers.13.self_attn.k_proj": {
457
+ "group_size": 32,
458
+ "bits": 6
459
+ },
460
+ "model.layers.13.self_attn.v_proj": {
461
+ "group_size": 32,
462
+ "bits": 8
463
+ },
464
+ "model.layers.13.self_attn.o_proj": {
465
+ "group_size": 32,
466
+ "bits": 6
467
+ },
468
+ "model.layers.13.mlp.gate": {
469
+ "group_size": 32,
470
+ "bits": 6
471
+ },
472
+ "model.layers.13.mlp.switch_mlp.gate_proj": {
473
+ "group_size": 32,
474
+ "bits": 6
475
+ },
476
+ "model.layers.13.mlp.switch_mlp.up_proj": {
477
+ "group_size": 32,
478
+ "bits": 6
479
+ },
480
+ "model.layers.13.mlp.switch_mlp.down_proj": {
481
+ "group_size": 32,
482
+ "bits": 6
483
+ },
484
+ "model.layers.14.self_attn.q_proj": {
485
+ "group_size": 32,
486
+ "bits": 6
487
+ },
488
+ "model.layers.14.self_attn.k_proj": {
489
+ "group_size": 32,
490
+ "bits": 6
491
+ },
492
+ "model.layers.14.self_attn.v_proj": {
493
+ "group_size": 32,
494
+ "bits": 8
495
+ },
496
+ "model.layers.14.self_attn.o_proj": {
497
+ "group_size": 32,
498
+ "bits": 6
499
+ },
500
+ "model.layers.14.mlp.gate": {
501
+ "group_size": 32,
502
+ "bits": 6
503
+ },
504
+ "model.layers.14.mlp.switch_mlp.gate_proj": {
505
+ "group_size": 32,
506
+ "bits": 6
507
+ },
508
+ "model.layers.14.mlp.switch_mlp.up_proj": {
509
+ "group_size": 32,
510
+ "bits": 6
511
+ },
512
+ "model.layers.14.mlp.switch_mlp.down_proj": {
513
+ "group_size": 32,
514
+ "bits": 8
515
+ },
516
+ "model.layers.15.self_attn.q_proj": {
517
+ "group_size": 32,
518
+ "bits": 6
519
+ },
520
+ "model.layers.15.self_attn.k_proj": {
521
+ "group_size": 32,
522
+ "bits": 6
523
+ },
524
+ "model.layers.15.self_attn.v_proj": {
525
+ "group_size": 32,
526
+ "bits": 8
527
+ },
528
+ "model.layers.15.self_attn.o_proj": {
529
+ "group_size": 32,
530
+ "bits": 6
531
+ },
532
+ "model.layers.15.mlp.gate": {
533
+ "group_size": 32,
534
+ "bits": 6
535
+ },
536
+ "model.layers.15.mlp.switch_mlp.gate_proj": {
537
+ "group_size": 32,
538
+ "bits": 6
539
+ },
540
+ "model.layers.15.mlp.switch_mlp.up_proj": {
541
+ "group_size": 32,
542
+ "bits": 6
543
+ },
544
+ "model.layers.15.mlp.switch_mlp.down_proj": {
545
+ "group_size": 32,
546
+ "bits": 6
547
+ },
548
+ "model.layers.16.self_attn.q_proj": {
549
+ "group_size": 32,
550
+ "bits": 6
551
+ },
552
+ "model.layers.16.self_attn.k_proj": {
553
+ "group_size": 32,
554
+ "bits": 6
555
+ },
556
+ "model.layers.16.self_attn.v_proj": {
557
+ "group_size": 32,
558
+ "bits": 8
559
+ },
560
+ "model.layers.16.self_attn.o_proj": {
561
+ "group_size": 32,
562
+ "bits": 6
563
+ },
564
+ "model.layers.16.mlp.gate": {
565
+ "group_size": 32,
566
+ "bits": 6
567
+ },
568
+ "model.layers.16.mlp.switch_mlp.gate_proj": {
569
+ "group_size": 32,
570
+ "bits": 6
571
+ },
572
+ "model.layers.16.mlp.switch_mlp.up_proj": {
573
+ "group_size": 32,
574
+ "bits": 6
575
+ },
576
+ "model.layers.16.mlp.switch_mlp.down_proj": {
577
+ "group_size": 32,
578
+ "bits": 6
579
+ },
580
+ "model.layers.17.self_attn.q_proj": {
581
+ "group_size": 32,
582
+ "bits": 6
583
+ },
584
+ "model.layers.17.self_attn.k_proj": {
585
+ "group_size": 32,
586
+ "bits": 6
587
+ },
588
+ "model.layers.17.self_attn.v_proj": {
589
+ "group_size": 32,
590
+ "bits": 8
591
+ },
592
+ "model.layers.17.self_attn.o_proj": {
593
+ "group_size": 32,
594
+ "bits": 6
595
+ },
596
+ "model.layers.17.mlp.gate": {
597
+ "group_size": 32,
598
+ "bits": 6
599
+ },
600
+ "model.layers.17.mlp.switch_mlp.gate_proj": {
601
+ "group_size": 32,
602
+ "bits": 6
603
+ },
604
+ "model.layers.17.mlp.switch_mlp.up_proj": {
605
+ "group_size": 32,
606
+ "bits": 6
607
+ },
608
+ "model.layers.17.mlp.switch_mlp.down_proj": {
609
+ "group_size": 32,
610
+ "bits": 8
611
+ },
612
+ "model.layers.18.self_attn.q_proj": {
613
+ "group_size": 32,
614
+ "bits": 6
615
+ },
616
+ "model.layers.18.self_attn.k_proj": {
617
+ "group_size": 32,
618
+ "bits": 6
619
+ },
620
+ "model.layers.18.self_attn.v_proj": {
621
+ "group_size": 32,
622
+ "bits": 8
623
+ },
624
+ "model.layers.18.self_attn.o_proj": {
625
+ "group_size": 32,
626
+ "bits": 6
627
+ },
628
+ "model.layers.18.mlp.gate": {
629
+ "group_size": 32,
630
+ "bits": 6
631
+ },
632
+ "model.layers.18.mlp.switch_mlp.gate_proj": {
633
+ "group_size": 32,
634
+ "bits": 6
635
+ },
636
+ "model.layers.18.mlp.switch_mlp.up_proj": {
637
+ "group_size": 32,
638
+ "bits": 6
639
+ },
640
+ "model.layers.18.mlp.switch_mlp.down_proj": {
641
+ "group_size": 32,
642
+ "bits": 6
643
+ },
644
+ "model.layers.19.self_attn.q_proj": {
645
+ "group_size": 32,
646
+ "bits": 6
647
+ },
648
+ "model.layers.19.self_attn.k_proj": {
649
+ "group_size": 32,
650
+ "bits": 6
651
+ },
652
+ "model.layers.19.self_attn.v_proj": {
653
+ "group_size": 32,
654
+ "bits": 8
655
+ },
656
+ "model.layers.19.self_attn.o_proj": {
657
+ "group_size": 32,
658
+ "bits": 6
659
+ },
660
+ "model.layers.19.mlp.gate": {
661
+ "group_size": 32,
662
+ "bits": 6
663
+ },
664
+ "model.layers.19.mlp.switch_mlp.gate_proj": {
665
+ "group_size": 32,
666
+ "bits": 6
667
+ },
668
+ "model.layers.19.mlp.switch_mlp.up_proj": {
669
+ "group_size": 32,
670
+ "bits": 6
671
+ },
672
+ "model.layers.19.mlp.switch_mlp.down_proj": {
673
+ "group_size": 32,
674
+ "bits": 6
675
+ },
676
+ "model.layers.20.self_attn.q_proj": {
677
+ "group_size": 32,
678
+ "bits": 6
679
+ },
680
+ "model.layers.20.self_attn.k_proj": {
681
+ "group_size": 32,
682
+ "bits": 6
683
+ },
684
+ "model.layers.20.self_attn.v_proj": {
685
+ "group_size": 32,
686
+ "bits": 8
687
+ },
688
+ "model.layers.20.self_attn.o_proj": {
689
+ "group_size": 32,
690
+ "bits": 6
691
+ },
692
+ "model.layers.20.mlp.gate": {
693
+ "group_size": 32,
694
+ "bits": 6
695
+ },
696
+ "model.layers.20.mlp.switch_mlp.gate_proj": {
697
+ "group_size": 32,
698
+ "bits": 6
699
+ },
700
+ "model.layers.20.mlp.switch_mlp.up_proj": {
701
+ "group_size": 32,
702
+ "bits": 6
703
+ },
704
+ "model.layers.20.mlp.switch_mlp.down_proj": {
705
+ "group_size": 32,
706
+ "bits": 8
707
+ },
708
+ "model.layers.21.self_attn.q_proj": {
709
+ "group_size": 32,
710
+ "bits": 6
711
+ },
712
+ "model.layers.21.self_attn.k_proj": {
713
+ "group_size": 32,
714
+ "bits": 6
715
+ },
716
+ "model.layers.21.self_attn.v_proj": {
717
+ "group_size": 32,
718
+ "bits": 8
719
+ },
720
+ "model.layers.21.self_attn.o_proj": {
721
+ "group_size": 32,
722
+ "bits": 6
723
+ },
724
+ "model.layers.21.mlp.gate": {
725
+ "group_size": 32,
726
+ "bits": 6
727
+ },
728
+ "model.layers.21.mlp.switch_mlp.gate_proj": {
729
+ "group_size": 32,
730
+ "bits": 6
731
+ },
732
+ "model.layers.21.mlp.switch_mlp.up_proj": {
733
+ "group_size": 32,
734
+ "bits": 6
735
+ },
736
+ "model.layers.21.mlp.switch_mlp.down_proj": {
737
+ "group_size": 32,
738
+ "bits": 6
739
+ },
740
+ "model.layers.22.self_attn.q_proj": {
741
+ "group_size": 32,
742
+ "bits": 6
743
+ },
744
+ "model.layers.22.self_attn.k_proj": {
745
+ "group_size": 32,
746
+ "bits": 6
747
+ },
748
+ "model.layers.22.self_attn.v_proj": {
749
+ "group_size": 32,
750
+ "bits": 8
751
+ },
752
+ "model.layers.22.self_attn.o_proj": {
753
+ "group_size": 32,
754
+ "bits": 6
755
+ },
756
+ "model.layers.22.mlp.gate": {
757
+ "group_size": 32,
758
+ "bits": 6
759
+ },
760
+ "model.layers.22.mlp.switch_mlp.gate_proj": {
761
+ "group_size": 32,
762
+ "bits": 6
763
+ },
764
+ "model.layers.22.mlp.switch_mlp.up_proj": {
765
+ "group_size": 32,
766
+ "bits": 6
767
+ },
768
+ "model.layers.22.mlp.switch_mlp.down_proj": {
769
+ "group_size": 32,
770
+ "bits": 6
771
+ },
772
+ "model.layers.23.self_attn.q_proj": {
773
+ "group_size": 32,
774
+ "bits": 6
775
+ },
776
+ "model.layers.23.self_attn.k_proj": {
777
+ "group_size": 32,
778
+ "bits": 6
779
+ },
780
+ "model.layers.23.self_attn.v_proj": {
781
+ "group_size": 32,
782
+ "bits": 8
783
+ },
784
+ "model.layers.23.self_attn.o_proj": {
785
+ "group_size": 32,
786
+ "bits": 6
787
+ },
788
+ "model.layers.23.mlp.gate": {
789
+ "group_size": 32,
790
+ "bits": 6
791
+ },
792
+ "model.layers.23.mlp.switch_mlp.gate_proj": {
793
+ "group_size": 32,
794
+ "bits": 6
795
+ },
796
+ "model.layers.23.mlp.switch_mlp.up_proj": {
797
+ "group_size": 32,
798
+ "bits": 6
799
+ },
800
+ "model.layers.23.mlp.switch_mlp.down_proj": {
801
+ "group_size": 32,
802
+ "bits": 8
803
+ },
804
+ "model.layers.24.self_attn.q_proj": {
805
+ "group_size": 32,
806
+ "bits": 6
807
+ },
808
+ "model.layers.24.self_attn.k_proj": {
809
+ "group_size": 32,
810
+ "bits": 6
811
+ },
812
+ "model.layers.24.self_attn.v_proj": {
813
+ "group_size": 32,
814
+ "bits": 8
815
+ },
816
+ "model.layers.24.self_attn.o_proj": {
817
+ "group_size": 32,
818
+ "bits": 6
819
+ },
820
+ "model.layers.24.mlp.gate": {
821
+ "group_size": 32,
822
+ "bits": 6
823
+ },
824
+ "model.layers.24.mlp.switch_mlp.gate_proj": {
825
+ "group_size": 32,
826
+ "bits": 6
827
+ },
828
+ "model.layers.24.mlp.switch_mlp.up_proj": {
829
+ "group_size": 32,
830
+ "bits": 6
831
+ },
832
+ "model.layers.24.mlp.switch_mlp.down_proj": {
833
+ "group_size": 32,
834
+ "bits": 6
835
+ },
836
+ "model.layers.25.self_attn.q_proj": {
837
+ "group_size": 32,
838
+ "bits": 6
839
+ },
840
+ "model.layers.25.self_attn.k_proj": {
841
+ "group_size": 32,
842
+ "bits": 6
843
+ },
844
+ "model.layers.25.self_attn.v_proj": {
845
+ "group_size": 32,
846
+ "bits": 8
847
+ },
848
+ "model.layers.25.self_attn.o_proj": {
849
+ "group_size": 32,
850
+ "bits": 6
851
+ },
852
+ "model.layers.25.mlp.gate": {
853
+ "group_size": 32,
854
+ "bits": 6
855
+ },
856
+ "model.layers.25.mlp.switch_mlp.gate_proj": {
857
+ "group_size": 32,
858
+ "bits": 6
859
+ },
860
+ "model.layers.25.mlp.switch_mlp.up_proj": {
861
+ "group_size": 32,
862
+ "bits": 6
863
+ },
864
+ "model.layers.25.mlp.switch_mlp.down_proj": {
865
+ "group_size": 32,
866
+ "bits": 6
867
+ },
868
+ "model.layers.26.self_attn.q_proj": {
869
+ "group_size": 32,
870
+ "bits": 6
871
+ },
872
+ "model.layers.26.self_attn.k_proj": {
873
+ "group_size": 32,
874
+ "bits": 6
875
+ },
876
+ "model.layers.26.self_attn.v_proj": {
877
+ "group_size": 32,
878
+ "bits": 8
879
+ },
880
+ "model.layers.26.self_attn.o_proj": {
881
+ "group_size": 32,
882
+ "bits": 6
883
+ },
884
+ "model.layers.26.mlp.gate": {
885
+ "group_size": 32,
886
+ "bits": 6
887
+ },
888
+ "model.layers.26.mlp.switch_mlp.gate_proj": {
889
+ "group_size": 32,
890
+ "bits": 6
891
+ },
892
+ "model.layers.26.mlp.switch_mlp.up_proj": {
893
+ "group_size": 32,
894
+ "bits": 6
895
+ },
896
+ "model.layers.26.mlp.switch_mlp.down_proj": {
897
+ "group_size": 32,
898
+ "bits": 8
899
+ },
900
+ "model.layers.27.self_attn.q_proj": {
901
+ "group_size": 32,
902
+ "bits": 6
903
+ },
904
+ "model.layers.27.self_attn.k_proj": {
905
+ "group_size": 32,
906
+ "bits": 6
907
+ },
908
+ "model.layers.27.self_attn.v_proj": {
909
+ "group_size": 32,
910
+ "bits": 8
911
+ },
912
+ "model.layers.27.self_attn.o_proj": {
913
+ "group_size": 32,
914
+ "bits": 6
915
+ },
916
+ "model.layers.27.mlp.gate": {
917
+ "group_size": 32,
918
+ "bits": 6
919
+ },
920
+ "model.layers.27.mlp.switch_mlp.gate_proj": {
921
+ "group_size": 32,
922
+ "bits": 6
923
+ },
924
+ "model.layers.27.mlp.switch_mlp.up_proj": {
925
+ "group_size": 32,
926
+ "bits": 6
927
+ },
928
+ "model.layers.27.mlp.switch_mlp.down_proj": {
929
+ "group_size": 32,
930
+ "bits": 6
931
+ },
932
+ "model.layers.28.self_attn.q_proj": {
933
+ "group_size": 32,
934
+ "bits": 6
935
+ },
936
+ "model.layers.28.self_attn.k_proj": {
937
+ "group_size": 32,
938
+ "bits": 6
939
+ },
940
+ "model.layers.28.self_attn.v_proj": {
941
+ "group_size": 32,
942
+ "bits": 8
943
+ },
944
+ "model.layers.28.self_attn.o_proj": {
945
+ "group_size": 32,
946
+ "bits": 6
947
+ },
948
+ "model.layers.28.mlp.gate": {
949
+ "group_size": 32,
950
+ "bits": 6
951
+ },
952
+ "model.layers.28.mlp.switch_mlp.gate_proj": {
953
+ "group_size": 32,
954
+ "bits": 6
955
+ },
956
+ "model.layers.28.mlp.switch_mlp.up_proj": {
957
+ "group_size": 32,
958
+ "bits": 6
959
+ },
960
+ "model.layers.28.mlp.switch_mlp.down_proj": {
961
+ "group_size": 32,
962
+ "bits": 6
963
+ },
964
+ "model.layers.29.self_attn.q_proj": {
965
+ "group_size": 32,
966
+ "bits": 6
967
+ },
968
+ "model.layers.29.self_attn.k_proj": {
969
+ "group_size": 32,
970
+ "bits": 6
971
+ },
972
+ "model.layers.29.self_attn.v_proj": {
973
+ "group_size": 32,
974
+ "bits": 8
975
+ },
976
+ "model.layers.29.self_attn.o_proj": {
977
+ "group_size": 32,
978
+ "bits": 6
979
+ },
980
+ "model.layers.29.mlp.gate": {
981
+ "group_size": 32,
982
+ "bits": 6
983
+ },
984
+ "model.layers.29.mlp.switch_mlp.gate_proj": {
985
+ "group_size": 32,
986
+ "bits": 6
987
+ },
988
+ "model.layers.29.mlp.switch_mlp.up_proj": {
989
+ "group_size": 32,
990
+ "bits": 6
991
+ },
992
+ "model.layers.29.mlp.switch_mlp.down_proj": {
993
+ "group_size": 32,
994
+ "bits": 8
995
+ },
996
+ "model.layers.30.self_attn.q_proj": {
997
+ "group_size": 32,
998
+ "bits": 6
999
+ },
1000
+ "model.layers.30.self_attn.k_proj": {
1001
+ "group_size": 32,
1002
+ "bits": 6
1003
+ },
1004
+ "model.layers.30.self_attn.v_proj": {
1005
+ "group_size": 32,
1006
+ "bits": 8
1007
+ },
1008
+ "model.layers.30.self_attn.o_proj": {
1009
+ "group_size": 32,
1010
+ "bits": 6
1011
+ },
1012
+ "model.layers.30.mlp.gate": {
1013
+ "group_size": 32,
1014
+ "bits": 6
1015
+ },
1016
+ "model.layers.30.mlp.switch_mlp.gate_proj": {
1017
+ "group_size": 32,
1018
+ "bits": 6
1019
+ },
1020
+ "model.layers.30.mlp.switch_mlp.up_proj": {
1021
+ "group_size": 32,
1022
+ "bits": 6
1023
+ },
1024
+ "model.layers.30.mlp.switch_mlp.down_proj": {
1025
+ "group_size": 32,
1026
+ "bits": 6
1027
+ },
1028
+ "model.layers.31.self_attn.q_proj": {
1029
+ "group_size": 32,
1030
+ "bits": 6
1031
+ },
1032
+ "model.layers.31.self_attn.k_proj": {
1033
+ "group_size": 32,
1034
+ "bits": 6
1035
+ },
1036
+ "model.layers.31.self_attn.v_proj": {
1037
+ "group_size": 32,
1038
+ "bits": 8
1039
+ },
1040
+ "model.layers.31.self_attn.o_proj": {
1041
+ "group_size": 32,
1042
+ "bits": 6
1043
+ },
1044
+ "model.layers.31.mlp.gate": {
1045
+ "group_size": 32,
1046
+ "bits": 6
1047
+ },
1048
+ "model.layers.31.mlp.switch_mlp.gate_proj": {
1049
+ "group_size": 32,
1050
+ "bits": 6
1051
+ },
1052
+ "model.layers.31.mlp.switch_mlp.up_proj": {
1053
+ "group_size": 32,
1054
+ "bits": 6
1055
+ },
1056
+ "model.layers.31.mlp.switch_mlp.down_proj": {
1057
+ "group_size": 32,
1058
+ "bits": 6
1059
+ },
1060
+ "model.layers.32.self_attn.q_proj": {
1061
+ "group_size": 32,
1062
+ "bits": 6
1063
+ },
1064
+ "model.layers.32.self_attn.k_proj": {
1065
+ "group_size": 32,
1066
+ "bits": 6
1067
+ },
1068
+ "model.layers.32.self_attn.v_proj": {
1069
+ "group_size": 32,
1070
+ "bits": 8
1071
+ },
1072
+ "model.layers.32.self_attn.o_proj": {
1073
+ "group_size": 32,
1074
+ "bits": 6
1075
+ },
1076
+ "model.layers.32.mlp.gate": {
1077
+ "group_size": 32,
1078
+ "bits": 6
1079
+ },
1080
+ "model.layers.32.mlp.switch_mlp.gate_proj": {
1081
+ "group_size": 32,
1082
+ "bits": 6
1083
+ },
1084
+ "model.layers.32.mlp.switch_mlp.up_proj": {
1085
+ "group_size": 32,
1086
+ "bits": 6
1087
+ },
1088
+ "model.layers.32.mlp.switch_mlp.down_proj": {
1089
+ "group_size": 32,
1090
+ "bits": 8
1091
+ },
1092
+ "model.layers.33.self_attn.q_proj": {
1093
+ "group_size": 32,
1094
+ "bits": 6
1095
+ },
1096
+ "model.layers.33.self_attn.k_proj": {
1097
+ "group_size": 32,
1098
+ "bits": 6
1099
+ },
1100
+ "model.layers.33.self_attn.v_proj": {
1101
+ "group_size": 32,
1102
+ "bits": 8
1103
+ },
1104
+ "model.layers.33.self_attn.o_proj": {
1105
+ "group_size": 32,
1106
+ "bits": 6
1107
+ },
1108
+ "model.layers.33.mlp.gate": {
1109
+ "group_size": 32,
1110
+ "bits": 6
1111
+ },
1112
+ "model.layers.33.mlp.switch_mlp.gate_proj": {
1113
+ "group_size": 32,
1114
+ "bits": 6
1115
+ },
1116
+ "model.layers.33.mlp.switch_mlp.up_proj": {
1117
+ "group_size": 32,
1118
+ "bits": 6
1119
+ },
1120
+ "model.layers.33.mlp.switch_mlp.down_proj": {
1121
+ "group_size": 32,
1122
+ "bits": 6
1123
+ },
1124
+ "model.layers.34.self_attn.q_proj": {
1125
+ "group_size": 32,
1126
+ "bits": 6
1127
+ },
1128
+ "model.layers.34.self_attn.k_proj": {
1129
+ "group_size": 32,
1130
+ "bits": 6
1131
+ },
1132
+ "model.layers.34.self_attn.v_proj": {
1133
+ "group_size": 32,
1134
+ "bits": 8
1135
+ },
1136
+ "model.layers.34.self_attn.o_proj": {
1137
+ "group_size": 32,
1138
+ "bits": 6
1139
+ },
1140
+ "model.layers.34.mlp.gate": {
1141
+ "group_size": 32,
1142
+ "bits": 6
1143
+ },
1144
+ "model.layers.34.mlp.switch_mlp.gate_proj": {
1145
+ "group_size": 32,
1146
+ "bits": 6
1147
+ },
1148
+ "model.layers.34.mlp.switch_mlp.up_proj": {
1149
+ "group_size": 32,
1150
+ "bits": 6
1151
+ },
1152
+ "model.layers.34.mlp.switch_mlp.down_proj": {
1153
+ "group_size": 32,
1154
+ "bits": 6
1155
+ },
1156
+ "model.layers.35.self_attn.q_proj": {
1157
+ "group_size": 32,
1158
+ "bits": 6
1159
+ },
1160
+ "model.layers.35.self_attn.k_proj": {
1161
+ "group_size": 32,
1162
+ "bits": 6
1163
+ },
1164
+ "model.layers.35.self_attn.v_proj": {
1165
+ "group_size": 32,
1166
+ "bits": 8
1167
+ },
1168
+ "model.layers.35.self_attn.o_proj": {
1169
+ "group_size": 32,
1170
+ "bits": 6
1171
+ },
1172
+ "model.layers.35.mlp.gate": {
1173
+ "group_size": 32,
1174
+ "bits": 6
1175
+ },
1176
+ "model.layers.35.mlp.switch_mlp.gate_proj": {
1177
+ "group_size": 32,
1178
+ "bits": 6
1179
+ },
1180
+ "model.layers.35.mlp.switch_mlp.up_proj": {
1181
+ "group_size": 32,
1182
+ "bits": 6
1183
+ },
1184
+ "model.layers.35.mlp.switch_mlp.down_proj": {
1185
+ "group_size": 32,
1186
+ "bits": 8
1187
+ },
1188
+ "model.layers.36.self_attn.q_proj": {
1189
+ "group_size": 32,
1190
+ "bits": 6
1191
+ },
1192
+ "model.layers.36.self_attn.k_proj": {
1193
+ "group_size": 32,
1194
+ "bits": 6
1195
+ },
1196
+ "model.layers.36.self_attn.v_proj": {
1197
+ "group_size": 32,
1198
+ "bits": 8
1199
+ },
1200
+ "model.layers.36.self_attn.o_proj": {
1201
+ "group_size": 32,
1202
+ "bits": 6
1203
+ },
1204
+ "model.layers.36.mlp.gate": {
1205
+ "group_size": 32,
1206
+ "bits": 6
1207
+ },
1208
+ "model.layers.36.mlp.switch_mlp.gate_proj": {
1209
+ "group_size": 32,
1210
+ "bits": 6
1211
+ },
1212
+ "model.layers.36.mlp.switch_mlp.up_proj": {
1213
+ "group_size": 32,
1214
+ "bits": 6
1215
+ },
1216
+ "model.layers.36.mlp.switch_mlp.down_proj": {
1217
+ "group_size": 32,
1218
+ "bits": 6
1219
+ },
1220
+ "model.layers.37.self_attn.q_proj": {
1221
+ "group_size": 32,
1222
+ "bits": 6
1223
+ },
1224
+ "model.layers.37.self_attn.k_proj": {
1225
+ "group_size": 32,
1226
+ "bits": 6
1227
+ },
1228
+ "model.layers.37.self_attn.v_proj": {
1229
+ "group_size": 32,
1230
+ "bits": 8
1231
+ },
1232
+ "model.layers.37.self_attn.o_proj": {
1233
+ "group_size": 32,
1234
+ "bits": 6
1235
+ },
1236
+ "model.layers.37.mlp.gate": {
1237
+ "group_size": 32,
1238
+ "bits": 6
1239
+ },
1240
+ "model.layers.37.mlp.switch_mlp.gate_proj": {
1241
+ "group_size": 32,
1242
+ "bits": 6
1243
+ },
1244
+ "model.layers.37.mlp.switch_mlp.up_proj": {
1245
+ "group_size": 32,
1246
+ "bits": 6
1247
+ },
1248
+ "model.layers.37.mlp.switch_mlp.down_proj": {
1249
+ "group_size": 32,
1250
+ "bits": 6
1251
+ },
1252
+ "model.layers.38.self_attn.q_proj": {
1253
+ "group_size": 32,
1254
+ "bits": 6
1255
+ },
1256
+ "model.layers.38.self_attn.k_proj": {
1257
+ "group_size": 32,
1258
+ "bits": 6
1259
+ },
1260
+ "model.layers.38.self_attn.v_proj": {
1261
+ "group_size": 32,
1262
+ "bits": 8
1263
+ },
1264
+ "model.layers.38.self_attn.o_proj": {
1265
+ "group_size": 32,
1266
+ "bits": 6
1267
+ },
1268
+ "model.layers.38.mlp.gate": {
1269
+ "group_size": 32,
1270
+ "bits": 6
1271
+ },
1272
+ "model.layers.38.mlp.switch_mlp.gate_proj": {
1273
+ "group_size": 32,
1274
+ "bits": 6
1275
+ },
1276
+ "model.layers.38.mlp.switch_mlp.up_proj": {
1277
+ "group_size": 32,
1278
+ "bits": 6
1279
+ },
1280
+ "model.layers.38.mlp.switch_mlp.down_proj": {
1281
+ "group_size": 32,
1282
+ "bits": 8
1283
+ },
1284
+ "model.layers.39.self_attn.q_proj": {
1285
+ "group_size": 32,
1286
+ "bits": 6
1287
+ },
1288
+ "model.layers.39.self_attn.k_proj": {
1289
+ "group_size": 32,
1290
+ "bits": 6
1291
+ },
1292
+ "model.layers.39.self_attn.v_proj": {
1293
+ "group_size": 32,
1294
+ "bits": 8
1295
+ },
1296
+ "model.layers.39.self_attn.o_proj": {
1297
+ "group_size": 32,
1298
+ "bits": 6
1299
+ },
1300
+ "model.layers.39.mlp.gate": {
1301
+ "group_size": 32,
1302
+ "bits": 6
1303
+ },
1304
+ "model.layers.39.mlp.switch_mlp.gate_proj": {
1305
+ "group_size": 32,
1306
+ "bits": 6
1307
+ },
1308
+ "model.layers.39.mlp.switch_mlp.up_proj": {
1309
+ "group_size": 32,
1310
+ "bits": 6
1311
+ },
1312
+ "model.layers.39.mlp.switch_mlp.down_proj": {
1313
+ "group_size": 32,
1314
+ "bits": 6
1315
+ },
1316
+ "model.layers.40.self_attn.q_proj": {
1317
+ "group_size": 32,
1318
+ "bits": 6
1319
+ },
1320
+ "model.layers.40.self_attn.k_proj": {
1321
+ "group_size": 32,
1322
+ "bits": 6
1323
+ },
1324
+ "model.layers.40.self_attn.v_proj": {
1325
+ "group_size": 32,
1326
+ "bits": 8
1327
+ },
1328
+ "model.layers.40.self_attn.o_proj": {
1329
+ "group_size": 32,
1330
+ "bits": 6
1331
+ },
1332
+ "model.layers.40.mlp.gate": {
1333
+ "group_size": 32,
1334
+ "bits": 6
1335
+ },
1336
+ "model.layers.40.mlp.switch_mlp.gate_proj": {
1337
+ "group_size": 32,
1338
+ "bits": 6
1339
+ },
1340
+ "model.layers.40.mlp.switch_mlp.up_proj": {
1341
+ "group_size": 32,
1342
+ "bits": 6
1343
+ },
1344
+ "model.layers.40.mlp.switch_mlp.down_proj": {
1345
+ "group_size": 32,
1346
+ "bits": 6
1347
+ },
1348
+ "model.layers.41.self_attn.q_proj": {
1349
+ "group_size": 32,
1350
+ "bits": 6
1351
+ },
1352
+ "model.layers.41.self_attn.k_proj": {
1353
+ "group_size": 32,
1354
+ "bits": 6
1355
+ },
1356
+ "model.layers.41.self_attn.v_proj": {
1357
+ "group_size": 32,
1358
+ "bits": 8
1359
+ },
1360
+ "model.layers.41.self_attn.o_proj": {
1361
+ "group_size": 32,
1362
+ "bits": 6
1363
+ },
1364
+ "model.layers.41.mlp.gate": {
1365
+ "group_size": 32,
1366
+ "bits": 6
1367
+ },
1368
+ "model.layers.41.mlp.switch_mlp.gate_proj": {
1369
+ "group_size": 32,
1370
+ "bits": 6
1371
+ },
1372
+ "model.layers.41.mlp.switch_mlp.up_proj": {
1373
+ "group_size": 32,
1374
+ "bits": 6
1375
+ },
1376
+ "model.layers.41.mlp.switch_mlp.down_proj": {
1377
+ "group_size": 32,
1378
+ "bits": 8
1379
+ },
1380
+ "model.layers.42.self_attn.q_proj": {
1381
+ "group_size": 32,
1382
+ "bits": 6
1383
+ },
1384
+ "model.layers.42.self_attn.k_proj": {
1385
+ "group_size": 32,
1386
+ "bits": 6
1387
+ },
1388
+ "model.layers.42.self_attn.v_proj": {
1389
+ "group_size": 32,
1390
+ "bits": 8
1391
+ },
1392
+ "model.layers.42.self_attn.o_proj": {
1393
+ "group_size": 32,
1394
+ "bits": 6
1395
+ },
1396
+ "model.layers.42.mlp.gate": {
1397
+ "group_size": 32,
1398
+ "bits": 6
1399
+ },
1400
+ "model.layers.42.mlp.switch_mlp.gate_proj": {
1401
+ "group_size": 32,
1402
+ "bits": 6
1403
+ },
1404
+ "model.layers.42.mlp.switch_mlp.up_proj": {
1405
+ "group_size": 32,
1406
+ "bits": 6
1407
+ },
1408
+ "model.layers.42.mlp.switch_mlp.down_proj": {
1409
+ "group_size": 32,
1410
+ "bits": 8
1411
+ },
1412
+ "model.layers.43.self_attn.q_proj": {
1413
+ "group_size": 32,
1414
+ "bits": 6
1415
+ },
1416
+ "model.layers.43.self_attn.k_proj": {
1417
+ "group_size": 32,
1418
+ "bits": 6
1419
+ },
1420
+ "model.layers.43.self_attn.v_proj": {
1421
+ "group_size": 32,
1422
+ "bits": 8
1423
+ },
1424
+ "model.layers.43.self_attn.o_proj": {
1425
+ "group_size": 32,
1426
+ "bits": 6
1427
+ },
1428
+ "model.layers.43.mlp.gate": {
1429
+ "group_size": 32,
1430
+ "bits": 6
1431
+ },
1432
+ "model.layers.43.mlp.switch_mlp.gate_proj": {
1433
+ "group_size": 32,
1434
+ "bits": 6
1435
+ },
1436
+ "model.layers.43.mlp.switch_mlp.up_proj": {
1437
+ "group_size": 32,
1438
+ "bits": 6
1439
+ },
1440
+ "model.layers.43.mlp.switch_mlp.down_proj": {
1441
+ "group_size": 32,
1442
+ "bits": 8
1443
+ },
1444
+ "model.layers.44.self_attn.q_proj": {
1445
+ "group_size": 32,
1446
+ "bits": 6
1447
+ },
1448
+ "model.layers.44.self_attn.k_proj": {
1449
+ "group_size": 32,
1450
+ "bits": 6
1451
+ },
1452
+ "model.layers.44.self_attn.v_proj": {
1453
+ "group_size": 32,
1454
+ "bits": 8
1455
+ },
1456
+ "model.layers.44.self_attn.o_proj": {
1457
+ "group_size": 32,
1458
+ "bits": 6
1459
+ },
1460
+ "model.layers.44.mlp.gate": {
1461
+ "group_size": 32,
1462
+ "bits": 6
1463
+ },
1464
+ "model.layers.44.mlp.switch_mlp.gate_proj": {
1465
+ "group_size": 32,
1466
+ "bits": 6
1467
+ },
1468
+ "model.layers.44.mlp.switch_mlp.up_proj": {
1469
+ "group_size": 32,
1470
+ "bits": 6
1471
+ },
1472
+ "model.layers.44.mlp.switch_mlp.down_proj": {
1473
+ "group_size": 32,
1474
+ "bits": 8
1475
+ },
1476
+ "model.layers.45.self_attn.q_proj": {
1477
+ "group_size": 32,
1478
+ "bits": 6
1479
+ },
1480
+ "model.layers.45.self_attn.k_proj": {
1481
+ "group_size": 32,
1482
+ "bits": 6
1483
+ },
1484
+ "model.layers.45.self_attn.v_proj": {
1485
+ "group_size": 32,
1486
+ "bits": 8
1487
+ },
1488
+ "model.layers.45.self_attn.o_proj": {
1489
+ "group_size": 32,
1490
+ "bits": 6
1491
+ },
1492
+ "model.layers.45.mlp.gate": {
1493
+ "group_size": 32,
1494
+ "bits": 6
1495
+ },
1496
+ "model.layers.45.mlp.switch_mlp.gate_proj": {
1497
+ "group_size": 32,
1498
+ "bits": 6
1499
+ },
1500
+ "model.layers.45.mlp.switch_mlp.up_proj": {
1501
+ "group_size": 32,
1502
+ "bits": 6
1503
+ },
1504
+ "model.layers.45.mlp.switch_mlp.down_proj": {
1505
+ "group_size": 32,
1506
+ "bits": 8
1507
+ },
1508
+ "model.layers.46.self_attn.q_proj": {
1509
+ "group_size": 32,
1510
+ "bits": 6
1511
+ },
1512
+ "model.layers.46.self_attn.k_proj": {
1513
+ "group_size": 32,
1514
+ "bits": 6
1515
+ },
1516
+ "model.layers.46.self_attn.v_proj": {
1517
+ "group_size": 32,
1518
+ "bits": 8
1519
+ },
1520
+ "model.layers.46.self_attn.o_proj": {
1521
+ "group_size": 32,
1522
+ "bits": 6
1523
+ },
1524
+ "model.layers.46.mlp.gate": {
1525
+ "group_size": 32,
1526
+ "bits": 6
1527
+ },
1528
+ "model.layers.46.mlp.switch_mlp.gate_proj": {
1529
+ "group_size": 32,
1530
+ "bits": 6
1531
+ },
1532
+ "model.layers.46.mlp.switch_mlp.up_proj": {
1533
+ "group_size": 32,
1534
+ "bits": 6
1535
+ },
1536
+ "model.layers.46.mlp.switch_mlp.down_proj": {
1537
+ "group_size": 32,
1538
+ "bits": 8
1539
+ },
1540
+ "model.layers.47.self_attn.q_proj": {
1541
+ "group_size": 32,
1542
+ "bits": 6
1543
+ },
1544
+ "model.layers.47.self_attn.k_proj": {
1545
+ "group_size": 32,
1546
+ "bits": 6
1547
+ },
1548
+ "model.layers.47.self_attn.v_proj": {
1549
+ "group_size": 32,
1550
+ "bits": 8
1551
+ },
1552
+ "model.layers.47.self_attn.o_proj": {
1553
+ "group_size": 32,
1554
+ "bits": 6
1555
+ },
1556
+ "model.layers.47.mlp.gate": {
1557
+ "group_size": 32,
1558
+ "bits": 6
1559
+ },
1560
+ "model.layers.47.mlp.switch_mlp.gate_proj": {
1561
+ "group_size": 32,
1562
+ "bits": 6
1563
+ },
1564
+ "model.layers.47.mlp.switch_mlp.up_proj": {
1565
+ "group_size": 32,
1566
+ "bits": 6
1567
+ },
1568
+ "model.layers.47.mlp.switch_mlp.down_proj": {
1569
+ "group_size": 32,
1570
+ "bits": 8
1571
+ },
1572
+ "lm_head": {
1573
+ "group_size": 32,
1574
+ "bits": 8
1575
+ }
1576
+ },
1577
+ "quantization_config": {
1578
+ "group_size": 32,
1579
+ "bits": 8,
1580
+ "mode": "affine",
1581
+ "model.embed_tokens": {
1582
+ "group_size": 32,
1583
+ "bits": 6
1584
+ },
1585
+ "model.layers.0.self_attn.q_proj": {
1586
+ "group_size": 32,
1587
+ "bits": 8
1588
+ },
1589
+ "model.layers.0.self_attn.k_proj": {
1590
+ "group_size": 32,
1591
+ "bits": 8
1592
+ },
1593
+ "model.layers.0.self_attn.v_proj": {
1594
+ "group_size": 32,
1595
+ "bits": 8
1596
+ },
1597
+ "model.layers.0.self_attn.o_proj": {
1598
+ "group_size": 32,
1599
+ "bits": 8
1600
+ },
1601
+ "model.layers.0.mlp.gate": {
1602
+ "group_size": 32,
1603
+ "bits": 8
1604
+ },
1605
+ "model.layers.0.mlp.switch_mlp.gate_proj": {
1606
+ "group_size": 32,
1607
+ "bits": 8
1608
+ },
1609
+ "model.layers.0.mlp.switch_mlp.up_proj": {
1610
+ "group_size": 32,
1611
+ "bits": 8
1612
+ },
1613
+ "model.layers.0.mlp.switch_mlp.down_proj": {
1614
+ "group_size": 32,
1615
+ "bits": 8
1616
+ },
1617
+ "model.layers.1.self_attn.q_proj": {
1618
+ "group_size": 32,
1619
+ "bits": 6
1620
+ },
1621
+ "model.layers.1.self_attn.k_proj": {
1622
+ "group_size": 32,
1623
+ "bits": 6
1624
+ },
1625
+ "model.layers.1.self_attn.v_proj": {
1626
+ "group_size": 32,
1627
+ "bits": 8
1628
+ },
1629
+ "model.layers.1.self_attn.o_proj": {
1630
+ "group_size": 32,
1631
+ "bits": 6
1632
+ },
1633
+ "model.layers.1.mlp.gate": {
1634
+ "group_size": 32,
1635
+ "bits": 6
1636
+ },
1637
+ "model.layers.1.mlp.switch_mlp.gate_proj": {
1638
+ "group_size": 32,
1639
+ "bits": 6
1640
+ },
1641
+ "model.layers.1.mlp.switch_mlp.up_proj": {
1642
+ "group_size": 32,
1643
+ "bits": 6
1644
+ },
1645
+ "model.layers.1.mlp.switch_mlp.down_proj": {
1646
+ "group_size": 32,
1647
+ "bits": 8
1648
+ },
1649
+ "model.layers.2.self_attn.q_proj": {
1650
+ "group_size": 32,
1651
+ "bits": 6
1652
+ },
1653
+ "model.layers.2.self_attn.k_proj": {
1654
+ "group_size": 32,
1655
+ "bits": 6
1656
+ },
1657
+ "model.layers.2.self_attn.v_proj": {
1658
+ "group_size": 32,
1659
+ "bits": 8
1660
+ },
1661
+ "model.layers.2.self_attn.o_proj": {
1662
+ "group_size": 32,
1663
+ "bits": 6
1664
+ },
1665
+ "model.layers.2.mlp.gate": {
1666
+ "group_size": 32,
1667
+ "bits": 6
1668
+ },
1669
+ "model.layers.2.mlp.switch_mlp.gate_proj": {
1670
+ "group_size": 32,
1671
+ "bits": 6
1672
+ },
1673
+ "model.layers.2.mlp.switch_mlp.up_proj": {
1674
+ "group_size": 32,
1675
+ "bits": 6
1676
+ },
1677
+ "model.layers.2.mlp.switch_mlp.down_proj": {
1678
+ "group_size": 32,
1679
+ "bits": 8
1680
+ },
1681
+ "model.layers.3.self_attn.q_proj": {
1682
+ "group_size": 32,
1683
+ "bits": 6
1684
+ },
1685
+ "model.layers.3.self_attn.k_proj": {
1686
+ "group_size": 32,
1687
+ "bits": 6
1688
+ },
1689
+ "model.layers.3.self_attn.v_proj": {
1690
+ "group_size": 32,
1691
+ "bits": 8
1692
+ },
1693
+ "model.layers.3.self_attn.o_proj": {
1694
+ "group_size": 32,
1695
+ "bits": 6
1696
+ },
1697
+ "model.layers.3.mlp.gate": {
1698
+ "group_size": 32,
1699
+ "bits": 6
1700
+ },
1701
+ "model.layers.3.mlp.switch_mlp.gate_proj": {
1702
+ "group_size": 32,
1703
+ "bits": 6
1704
+ },
1705
+ "model.layers.3.mlp.switch_mlp.up_proj": {
1706
+ "group_size": 32,
1707
+ "bits": 6
1708
+ },
1709
+ "model.layers.3.mlp.switch_mlp.down_proj": {
1710
+ "group_size": 32,
1711
+ "bits": 8
1712
+ },
1713
+ "model.layers.4.self_attn.q_proj": {
1714
+ "group_size": 32,
1715
+ "bits": 6
1716
+ },
1717
+ "model.layers.4.self_attn.k_proj": {
1718
+ "group_size": 32,
1719
+ "bits": 6
1720
+ },
1721
+ "model.layers.4.self_attn.v_proj": {
1722
+ "group_size": 32,
1723
+ "bits": 8
1724
+ },
1725
+ "model.layers.4.self_attn.o_proj": {
1726
+ "group_size": 32,
1727
+ "bits": 6
1728
+ },
1729
+ "model.layers.4.mlp.gate": {
1730
+ "group_size": 32,
1731
+ "bits": 6
1732
+ },
1733
+ "model.layers.4.mlp.switch_mlp.gate_proj": {
1734
+ "group_size": 32,
1735
+ "bits": 6
1736
+ },
1737
+ "model.layers.4.mlp.switch_mlp.up_proj": {
1738
+ "group_size": 32,
1739
+ "bits": 6
1740
+ },
1741
+ "model.layers.4.mlp.switch_mlp.down_proj": {
1742
+ "group_size": 32,
1743
+ "bits": 8
1744
+ },
1745
+ "model.layers.5.self_attn.q_proj": {
1746
+ "group_size": 32,
1747
+ "bits": 6
1748
+ },
1749
+ "model.layers.5.self_attn.k_proj": {
1750
+ "group_size": 32,
1751
+ "bits": 6
1752
+ },
1753
+ "model.layers.5.self_attn.v_proj": {
1754
+ "group_size": 32,
1755
+ "bits": 8
1756
+ },
1757
+ "model.layers.5.self_attn.o_proj": {
1758
+ "group_size": 32,
1759
+ "bits": 6
1760
+ },
1761
+ "model.layers.5.mlp.gate": {
1762
+ "group_size": 32,
1763
+ "bits": 6
1764
+ },
1765
+ "model.layers.5.mlp.switch_mlp.gate_proj": {
1766
+ "group_size": 32,
1767
+ "bits": 6
1768
+ },
1769
+ "model.layers.5.mlp.switch_mlp.up_proj": {
1770
+ "group_size": 32,
1771
+ "bits": 6
1772
+ },
1773
+ "model.layers.5.mlp.switch_mlp.down_proj": {
1774
+ "group_size": 32,
1775
+ "bits": 8
1776
+ },
1777
+ "model.layers.6.self_attn.q_proj": {
1778
+ "group_size": 32,
1779
+ "bits": 6
1780
+ },
1781
+ "model.layers.6.self_attn.k_proj": {
1782
+ "group_size": 32,
1783
+ "bits": 6
1784
+ },
1785
+ "model.layers.6.self_attn.v_proj": {
1786
+ "group_size": 32,
1787
+ "bits": 8
1788
+ },
1789
+ "model.layers.6.self_attn.o_proj": {
1790
+ "group_size": 32,
1791
+ "bits": 6
1792
+ },
1793
+ "model.layers.6.mlp.gate": {
1794
+ "group_size": 32,
1795
+ "bits": 6
1796
+ },
1797
+ "model.layers.6.mlp.switch_mlp.gate_proj": {
1798
+ "group_size": 32,
1799
+ "bits": 6
1800
+ },
1801
+ "model.layers.6.mlp.switch_mlp.up_proj": {
1802
+ "group_size": 32,
1803
+ "bits": 6
1804
+ },
1805
+ "model.layers.6.mlp.switch_mlp.down_proj": {
1806
+ "group_size": 32,
1807
+ "bits": 6
1808
+ },
1809
+ "model.layers.7.self_attn.q_proj": {
1810
+ "group_size": 32,
1811
+ "bits": 6
1812
+ },
1813
+ "model.layers.7.self_attn.k_proj": {
1814
+ "group_size": 32,
1815
+ "bits": 6
1816
+ },
1817
+ "model.layers.7.self_attn.v_proj": {
1818
+ "group_size": 32,
1819
+ "bits": 8
1820
+ },
1821
+ "model.layers.7.self_attn.o_proj": {
1822
+ "group_size": 32,
1823
+ "bits": 6
1824
+ },
1825
+ "model.layers.7.mlp.gate": {
1826
+ "group_size": 32,
1827
+ "bits": 6
1828
+ },
1829
+ "model.layers.7.mlp.switch_mlp.gate_proj": {
1830
+ "group_size": 32,
1831
+ "bits": 6
1832
+ },
1833
+ "model.layers.7.mlp.switch_mlp.up_proj": {
1834
+ "group_size": 32,
1835
+ "bits": 6
1836
+ },
1837
+ "model.layers.7.mlp.switch_mlp.down_proj": {
1838
+ "group_size": 32,
1839
+ "bits": 6
1840
+ },
1841
+ "model.layers.8.self_attn.q_proj": {
1842
+ "group_size": 32,
1843
+ "bits": 6
1844
+ },
1845
+ "model.layers.8.self_attn.k_proj": {
1846
+ "group_size": 32,
1847
+ "bits": 6
1848
+ },
1849
+ "model.layers.8.self_attn.v_proj": {
1850
+ "group_size": 32,
1851
+ "bits": 8
1852
+ },
1853
+ "model.layers.8.self_attn.o_proj": {
1854
+ "group_size": 32,
1855
+ "bits": 6
1856
+ },
1857
+ "model.layers.8.mlp.gate": {
1858
+ "group_size": 32,
1859
+ "bits": 6
1860
+ },
1861
+ "model.layers.8.mlp.switch_mlp.gate_proj": {
1862
+ "group_size": 32,
1863
+ "bits": 6
1864
+ },
1865
+ "model.layers.8.mlp.switch_mlp.up_proj": {
1866
+ "group_size": 32,
1867
+ "bits": 6
1868
+ },
1869
+ "model.layers.8.mlp.switch_mlp.down_proj": {
1870
+ "group_size": 32,
1871
+ "bits": 8
1872
+ },
1873
+ "model.layers.9.self_attn.q_proj": {
1874
+ "group_size": 32,
1875
+ "bits": 6
1876
+ },
1877
+ "model.layers.9.self_attn.k_proj": {
1878
+ "group_size": 32,
1879
+ "bits": 6
1880
+ },
1881
+ "model.layers.9.self_attn.v_proj": {
1882
+ "group_size": 32,
1883
+ "bits": 8
1884
+ },
1885
+ "model.layers.9.self_attn.o_proj": {
1886
+ "group_size": 32,
1887
+ "bits": 6
1888
+ },
1889
+ "model.layers.9.mlp.gate": {
1890
+ "group_size": 32,
1891
+ "bits": 6
1892
+ },
1893
+ "model.layers.9.mlp.switch_mlp.gate_proj": {
1894
+ "group_size": 32,
1895
+ "bits": 6
1896
+ },
1897
+ "model.layers.9.mlp.switch_mlp.up_proj": {
1898
+ "group_size": 32,
1899
+ "bits": 6
1900
+ },
1901
+ "model.layers.9.mlp.switch_mlp.down_proj": {
1902
+ "group_size": 32,
1903
+ "bits": 6
1904
+ },
1905
+ "model.layers.10.self_attn.q_proj": {
1906
+ "group_size": 32,
1907
+ "bits": 6
1908
+ },
1909
+ "model.layers.10.self_attn.k_proj": {
1910
+ "group_size": 32,
1911
+ "bits": 6
1912
+ },
1913
+ "model.layers.10.self_attn.v_proj": {
1914
+ "group_size": 32,
1915
+ "bits": 8
1916
+ },
1917
+ "model.layers.10.self_attn.o_proj": {
1918
+ "group_size": 32,
1919
+ "bits": 6
1920
+ },
1921
+ "model.layers.10.mlp.gate": {
1922
+ "group_size": 32,
1923
+ "bits": 6
1924
+ },
1925
+ "model.layers.10.mlp.switch_mlp.gate_proj": {
1926
+ "group_size": 32,
1927
+ "bits": 6
1928
+ },
1929
+ "model.layers.10.mlp.switch_mlp.up_proj": {
1930
+ "group_size": 32,
1931
+ "bits": 6
1932
+ },
1933
+ "model.layers.10.mlp.switch_mlp.down_proj": {
1934
+ "group_size": 32,
1935
+ "bits": 6
1936
+ },
1937
+ "model.layers.11.self_attn.q_proj": {
1938
+ "group_size": 32,
1939
+ "bits": 6
1940
+ },
1941
+ "model.layers.11.self_attn.k_proj": {
1942
+ "group_size": 32,
1943
+ "bits": 6
1944
+ },
1945
+ "model.layers.11.self_attn.v_proj": {
1946
+ "group_size": 32,
1947
+ "bits": 8
1948
+ },
1949
+ "model.layers.11.self_attn.o_proj": {
1950
+ "group_size": 32,
1951
+ "bits": 6
1952
+ },
1953
+ "model.layers.11.mlp.gate": {
1954
+ "group_size": 32,
1955
+ "bits": 6
1956
+ },
1957
+ "model.layers.11.mlp.switch_mlp.gate_proj": {
1958
+ "group_size": 32,
1959
+ "bits": 6
1960
+ },
1961
+ "model.layers.11.mlp.switch_mlp.up_proj": {
1962
+ "group_size": 32,
1963
+ "bits": 6
1964
+ },
1965
+ "model.layers.11.mlp.switch_mlp.down_proj": {
1966
+ "group_size": 32,
1967
+ "bits": 8
1968
+ },
1969
+ "model.layers.12.self_attn.q_proj": {
1970
+ "group_size": 32,
1971
+ "bits": 6
1972
+ },
1973
+ "model.layers.12.self_attn.k_proj": {
1974
+ "group_size": 32,
1975
+ "bits": 6
1976
+ },
1977
+ "model.layers.12.self_attn.v_proj": {
1978
+ "group_size": 32,
1979
+ "bits": 8
1980
+ },
1981
+ "model.layers.12.self_attn.o_proj": {
1982
+ "group_size": 32,
1983
+ "bits": 6
1984
+ },
1985
+ "model.layers.12.mlp.gate": {
1986
+ "group_size": 32,
1987
+ "bits": 6
1988
+ },
1989
+ "model.layers.12.mlp.switch_mlp.gate_proj": {
1990
+ "group_size": 32,
1991
+ "bits": 6
1992
+ },
1993
+ "model.layers.12.mlp.switch_mlp.up_proj": {
1994
+ "group_size": 32,
1995
+ "bits": 6
1996
+ },
1997
+ "model.layers.12.mlp.switch_mlp.down_proj": {
1998
+ "group_size": 32,
1999
+ "bits": 6
2000
+ },
2001
+ "model.layers.13.self_attn.q_proj": {
2002
+ "group_size": 32,
2003
+ "bits": 6
2004
+ },
2005
+ "model.layers.13.self_attn.k_proj": {
2006
+ "group_size": 32,
2007
+ "bits": 6
2008
+ },
2009
+ "model.layers.13.self_attn.v_proj": {
2010
+ "group_size": 32,
2011
+ "bits": 8
2012
+ },
2013
+ "model.layers.13.self_attn.o_proj": {
2014
+ "group_size": 32,
2015
+ "bits": 6
2016
+ },
2017
+ "model.layers.13.mlp.gate": {
2018
+ "group_size": 32,
2019
+ "bits": 6
2020
+ },
2021
+ "model.layers.13.mlp.switch_mlp.gate_proj": {
2022
+ "group_size": 32,
2023
+ "bits": 6
2024
+ },
2025
+ "model.layers.13.mlp.switch_mlp.up_proj": {
2026
+ "group_size": 32,
2027
+ "bits": 6
2028
+ },
2029
+ "model.layers.13.mlp.switch_mlp.down_proj": {
2030
+ "group_size": 32,
2031
+ "bits": 6
2032
+ },
2033
+ "model.layers.14.self_attn.q_proj": {
2034
+ "group_size": 32,
2035
+ "bits": 6
2036
+ },
2037
+ "model.layers.14.self_attn.k_proj": {
2038
+ "group_size": 32,
2039
+ "bits": 6
2040
+ },
2041
+ "model.layers.14.self_attn.v_proj": {
2042
+ "group_size": 32,
2043
+ "bits": 8
2044
+ },
2045
+ "model.layers.14.self_attn.o_proj": {
2046
+ "group_size": 32,
2047
+ "bits": 6
2048
+ },
2049
+ "model.layers.14.mlp.gate": {
2050
+ "group_size": 32,
2051
+ "bits": 6
2052
+ },
2053
+ "model.layers.14.mlp.switch_mlp.gate_proj": {
2054
+ "group_size": 32,
2055
+ "bits": 6
2056
+ },
2057
+ "model.layers.14.mlp.switch_mlp.up_proj": {
2058
+ "group_size": 32,
2059
+ "bits": 6
2060
+ },
2061
+ "model.layers.14.mlp.switch_mlp.down_proj": {
2062
+ "group_size": 32,
2063
+ "bits": 8
2064
+ },
2065
+ "model.layers.15.self_attn.q_proj": {
2066
+ "group_size": 32,
2067
+ "bits": 6
2068
+ },
2069
+ "model.layers.15.self_attn.k_proj": {
2070
+ "group_size": 32,
2071
+ "bits": 6
2072
+ },
2073
+ "model.layers.15.self_attn.v_proj": {
2074
+ "group_size": 32,
2075
+ "bits": 8
2076
+ },
2077
+ "model.layers.15.self_attn.o_proj": {
2078
+ "group_size": 32,
2079
+ "bits": 6
2080
+ },
2081
+ "model.layers.15.mlp.gate": {
2082
+ "group_size": 32,
2083
+ "bits": 6
2084
+ },
2085
+ "model.layers.15.mlp.switch_mlp.gate_proj": {
2086
+ "group_size": 32,
2087
+ "bits": 6
2088
+ },
2089
+ "model.layers.15.mlp.switch_mlp.up_proj": {
2090
+ "group_size": 32,
2091
+ "bits": 6
2092
+ },
2093
+ "model.layers.15.mlp.switch_mlp.down_proj": {
2094
+ "group_size": 32,
2095
+ "bits": 6
2096
+ },
2097
+ "model.layers.16.self_attn.q_proj": {
2098
+ "group_size": 32,
2099
+ "bits": 6
2100
+ },
2101
+ "model.layers.16.self_attn.k_proj": {
2102
+ "group_size": 32,
2103
+ "bits": 6
2104
+ },
2105
+ "model.layers.16.self_attn.v_proj": {
2106
+ "group_size": 32,
2107
+ "bits": 8
2108
+ },
2109
+ "model.layers.16.self_attn.o_proj": {
2110
+ "group_size": 32,
2111
+ "bits": 6
2112
+ },
2113
+ "model.layers.16.mlp.gate": {
2114
+ "group_size": 32,
2115
+ "bits": 6
2116
+ },
2117
+ "model.layers.16.mlp.switch_mlp.gate_proj": {
2118
+ "group_size": 32,
2119
+ "bits": 6
2120
+ },
2121
+ "model.layers.16.mlp.switch_mlp.up_proj": {
2122
+ "group_size": 32,
2123
+ "bits": 6
2124
+ },
2125
+ "model.layers.16.mlp.switch_mlp.down_proj": {
2126
+ "group_size": 32,
2127
+ "bits": 6
2128
+ },
2129
+ "model.layers.17.self_attn.q_proj": {
2130
+ "group_size": 32,
2131
+ "bits": 6
2132
+ },
2133
+ "model.layers.17.self_attn.k_proj": {
2134
+ "group_size": 32,
2135
+ "bits": 6
2136
+ },
2137
+ "model.layers.17.self_attn.v_proj": {
2138
+ "group_size": 32,
2139
+ "bits": 8
2140
+ },
2141
+ "model.layers.17.self_attn.o_proj": {
2142
+ "group_size": 32,
2143
+ "bits": 6
2144
+ },
2145
+ "model.layers.17.mlp.gate": {
2146
+ "group_size": 32,
2147
+ "bits": 6
2148
+ },
2149
+ "model.layers.17.mlp.switch_mlp.gate_proj": {
2150
+ "group_size": 32,
2151
+ "bits": 6
2152
+ },
2153
+ "model.layers.17.mlp.switch_mlp.up_proj": {
2154
+ "group_size": 32,
2155
+ "bits": 6
2156
+ },
2157
+ "model.layers.17.mlp.switch_mlp.down_proj": {
2158
+ "group_size": 32,
2159
+ "bits": 8
2160
+ },
2161
+ "model.layers.18.self_attn.q_proj": {
2162
+ "group_size": 32,
2163
+ "bits": 6
2164
+ },
2165
+ "model.layers.18.self_attn.k_proj": {
2166
+ "group_size": 32,
2167
+ "bits": 6
2168
+ },
2169
+ "model.layers.18.self_attn.v_proj": {
2170
+ "group_size": 32,
2171
+ "bits": 8
2172
+ },
2173
+ "model.layers.18.self_attn.o_proj": {
2174
+ "group_size": 32,
2175
+ "bits": 6
2176
+ },
2177
+ "model.layers.18.mlp.gate": {
2178
+ "group_size": 32,
2179
+ "bits": 6
2180
+ },
2181
+ "model.layers.18.mlp.switch_mlp.gate_proj": {
2182
+ "group_size": 32,
2183
+ "bits": 6
2184
+ },
2185
+ "model.layers.18.mlp.switch_mlp.up_proj": {
2186
+ "group_size": 32,
2187
+ "bits": 6
2188
+ },
2189
+ "model.layers.18.mlp.switch_mlp.down_proj": {
2190
+ "group_size": 32,
2191
+ "bits": 6
2192
+ },
2193
+ "model.layers.19.self_attn.q_proj": {
2194
+ "group_size": 32,
2195
+ "bits": 6
2196
+ },
2197
+ "model.layers.19.self_attn.k_proj": {
2198
+ "group_size": 32,
2199
+ "bits": 6
2200
+ },
2201
+ "model.layers.19.self_attn.v_proj": {
2202
+ "group_size": 32,
2203
+ "bits": 8
2204
+ },
2205
+ "model.layers.19.self_attn.o_proj": {
2206
+ "group_size": 32,
2207
+ "bits": 6
2208
+ },
2209
+ "model.layers.19.mlp.gate": {
2210
+ "group_size": 32,
2211
+ "bits": 6
2212
+ },
2213
+ "model.layers.19.mlp.switch_mlp.gate_proj": {
2214
+ "group_size": 32,
2215
+ "bits": 6
2216
+ },
2217
+ "model.layers.19.mlp.switch_mlp.up_proj": {
2218
+ "group_size": 32,
2219
+ "bits": 6
2220
+ },
2221
+ "model.layers.19.mlp.switch_mlp.down_proj": {
2222
+ "group_size": 32,
2223
+ "bits": 6
2224
+ },
2225
+ "model.layers.20.self_attn.q_proj": {
2226
+ "group_size": 32,
2227
+ "bits": 6
2228
+ },
2229
+ "model.layers.20.self_attn.k_proj": {
2230
+ "group_size": 32,
2231
+ "bits": 6
2232
+ },
2233
+ "model.layers.20.self_attn.v_proj": {
2234
+ "group_size": 32,
2235
+ "bits": 8
2236
+ },
2237
+ "model.layers.20.self_attn.o_proj": {
2238
+ "group_size": 32,
2239
+ "bits": 6
2240
+ },
2241
+ "model.layers.20.mlp.gate": {
2242
+ "group_size": 32,
2243
+ "bits": 6
2244
+ },
2245
+ "model.layers.20.mlp.switch_mlp.gate_proj": {
2246
+ "group_size": 32,
2247
+ "bits": 6
2248
+ },
2249
+ "model.layers.20.mlp.switch_mlp.up_proj": {
2250
+ "group_size": 32,
2251
+ "bits": 6
2252
+ },
2253
+ "model.layers.20.mlp.switch_mlp.down_proj": {
2254
+ "group_size": 32,
2255
+ "bits": 8
2256
+ },
2257
+ "model.layers.21.self_attn.q_proj": {
2258
+ "group_size": 32,
2259
+ "bits": 6
2260
+ },
2261
+ "model.layers.21.self_attn.k_proj": {
2262
+ "group_size": 32,
2263
+ "bits": 6
2264
+ },
2265
+ "model.layers.21.self_attn.v_proj": {
2266
+ "group_size": 32,
2267
+ "bits": 8
2268
+ },
2269
+ "model.layers.21.self_attn.o_proj": {
2270
+ "group_size": 32,
2271
+ "bits": 6
2272
+ },
2273
+ "model.layers.21.mlp.gate": {
2274
+ "group_size": 32,
2275
+ "bits": 6
2276
+ },
2277
+ "model.layers.21.mlp.switch_mlp.gate_proj": {
2278
+ "group_size": 32,
2279
+ "bits": 6
2280
+ },
2281
+ "model.layers.21.mlp.switch_mlp.up_proj": {
2282
+ "group_size": 32,
2283
+ "bits": 6
2284
+ },
2285
+ "model.layers.21.mlp.switch_mlp.down_proj": {
2286
+ "group_size": 32,
2287
+ "bits": 6
2288
+ },
2289
+ "model.layers.22.self_attn.q_proj": {
2290
+ "group_size": 32,
2291
+ "bits": 6
2292
+ },
2293
+ "model.layers.22.self_attn.k_proj": {
2294
+ "group_size": 32,
2295
+ "bits": 6
2296
+ },
2297
+ "model.layers.22.self_attn.v_proj": {
2298
+ "group_size": 32,
2299
+ "bits": 8
2300
+ },
2301
+ "model.layers.22.self_attn.o_proj": {
2302
+ "group_size": 32,
2303
+ "bits": 6
2304
+ },
2305
+ "model.layers.22.mlp.gate": {
2306
+ "group_size": 32,
2307
+ "bits": 6
2308
+ },
2309
+ "model.layers.22.mlp.switch_mlp.gate_proj": {
2310
+ "group_size": 32,
2311
+ "bits": 6
2312
+ },
2313
+ "model.layers.22.mlp.switch_mlp.up_proj": {
2314
+ "group_size": 32,
2315
+ "bits": 6
2316
+ },
2317
+ "model.layers.22.mlp.switch_mlp.down_proj": {
2318
+ "group_size": 32,
2319
+ "bits": 6
2320
+ },
2321
+ "model.layers.23.self_attn.q_proj": {
2322
+ "group_size": 32,
2323
+ "bits": 6
2324
+ },
2325
+ "model.layers.23.self_attn.k_proj": {
2326
+ "group_size": 32,
2327
+ "bits": 6
2328
+ },
2329
+ "model.layers.23.self_attn.v_proj": {
2330
+ "group_size": 32,
2331
+ "bits": 8
2332
+ },
2333
+ "model.layers.23.self_attn.o_proj": {
2334
+ "group_size": 32,
2335
+ "bits": 6
2336
+ },
2337
+ "model.layers.23.mlp.gate": {
2338
+ "group_size": 32,
2339
+ "bits": 6
2340
+ },
2341
+ "model.layers.23.mlp.switch_mlp.gate_proj": {
2342
+ "group_size": 32,
2343
+ "bits": 6
2344
+ },
2345
+ "model.layers.23.mlp.switch_mlp.up_proj": {
2346
+ "group_size": 32,
2347
+ "bits": 6
2348
+ },
2349
+ "model.layers.23.mlp.switch_mlp.down_proj": {
2350
+ "group_size": 32,
2351
+ "bits": 8
2352
+ },
2353
+ "model.layers.24.self_attn.q_proj": {
2354
+ "group_size": 32,
2355
+ "bits": 6
2356
+ },
2357
+ "model.layers.24.self_attn.k_proj": {
2358
+ "group_size": 32,
2359
+ "bits": 6
2360
+ },
2361
+ "model.layers.24.self_attn.v_proj": {
2362
+ "group_size": 32,
2363
+ "bits": 8
2364
+ },
2365
+ "model.layers.24.self_attn.o_proj": {
2366
+ "group_size": 32,
2367
+ "bits": 6
2368
+ },
2369
+ "model.layers.24.mlp.gate": {
2370
+ "group_size": 32,
2371
+ "bits": 6
2372
+ },
2373
+ "model.layers.24.mlp.switch_mlp.gate_proj": {
2374
+ "group_size": 32,
2375
+ "bits": 6
2376
+ },
2377
+ "model.layers.24.mlp.switch_mlp.up_proj": {
2378
+ "group_size": 32,
2379
+ "bits": 6
2380
+ },
2381
+ "model.layers.24.mlp.switch_mlp.down_proj": {
2382
+ "group_size": 32,
2383
+ "bits": 6
2384
+ },
2385
+ "model.layers.25.self_attn.q_proj": {
2386
+ "group_size": 32,
2387
+ "bits": 6
2388
+ },
2389
+ "model.layers.25.self_attn.k_proj": {
2390
+ "group_size": 32,
2391
+ "bits": 6
2392
+ },
2393
+ "model.layers.25.self_attn.v_proj": {
2394
+ "group_size": 32,
2395
+ "bits": 8
2396
+ },
2397
+ "model.layers.25.self_attn.o_proj": {
2398
+ "group_size": 32,
2399
+ "bits": 6
2400
+ },
2401
+ "model.layers.25.mlp.gate": {
2402
+ "group_size": 32,
2403
+ "bits": 6
2404
+ },
2405
+ "model.layers.25.mlp.switch_mlp.gate_proj": {
2406
+ "group_size": 32,
2407
+ "bits": 6
2408
+ },
2409
+ "model.layers.25.mlp.switch_mlp.up_proj": {
2410
+ "group_size": 32,
2411
+ "bits": 6
2412
+ },
2413
+ "model.layers.25.mlp.switch_mlp.down_proj": {
2414
+ "group_size": 32,
2415
+ "bits": 6
2416
+ },
2417
+ "model.layers.26.self_attn.q_proj": {
2418
+ "group_size": 32,
2419
+ "bits": 6
2420
+ },
2421
+ "model.layers.26.self_attn.k_proj": {
2422
+ "group_size": 32,
2423
+ "bits": 6
2424
+ },
2425
+ "model.layers.26.self_attn.v_proj": {
2426
+ "group_size": 32,
2427
+ "bits": 8
2428
+ },
2429
+ "model.layers.26.self_attn.o_proj": {
2430
+ "group_size": 32,
2431
+ "bits": 6
2432
+ },
2433
+ "model.layers.26.mlp.gate": {
2434
+ "group_size": 32,
2435
+ "bits": 6
2436
+ },
2437
+ "model.layers.26.mlp.switch_mlp.gate_proj": {
2438
+ "group_size": 32,
2439
+ "bits": 6
2440
+ },
2441
+ "model.layers.26.mlp.switch_mlp.up_proj": {
2442
+ "group_size": 32,
2443
+ "bits": 6
2444
+ },
2445
+ "model.layers.26.mlp.switch_mlp.down_proj": {
2446
+ "group_size": 32,
2447
+ "bits": 8
2448
+ },
2449
+ "model.layers.27.self_attn.q_proj": {
2450
+ "group_size": 32,
2451
+ "bits": 6
2452
+ },
2453
+ "model.layers.27.self_attn.k_proj": {
2454
+ "group_size": 32,
2455
+ "bits": 6
2456
+ },
2457
+ "model.layers.27.self_attn.v_proj": {
2458
+ "group_size": 32,
2459
+ "bits": 8
2460
+ },
2461
+ "model.layers.27.self_attn.o_proj": {
2462
+ "group_size": 32,
2463
+ "bits": 6
2464
+ },
2465
+ "model.layers.27.mlp.gate": {
2466
+ "group_size": 32,
2467
+ "bits": 6
2468
+ },
2469
+ "model.layers.27.mlp.switch_mlp.gate_proj": {
2470
+ "group_size": 32,
2471
+ "bits": 6
2472
+ },
2473
+ "model.layers.27.mlp.switch_mlp.up_proj": {
2474
+ "group_size": 32,
2475
+ "bits": 6
2476
+ },
2477
+ "model.layers.27.mlp.switch_mlp.down_proj": {
2478
+ "group_size": 32,
2479
+ "bits": 6
2480
+ },
2481
+ "model.layers.28.self_attn.q_proj": {
2482
+ "group_size": 32,
2483
+ "bits": 6
2484
+ },
2485
+ "model.layers.28.self_attn.k_proj": {
2486
+ "group_size": 32,
2487
+ "bits": 6
2488
+ },
2489
+ "model.layers.28.self_attn.v_proj": {
2490
+ "group_size": 32,
2491
+ "bits": 8
2492
+ },
2493
+ "model.layers.28.self_attn.o_proj": {
2494
+ "group_size": 32,
2495
+ "bits": 6
2496
+ },
2497
+ "model.layers.28.mlp.gate": {
2498
+ "group_size": 32,
2499
+ "bits": 6
2500
+ },
2501
+ "model.layers.28.mlp.switch_mlp.gate_proj": {
2502
+ "group_size": 32,
2503
+ "bits": 6
2504
+ },
2505
+ "model.layers.28.mlp.switch_mlp.up_proj": {
2506
+ "group_size": 32,
2507
+ "bits": 6
2508
+ },
2509
+ "model.layers.28.mlp.switch_mlp.down_proj": {
2510
+ "group_size": 32,
2511
+ "bits": 6
2512
+ },
2513
+ "model.layers.29.self_attn.q_proj": {
2514
+ "group_size": 32,
2515
+ "bits": 6
2516
+ },
2517
+ "model.layers.29.self_attn.k_proj": {
2518
+ "group_size": 32,
2519
+ "bits": 6
2520
+ },
2521
+ "model.layers.29.self_attn.v_proj": {
2522
+ "group_size": 32,
2523
+ "bits": 8
2524
+ },
2525
+ "model.layers.29.self_attn.o_proj": {
2526
+ "group_size": 32,
2527
+ "bits": 6
2528
+ },
2529
+ "model.layers.29.mlp.gate": {
2530
+ "group_size": 32,
2531
+ "bits": 6
2532
+ },
2533
+ "model.layers.29.mlp.switch_mlp.gate_proj": {
2534
+ "group_size": 32,
2535
+ "bits": 6
2536
+ },
2537
+ "model.layers.29.mlp.switch_mlp.up_proj": {
2538
+ "group_size": 32,
2539
+ "bits": 6
2540
+ },
2541
+ "model.layers.29.mlp.switch_mlp.down_proj": {
2542
+ "group_size": 32,
2543
+ "bits": 8
2544
+ },
2545
+ "model.layers.30.self_attn.q_proj": {
2546
+ "group_size": 32,
2547
+ "bits": 6
2548
+ },
2549
+ "model.layers.30.self_attn.k_proj": {
2550
+ "group_size": 32,
2551
+ "bits": 6
2552
+ },
2553
+ "model.layers.30.self_attn.v_proj": {
2554
+ "group_size": 32,
2555
+ "bits": 8
2556
+ },
2557
+ "model.layers.30.self_attn.o_proj": {
2558
+ "group_size": 32,
2559
+ "bits": 6
2560
+ },
2561
+ "model.layers.30.mlp.gate": {
2562
+ "group_size": 32,
2563
+ "bits": 6
2564
+ },
2565
+ "model.layers.30.mlp.switch_mlp.gate_proj": {
2566
+ "group_size": 32,
2567
+ "bits": 6
2568
+ },
2569
+ "model.layers.30.mlp.switch_mlp.up_proj": {
2570
+ "group_size": 32,
2571
+ "bits": 6
2572
+ },
2573
+ "model.layers.30.mlp.switch_mlp.down_proj": {
2574
+ "group_size": 32,
2575
+ "bits": 6
2576
+ },
2577
+ "model.layers.31.self_attn.q_proj": {
2578
+ "group_size": 32,
2579
+ "bits": 6
2580
+ },
2581
+ "model.layers.31.self_attn.k_proj": {
2582
+ "group_size": 32,
2583
+ "bits": 6
2584
+ },
2585
+ "model.layers.31.self_attn.v_proj": {
2586
+ "group_size": 32,
2587
+ "bits": 8
2588
+ },
2589
+ "model.layers.31.self_attn.o_proj": {
2590
+ "group_size": 32,
2591
+ "bits": 6
2592
+ },
2593
+ "model.layers.31.mlp.gate": {
2594
+ "group_size": 32,
2595
+ "bits": 6
2596
+ },
2597
+ "model.layers.31.mlp.switch_mlp.gate_proj": {
2598
+ "group_size": 32,
2599
+ "bits": 6
2600
+ },
2601
+ "model.layers.31.mlp.switch_mlp.up_proj": {
2602
+ "group_size": 32,
2603
+ "bits": 6
2604
+ },
2605
+ "model.layers.31.mlp.switch_mlp.down_proj": {
2606
+ "group_size": 32,
2607
+ "bits": 6
2608
+ },
2609
+ "model.layers.32.self_attn.q_proj": {
2610
+ "group_size": 32,
2611
+ "bits": 6
2612
+ },
2613
+ "model.layers.32.self_attn.k_proj": {
2614
+ "group_size": 32,
2615
+ "bits": 6
2616
+ },
2617
+ "model.layers.32.self_attn.v_proj": {
2618
+ "group_size": 32,
2619
+ "bits": 8
2620
+ },
2621
+ "model.layers.32.self_attn.o_proj": {
2622
+ "group_size": 32,
2623
+ "bits": 6
2624
+ },
2625
+ "model.layers.32.mlp.gate": {
2626
+ "group_size": 32,
2627
+ "bits": 6
2628
+ },
2629
+ "model.layers.32.mlp.switch_mlp.gate_proj": {
2630
+ "group_size": 32,
2631
+ "bits": 6
2632
+ },
2633
+ "model.layers.32.mlp.switch_mlp.up_proj": {
2634
+ "group_size": 32,
2635
+ "bits": 6
2636
+ },
2637
+ "model.layers.32.mlp.switch_mlp.down_proj": {
2638
+ "group_size": 32,
2639
+ "bits": 8
2640
+ },
2641
+ "model.layers.33.self_attn.q_proj": {
2642
+ "group_size": 32,
2643
+ "bits": 6
2644
+ },
2645
+ "model.layers.33.self_attn.k_proj": {
2646
+ "group_size": 32,
2647
+ "bits": 6
2648
+ },
2649
+ "model.layers.33.self_attn.v_proj": {
2650
+ "group_size": 32,
2651
+ "bits": 8
2652
+ },
2653
+ "model.layers.33.self_attn.o_proj": {
2654
+ "group_size": 32,
2655
+ "bits": 6
2656
+ },
2657
+ "model.layers.33.mlp.gate": {
2658
+ "group_size": 32,
2659
+ "bits": 6
2660
+ },
2661
+ "model.layers.33.mlp.switch_mlp.gate_proj": {
2662
+ "group_size": 32,
2663
+ "bits": 6
2664
+ },
2665
+ "model.layers.33.mlp.switch_mlp.up_proj": {
2666
+ "group_size": 32,
2667
+ "bits": 6
2668
+ },
2669
+ "model.layers.33.mlp.switch_mlp.down_proj": {
2670
+ "group_size": 32,
2671
+ "bits": 6
2672
+ },
2673
+ "model.layers.34.self_attn.q_proj": {
2674
+ "group_size": 32,
2675
+ "bits": 6
2676
+ },
2677
+ "model.layers.34.self_attn.k_proj": {
2678
+ "group_size": 32,
2679
+ "bits": 6
2680
+ },
2681
+ "model.layers.34.self_attn.v_proj": {
2682
+ "group_size": 32,
2683
+ "bits": 8
2684
+ },
2685
+ "model.layers.34.self_attn.o_proj": {
2686
+ "group_size": 32,
2687
+ "bits": 6
2688
+ },
2689
+ "model.layers.34.mlp.gate": {
2690
+ "group_size": 32,
2691
+ "bits": 6
2692
+ },
2693
+ "model.layers.34.mlp.switch_mlp.gate_proj": {
2694
+ "group_size": 32,
2695
+ "bits": 6
2696
+ },
2697
+ "model.layers.34.mlp.switch_mlp.up_proj": {
2698
+ "group_size": 32,
2699
+ "bits": 6
2700
+ },
2701
+ "model.layers.34.mlp.switch_mlp.down_proj": {
2702
+ "group_size": 32,
2703
+ "bits": 6
2704
+ },
2705
+ "model.layers.35.self_attn.q_proj": {
2706
+ "group_size": 32,
2707
+ "bits": 6
2708
+ },
2709
+ "model.layers.35.self_attn.k_proj": {
2710
+ "group_size": 32,
2711
+ "bits": 6
2712
+ },
2713
+ "model.layers.35.self_attn.v_proj": {
2714
+ "group_size": 32,
2715
+ "bits": 8
2716
+ },
2717
+ "model.layers.35.self_attn.o_proj": {
2718
+ "group_size": 32,
2719
+ "bits": 6
2720
+ },
2721
+ "model.layers.35.mlp.gate": {
2722
+ "group_size": 32,
2723
+ "bits": 6
2724
+ },
2725
+ "model.layers.35.mlp.switch_mlp.gate_proj": {
2726
+ "group_size": 32,
2727
+ "bits": 6
2728
+ },
2729
+ "model.layers.35.mlp.switch_mlp.up_proj": {
2730
+ "group_size": 32,
2731
+ "bits": 6
2732
+ },
2733
+ "model.layers.35.mlp.switch_mlp.down_proj": {
2734
+ "group_size": 32,
2735
+ "bits": 8
2736
+ },
2737
+ "model.layers.36.self_attn.q_proj": {
2738
+ "group_size": 32,
2739
+ "bits": 6
2740
+ },
2741
+ "model.layers.36.self_attn.k_proj": {
2742
+ "group_size": 32,
2743
+ "bits": 6
2744
+ },
2745
+ "model.layers.36.self_attn.v_proj": {
2746
+ "group_size": 32,
2747
+ "bits": 8
2748
+ },
2749
+ "model.layers.36.self_attn.o_proj": {
2750
+ "group_size": 32,
2751
+ "bits": 6
2752
+ },
2753
+ "model.layers.36.mlp.gate": {
2754
+ "group_size": 32,
2755
+ "bits": 6
2756
+ },
2757
+ "model.layers.36.mlp.switch_mlp.gate_proj": {
2758
+ "group_size": 32,
2759
+ "bits": 6
2760
+ },
2761
+ "model.layers.36.mlp.switch_mlp.up_proj": {
2762
+ "group_size": 32,
2763
+ "bits": 6
2764
+ },
2765
+ "model.layers.36.mlp.switch_mlp.down_proj": {
2766
+ "group_size": 32,
2767
+ "bits": 6
2768
+ },
2769
+ "model.layers.37.self_attn.q_proj": {
2770
+ "group_size": 32,
2771
+ "bits": 6
2772
+ },
2773
+ "model.layers.37.self_attn.k_proj": {
2774
+ "group_size": 32,
2775
+ "bits": 6
2776
+ },
2777
+ "model.layers.37.self_attn.v_proj": {
2778
+ "group_size": 32,
2779
+ "bits": 8
2780
+ },
2781
+ "model.layers.37.self_attn.o_proj": {
2782
+ "group_size": 32,
2783
+ "bits": 6
2784
+ },
2785
+ "model.layers.37.mlp.gate": {
2786
+ "group_size": 32,
2787
+ "bits": 6
2788
+ },
2789
+ "model.layers.37.mlp.switch_mlp.gate_proj": {
2790
+ "group_size": 32,
2791
+ "bits": 6
2792
+ },
2793
+ "model.layers.37.mlp.switch_mlp.up_proj": {
2794
+ "group_size": 32,
2795
+ "bits": 6
2796
+ },
2797
+ "model.layers.37.mlp.switch_mlp.down_proj": {
2798
+ "group_size": 32,
2799
+ "bits": 6
2800
+ },
2801
+ "model.layers.38.self_attn.q_proj": {
2802
+ "group_size": 32,
2803
+ "bits": 6
2804
+ },
2805
+ "model.layers.38.self_attn.k_proj": {
2806
+ "group_size": 32,
2807
+ "bits": 6
2808
+ },
2809
+ "model.layers.38.self_attn.v_proj": {
2810
+ "group_size": 32,
2811
+ "bits": 8
2812
+ },
2813
+ "model.layers.38.self_attn.o_proj": {
2814
+ "group_size": 32,
2815
+ "bits": 6
2816
+ },
2817
+ "model.layers.38.mlp.gate": {
2818
+ "group_size": 32,
2819
+ "bits": 6
2820
+ },
2821
+ "model.layers.38.mlp.switch_mlp.gate_proj": {
2822
+ "group_size": 32,
2823
+ "bits": 6
2824
+ },
2825
+ "model.layers.38.mlp.switch_mlp.up_proj": {
2826
+ "group_size": 32,
2827
+ "bits": 6
2828
+ },
2829
+ "model.layers.38.mlp.switch_mlp.down_proj": {
2830
+ "group_size": 32,
2831
+ "bits": 8
2832
+ },
2833
+ "model.layers.39.self_attn.q_proj": {
2834
+ "group_size": 32,
2835
+ "bits": 6
2836
+ },
2837
+ "model.layers.39.self_attn.k_proj": {
2838
+ "group_size": 32,
2839
+ "bits": 6
2840
+ },
2841
+ "model.layers.39.self_attn.v_proj": {
2842
+ "group_size": 32,
2843
+ "bits": 8
2844
+ },
2845
+ "model.layers.39.self_attn.o_proj": {
2846
+ "group_size": 32,
2847
+ "bits": 6
2848
+ },
2849
+ "model.layers.39.mlp.gate": {
2850
+ "group_size": 32,
2851
+ "bits": 6
2852
+ },
2853
+ "model.layers.39.mlp.switch_mlp.gate_proj": {
2854
+ "group_size": 32,
2855
+ "bits": 6
2856
+ },
2857
+ "model.layers.39.mlp.switch_mlp.up_proj": {
2858
+ "group_size": 32,
2859
+ "bits": 6
2860
+ },
2861
+ "model.layers.39.mlp.switch_mlp.down_proj": {
2862
+ "group_size": 32,
2863
+ "bits": 6
2864
+ },
2865
+ "model.layers.40.self_attn.q_proj": {
2866
+ "group_size": 32,
2867
+ "bits": 6
2868
+ },
2869
+ "model.layers.40.self_attn.k_proj": {
2870
+ "group_size": 32,
2871
+ "bits": 6
2872
+ },
2873
+ "model.layers.40.self_attn.v_proj": {
2874
+ "group_size": 32,
2875
+ "bits": 8
2876
+ },
2877
+ "model.layers.40.self_attn.o_proj": {
2878
+ "group_size": 32,
2879
+ "bits": 6
2880
+ },
2881
+ "model.layers.40.mlp.gate": {
2882
+ "group_size": 32,
2883
+ "bits": 6
2884
+ },
2885
+ "model.layers.40.mlp.switch_mlp.gate_proj": {
2886
+ "group_size": 32,
2887
+ "bits": 6
2888
+ },
2889
+ "model.layers.40.mlp.switch_mlp.up_proj": {
2890
+ "group_size": 32,
2891
+ "bits": 6
2892
+ },
2893
+ "model.layers.40.mlp.switch_mlp.down_proj": {
2894
+ "group_size": 32,
2895
+ "bits": 6
2896
+ },
2897
+ "model.layers.41.self_attn.q_proj": {
2898
+ "group_size": 32,
2899
+ "bits": 6
2900
+ },
2901
+ "model.layers.41.self_attn.k_proj": {
2902
+ "group_size": 32,
2903
+ "bits": 6
2904
+ },
2905
+ "model.layers.41.self_attn.v_proj": {
2906
+ "group_size": 32,
2907
+ "bits": 8
2908
+ },
2909
+ "model.layers.41.self_attn.o_proj": {
2910
+ "group_size": 32,
2911
+ "bits": 6
2912
+ },
2913
+ "model.layers.41.mlp.gate": {
2914
+ "group_size": 32,
2915
+ "bits": 6
2916
+ },
2917
+ "model.layers.41.mlp.switch_mlp.gate_proj": {
2918
+ "group_size": 32,
2919
+ "bits": 6
2920
+ },
2921
+ "model.layers.41.mlp.switch_mlp.up_proj": {
2922
+ "group_size": 32,
2923
+ "bits": 6
2924
+ },
2925
+ "model.layers.41.mlp.switch_mlp.down_proj": {
2926
+ "group_size": 32,
2927
+ "bits": 8
2928
+ },
2929
+ "model.layers.42.self_attn.q_proj": {
2930
+ "group_size": 32,
2931
+ "bits": 6
2932
+ },
2933
+ "model.layers.42.self_attn.k_proj": {
2934
+ "group_size": 32,
2935
+ "bits": 6
2936
+ },
2937
+ "model.layers.42.self_attn.v_proj": {
2938
+ "group_size": 32,
2939
+ "bits": 8
2940
+ },
2941
+ "model.layers.42.self_attn.o_proj": {
2942
+ "group_size": 32,
2943
+ "bits": 6
2944
+ },
2945
+ "model.layers.42.mlp.gate": {
2946
+ "group_size": 32,
2947
+ "bits": 6
2948
+ },
2949
+ "model.layers.42.mlp.switch_mlp.gate_proj": {
2950
+ "group_size": 32,
2951
+ "bits": 6
2952
+ },
2953
+ "model.layers.42.mlp.switch_mlp.up_proj": {
2954
+ "group_size": 32,
2955
+ "bits": 6
2956
+ },
2957
+ "model.layers.42.mlp.switch_mlp.down_proj": {
2958
+ "group_size": 32,
2959
+ "bits": 8
2960
+ },
2961
+ "model.layers.43.self_attn.q_proj": {
2962
+ "group_size": 32,
2963
+ "bits": 6
2964
+ },
2965
+ "model.layers.43.self_attn.k_proj": {
2966
+ "group_size": 32,
2967
+ "bits": 6
2968
+ },
2969
+ "model.layers.43.self_attn.v_proj": {
2970
+ "group_size": 32,
2971
+ "bits": 8
2972
+ },
2973
+ "model.layers.43.self_attn.o_proj": {
2974
+ "group_size": 32,
2975
+ "bits": 6
2976
+ },
2977
+ "model.layers.43.mlp.gate": {
2978
+ "group_size": 32,
2979
+ "bits": 6
2980
+ },
2981
+ "model.layers.43.mlp.switch_mlp.gate_proj": {
2982
+ "group_size": 32,
2983
+ "bits": 6
2984
+ },
2985
+ "model.layers.43.mlp.switch_mlp.up_proj": {
2986
+ "group_size": 32,
2987
+ "bits": 6
2988
+ },
2989
+ "model.layers.43.mlp.switch_mlp.down_proj": {
2990
+ "group_size": 32,
2991
+ "bits": 8
2992
+ },
2993
+ "model.layers.44.self_attn.q_proj": {
2994
+ "group_size": 32,
2995
+ "bits": 6
2996
+ },
2997
+ "model.layers.44.self_attn.k_proj": {
2998
+ "group_size": 32,
2999
+ "bits": 6
3000
+ },
3001
+ "model.layers.44.self_attn.v_proj": {
3002
+ "group_size": 32,
3003
+ "bits": 8
3004
+ },
3005
+ "model.layers.44.self_attn.o_proj": {
3006
+ "group_size": 32,
3007
+ "bits": 6
3008
+ },
3009
+ "model.layers.44.mlp.gate": {
3010
+ "group_size": 32,
3011
+ "bits": 6
3012
+ },
3013
+ "model.layers.44.mlp.switch_mlp.gate_proj": {
3014
+ "group_size": 32,
3015
+ "bits": 6
3016
+ },
3017
+ "model.layers.44.mlp.switch_mlp.up_proj": {
3018
+ "group_size": 32,
3019
+ "bits": 6
3020
+ },
3021
+ "model.layers.44.mlp.switch_mlp.down_proj": {
3022
+ "group_size": 32,
3023
+ "bits": 8
3024
+ },
3025
+ "model.layers.45.self_attn.q_proj": {
3026
+ "group_size": 32,
3027
+ "bits": 6
3028
+ },
3029
+ "model.layers.45.self_attn.k_proj": {
3030
+ "group_size": 32,
3031
+ "bits": 6
3032
+ },
3033
+ "model.layers.45.self_attn.v_proj": {
3034
+ "group_size": 32,
3035
+ "bits": 8
3036
+ },
3037
+ "model.layers.45.self_attn.o_proj": {
3038
+ "group_size": 32,
3039
+ "bits": 6
3040
+ },
3041
+ "model.layers.45.mlp.gate": {
3042
+ "group_size": 32,
3043
+ "bits": 6
3044
+ },
3045
+ "model.layers.45.mlp.switch_mlp.gate_proj": {
3046
+ "group_size": 32,
3047
+ "bits": 6
3048
+ },
3049
+ "model.layers.45.mlp.switch_mlp.up_proj": {
3050
+ "group_size": 32,
3051
+ "bits": 6
3052
+ },
3053
+ "model.layers.45.mlp.switch_mlp.down_proj": {
3054
+ "group_size": 32,
3055
+ "bits": 8
3056
+ },
3057
+ "model.layers.46.self_attn.q_proj": {
3058
+ "group_size": 32,
3059
+ "bits": 6
3060
+ },
3061
+ "model.layers.46.self_attn.k_proj": {
3062
+ "group_size": 32,
3063
+ "bits": 6
3064
+ },
3065
+ "model.layers.46.self_attn.v_proj": {
3066
+ "group_size": 32,
3067
+ "bits": 8
3068
+ },
3069
+ "model.layers.46.self_attn.o_proj": {
3070
+ "group_size": 32,
3071
+ "bits": 6
3072
+ },
3073
+ "model.layers.46.mlp.gate": {
3074
+ "group_size": 32,
3075
+ "bits": 6
3076
+ },
3077
+ "model.layers.46.mlp.switch_mlp.gate_proj": {
3078
+ "group_size": 32,
3079
+ "bits": 6
3080
+ },
3081
+ "model.layers.46.mlp.switch_mlp.up_proj": {
3082
+ "group_size": 32,
3083
+ "bits": 6
3084
+ },
3085
+ "model.layers.46.mlp.switch_mlp.down_proj": {
3086
+ "group_size": 32,
3087
+ "bits": 8
3088
+ },
3089
+ "model.layers.47.self_attn.q_proj": {
3090
+ "group_size": 32,
3091
+ "bits": 6
3092
+ },
3093
+ "model.layers.47.self_attn.k_proj": {
3094
+ "group_size": 32,
3095
+ "bits": 6
3096
+ },
3097
+ "model.layers.47.self_attn.v_proj": {
3098
+ "group_size": 32,
3099
+ "bits": 8
3100
+ },
3101
+ "model.layers.47.self_attn.o_proj": {
3102
+ "group_size": 32,
3103
+ "bits": 6
3104
+ },
3105
+ "model.layers.47.mlp.gate": {
3106
+ "group_size": 32,
3107
+ "bits": 6
3108
+ },
3109
+ "model.layers.47.mlp.switch_mlp.gate_proj": {
3110
+ "group_size": 32,
3111
+ "bits": 6
3112
+ },
3113
+ "model.layers.47.mlp.switch_mlp.up_proj": {
3114
+ "group_size": 32,
3115
+ "bits": 6
3116
+ },
3117
+ "model.layers.47.mlp.switch_mlp.down_proj": {
3118
+ "group_size": 32,
3119
+ "bits": 8
3120
+ },
3121
+ "lm_head": {
3122
+ "group_size": 32,
3123
+ "bits": 8
3124
+ }
3125
+ },
3126
+ "rms_norm_eps": 1e-06,
3127
+ "rope_scaling": null,
3128
+ "rope_theta": 10000000,
3129
+ "router_aux_loss_coef": 0.0,
3130
+ "shared_expert_intermediate_size": 0,
3131
+ "sliding_window": null,
3132
+ "tie_word_embeddings": false,
3133
+ "torch_dtype": "bfloat16",
3134
+ "transformers_version": "4.54.0",
3135
+ "use_cache": true,
3136
+ "use_qk_norm": true,
3137
+ "use_sliding_window": false,
3138
+ "vocab_size": 151936
3139
+ }
generation_config.json ADDED
@@ -0,0 +1,13 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "do_sample": true,
3
+ "eos_token_id": [
4
+ 151645,
5
+ 151643
6
+ ],
7
+ "pad_token_id": 151643,
8
+ "repetition_penalty": 1.05,
9
+ "temperature": 0.7,
10
+ "top_k": 20,
11
+ "top_p": 0.8,
12
+ "transformers_version": "4.54.0"
13
+ }
merges.txt ADDED
The diff for this file is too large to render. See raw diff
 
model-00001-of-00006.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:154d45a95b9b49efa1752c15c54b32d85b951f6f0e3f5b380d2c06ddff00bf23
3
+ size 5236630306
model-00002-of-00006.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a38eb83b4f2db6b3094f413b347d9a32d7780ac24cca065915c8574d236808ae
3
+ size 5236664153
model-00003-of-00006.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f01c1f97e8644b558db1bebaf5e8c01eeabb576942053355431ed1c39a359d24
3
+ size 5304013258
model-00004-of-00006.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b13e4870b87024626d01aaaae1ded777fa71422948111b15995672dead2baf56
3
+ size 5236664171
model-00005-of-00006.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c8a3898f57d6ed5685b2cf66ee893734240dd040ca953a34999d051111999fc1
3
+ size 5337327481
model-00006-of-00006.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b0966f594223f27a3a693e2242c7d5abd3818af20f36226b81fea93d47db5785
3
+ size 1768222848
model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
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": "<|endoftext|>",
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 +1,239 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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": 1048576,
235
+ "pad_token": "<|endoftext|>",
236
+ "split_special_tokens": false,
237
+ "tokenizer_class": "Qwen2Tokenizer",
238
+ "unk_token": null
239
+ }
vocab.json ADDED
The diff for this file is too large to render. See raw diff