Hugging Face's logo

nicolasembleton
/
Apodex-1.1-mini-MLX-5bit

Text Generation
MLX
Safetensors
qwen3_5_moe
Mixture of Experts
5-bit
conversational
Model card Files
xet
 Community
nicolasembleton commited on
Commit
b99c85c
·
verified ·
1 Parent(s): 15d368e

Upload folder using huggingface_hub

Browse files
Files changed (13) hide show
  1. .gitattributes +1 -0
  2. README.md +31 -0
  3. chat_template.jinja +173 -0
  4. config.json +753 -0
  5. generation_config.json +13 -0
  6. model-00001-of-00005.safetensors +3 -0
  7. model-00002-of-00005.safetensors +3 -0
  8. model-00003-of-00005.safetensors +3 -0
  9. model-00004-of-00005.safetensors +3 -0
  10. model-00005-of-00005.safetensors +3 -0
  11. model.safetensors.index.json +0 -0
  12. tokenizer.json +3 -0
  13. tokenizer_config.json +33 -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 +1,31 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # Apodex-1.1-mini MLX (5-bit)
2
+
3
+ Converted from [apodex/Apodex-1.1-mini](https://huggingface.co/apodex/Apodex-1.1-mini) (Qwen3.5-35B-A3B MoE, 36B params, 3B active).
4
+
5
+ | Property | Value |
6
+ |----------|-------|
7
+ | Quantization | 5-bit affine |
8
+ | Group size | 64 |
9
+ | ~Size | ~23 GB |
10
+
11
+ ## Usage
12
+
13
+ ```bash
14
+ mlx_lm.generate --model ./Apodex-1.1-mini-MLX-5bit --prompt "Hello"
15
+ ```
16
+
17
+ ## Architecture
18
+
19
+ - **Base:** Qwen3.5-35B-A3B (MoE)
20
+ - **Experts:** 256 total, 8 active per token
21
+ - **Layers:** 40 (hybrid linear/full attention)
22
+ - **Context:** 262144 max position embeddings
23
+ - **Vision tower:** Stripped during conversion (text-only)
24
+
25
+ ## Known Issues
26
+
27
+ - Metal OOM during long-context prefill on <48GB unified memory. Keep context under 32k tokens.
28
+
29
+ ## License
30
+
31
+ Apache 2.0
chat_template.jinja ADDED
@@ -0,0 +1,173 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {%- set apodex_identity = '# Role Definition\n- Name: Apodex\n- Developer: Apodex AI\n- Identity: An AI assistant developed by Apodex AI\n- Core Principle: Always respond as Apodex and never pretend or claim to be any other AI model\n- Please interact with the user strictly in the identity of Apodex.' %}
2
+ {%- set identity_mode = identity_mode if identity_mode is defined else 'fallback' %}
3
+ {%- set has_caller_system = (messages and messages[0].role == 'system') %}
4
+ {%- set inject_identity = (identity_mode == 'force') or (identity_mode == 'fallback' and not has_caller_system) %}
5
+ {%- set image_count = namespace(value=0) %}
6
+ {%- set video_count = namespace(value=0) %}
7
+ {%- macro render_content(content, do_vision_count, is_system_content=false) %}
8
+ {%- if content is string %}
9
+ {{- content }}
10
+ {%- elif content is iterable and content is not mapping %}
11
+ {%- for item in content %}
12
+ {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
13
+ {%- if is_system_content %}
14
+ {{- raise_exception('System message cannot contain images.') }}
15
+ {%- endif %}
16
+ {%- if do_vision_count %}
17
+ {%- set image_count.value = image_count.value + 1 %}
18
+ {%- endif %}
19
+ {%- if add_vision_id %}
20
+ {{- 'Picture ' ~ image_count.value ~ ': ' }}
21
+ {%- endif %}
22
+ {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
23
+ {%- elif 'video' in item or item.type == 'video' %}
24
+ {%- if is_system_content %}
25
+ {{- raise_exception('System message cannot contain videos.') }}
26
+ {%- endif %}
27
+ {%- if do_vision_count %}
28
+ {%- set video_count.value = video_count.value + 1 %}
29
+ {%- endif %}
30
+ {%- if add_vision_id %}
31
+ {{- 'Video ' ~ video_count.value ~ ': ' }}
32
+ {%- endif %}
33
+ {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
34
+ {%- elif 'text' in item %}
35
+ {{- item.text }}
36
+ {%- else %}
37
+ {{- raise_exception('Unexpected item type in content.') }}
38
+ {%- endif %}
39
+ {%- endfor %}
40
+ {%- elif content is none or content is undefined %}
41
+ {{- '' }}
42
+ {%- else %}
43
+ {{- raise_exception('Unexpected content type.') }}
44
+ {%- endif %}
45
+ {%- endmacro %}
46
+ {%- if not messages %}
47
+ {{- raise_exception('No messages provided.') }}
48
+ {%- endif %}
49
+ {%- if tools and tools is iterable and tools is not mapping %}
50
+ {{- '<|im_start|>system\n' }}
51
+ {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
52
+ {%- for tool in tools %}
53
+ {{- "\n" }}
54
+ {{- tool | tojson }}
55
+ {%- endfor %}
56
+ {{- "\n</tools>" }}
57
+ {{- '\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>' }}
58
+ {%- if inject_identity %}
59
+ {{- '\n\n' + apodex_identity }}
60
+ {%- endif %}
61
+ {%- if has_caller_system %}
62
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
63
+ {%- if content %}
64
+ {{- '\n\n' + content }}
65
+ {%- endif %}
66
+ {%- endif %}
67
+ {{- '<|im_end|>\n' }}
68
+ {%- else %}
69
+ {%- if inject_identity or has_caller_system %}
70
+ {{- '<|im_start|>system\n' }}
71
+ {%- set parts = namespace(first=true) %}
72
+ {%- if inject_identity %}
73
+ {{- apodex_identity }}
74
+ {%- set parts.first = false %}
75
+ {%- endif %}
76
+ {%- if has_caller_system %}
77
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
78
+ {%- if content %}
79
+ {%- if not parts.first %}{{- '\n\n' }}{%- endif %}
80
+ {{- content }}
81
+ {%- endif %}
82
+ {%- endif %}
83
+ {{- '<|im_end|>\n' }}
84
+ {%- endif %}
85
+ {%- endif %}
86
+ {%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
87
+ {%- for message in messages[::-1] %}
88
+ {%- set index = (messages|length - 1) - loop.index0 %}
89
+ {%- if ns.multi_step_tool and message.role == "user" %}
90
+ {%- set content = render_content(message.content, false)|trim %}
91
+ {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
92
+ {%- set ns.multi_step_tool = false %}
93
+ {%- set ns.last_query_index = index %}
94
+ {%- endif %}
95
+ {%- endif %}
96
+ {%- endfor %}
97
+ {%- if ns.multi_step_tool %}
98
+ {{- raise_exception('No user query found in messages.') }}
99
+ {%- endif %}
100
+ {%- for message in messages %}
101
+ {%- set content = render_content(message.content, true)|trim %}
102
+ {%- if message.role == "system" %}
103
+ {%- if not loop.first %}
104
+ {{- raise_exception('System message must be at the beginning.') }}
105
+ {%- endif %}
106
+ {%- elif message.role == "user" %}
107
+ {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
108
+ {%- elif message.role == "assistant" %}
109
+ {%- set reasoning_content = '' %}
110
+ {%- if message.reasoning_content is string %}
111
+ {%- set reasoning_content = message.reasoning_content %}
112
+ {%- else %}
113
+ {%- if '</think>' in content %}
114
+ {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
115
+ {%- set content = content.split('</think>')[-1].lstrip('\n') %}
116
+ {%- endif %}
117
+ {%- endif %}
118
+ {%- set reasoning_content = reasoning_content|trim %}
119
+ {%- if (preserve_thinking is not defined or preserve_thinking is true) or (loop.index0 > ns.last_query_index) %}
120
+ {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
121
+ {%- else %}
122
+ {{- '<|im_start|>' + message.role + '\n' + content }}
123
+ {%- endif %}
124
+ {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
125
+ {%- for tool_call in message.tool_calls %}
126
+ {%- if tool_call.function is defined %}
127
+ {%- set tool_call = tool_call.function %}
128
+ {%- endif %}
129
+ {%- if loop.first %}
130
+ {%- if content|trim %}
131
+ {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
132
+ {%- else %}
133
+ {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
134
+ {%- endif %}
135
+ {%- else %}
136
+ {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
137
+ {%- endif %}
138
+ {%- if tool_call.arguments is defined %}
139
+ {%- for args_name, args_value in tool_call.arguments|items %}
140
+ {{- '<parameter=' + args_name + '>\n' }}
141
+ {%- set args_value = args_value | string if args_value is string else args_value | tojson | safe %}
142
+ {{- args_value }}
143
+ {{- '\n</parameter>\n' }}
144
+ {%- endfor %}
145
+ {%- endif %}
146
+ {{- '</function>\n</tool_call>' }}
147
+ {%- endfor %}
148
+ {%- endif %}
149
+ {{- '<|im_end|>\n' }}
150
+ {%- elif message.role == "tool" %}
151
+ {%- if loop.previtem and loop.previtem.role != "tool" %}
152
+ {{- '<|im_start|>user' }}
153
+ {%- endif %}
154
+ {{- '\n<tool_response>\n' }}
155
+ {{- content }}
156
+ {{- '\n</tool_response>' }}
157
+ {%- if not loop.last and loop.nextitem.role != "tool" %}
158
+ {{- '<|im_end|>\n' }}
159
+ {%- elif loop.last %}
160
+ {{- '<|im_end|>\n' }}
161
+ {%- endif %}
162
+ {%- else %}
163
+ {{- raise_exception('Unexpected message role.') }}
164
+ {%- endif %}
165
+ {%- endfor %}
166
+ {%- if add_generation_prompt %}
167
+ {{- '<|im_start|>assistant\n' }}
168
+ {%- if enable_thinking is defined and enable_thinking is false %}
169
+ {{- '<think>\n\n</think>\n\n' }}
170
+ {%- else %}
171
+ {{- '<think>\n' }}
172
+ {%- endif %}
173
+ {%- endif %}
config.json ADDED
@@ -0,0 +1,753 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "Qwen3_5MoeForConditionalGeneration"
4
+ ],
5
+ "eos_token_id": [
6
+ 248046,
7
+ 248044
8
+ ],
9
+ "image_token_id": 248056,
10
+ "model_type": "qwen3_5_moe",
11
+ "quantization": {
12
+ "group_size": 64,
13
+ "bits": 5,
14
+ "mode": "affine",
15
+ "language_model.model.layers.0.mlp.gate": {
16
+ "group_size": 64,
17
+ "bits": 8
18
+ },
19
+ "language_model.model.layers.0.mlp.shared_expert_gate": {
20
+ "group_size": 64,
21
+ "bits": 8
22
+ },
23
+ "language_model.model.layers.1.mlp.gate": {
24
+ "group_size": 64,
25
+ "bits": 8
26
+ },
27
+ "language_model.model.layers.1.mlp.shared_expert_gate": {
28
+ "group_size": 64,
29
+ "bits": 8
30
+ },
31
+ "language_model.model.layers.2.mlp.gate": {
32
+ "group_size": 64,
33
+ "bits": 8
34
+ },
35
+ "language_model.model.layers.2.mlp.shared_expert_gate": {
36
+ "group_size": 64,
37
+ "bits": 8
38
+ },
39
+ "language_model.model.layers.3.mlp.gate": {
40
+ "group_size": 64,
41
+ "bits": 8
42
+ },
43
+ "language_model.model.layers.3.mlp.shared_expert_gate": {
44
+ "group_size": 64,
45
+ "bits": 8
46
+ },
47
+ "language_model.model.layers.4.mlp.gate": {
48
+ "group_size": 64,
49
+ "bits": 8
50
+ },
51
+ "language_model.model.layers.4.mlp.shared_expert_gate": {
52
+ "group_size": 64,
53
+ "bits": 8
54
+ },
55
+ "language_model.model.layers.5.mlp.gate": {
56
+ "group_size": 64,
57
+ "bits": 8
58
+ },
59
+ "language_model.model.layers.5.mlp.shared_expert_gate": {
60
+ "group_size": 64,
61
+ "bits": 8
62
+ },
63
+ "language_model.model.layers.6.mlp.gate": {
64
+ "group_size": 64,
65
+ "bits": 8
66
+ },
67
+ "language_model.model.layers.6.mlp.shared_expert_gate": {
68
+ "group_size": 64,
69
+ "bits": 8
70
+ },
71
+ "language_model.model.layers.7.mlp.gate": {
72
+ "group_size": 64,
73
+ "bits": 8
74
+ },
75
+ "language_model.model.layers.7.mlp.shared_expert_gate": {
76
+ "group_size": 64,
77
+ "bits": 8
78
+ },
79
+ "language_model.model.layers.8.mlp.gate": {
80
+ "group_size": 64,
81
+ "bits": 8
82
+ },
83
+ "language_model.model.layers.8.mlp.shared_expert_gate": {
84
+ "group_size": 64,
85
+ "bits": 8
86
+ },
87
+ "language_model.model.layers.9.mlp.gate": {
88
+ "group_size": 64,
89
+ "bits": 8
90
+ },
91
+ "language_model.model.layers.9.mlp.shared_expert_gate": {
92
+ "group_size": 64,
93
+ "bits": 8
94
+ },
95
+ "language_model.model.layers.10.mlp.gate": {
96
+ "group_size": 64,
97
+ "bits": 8
98
+ },
99
+ "language_model.model.layers.10.mlp.shared_expert_gate": {
100
+ "group_size": 64,
101
+ "bits": 8
102
+ },
103
+ "language_model.model.layers.11.mlp.gate": {
104
+ "group_size": 64,
105
+ "bits": 8
106
+ },
107
+ "language_model.model.layers.11.mlp.shared_expert_gate": {
108
+ "group_size": 64,
109
+ "bits": 8
110
+ },
111
+ "language_model.model.layers.12.mlp.gate": {
112
+ "group_size": 64,
113
+ "bits": 8
114
+ },
115
+ "language_model.model.layers.12.mlp.shared_expert_gate": {
116
+ "group_size": 64,
117
+ "bits": 8
118
+ },
119
+ "language_model.model.layers.13.mlp.gate": {
120
+ "group_size": 64,
121
+ "bits": 8
122
+ },
123
+ "language_model.model.layers.13.mlp.shared_expert_gate": {
124
+ "group_size": 64,
125
+ "bits": 8
126
+ },
127
+ "language_model.model.layers.14.mlp.gate": {
128
+ "group_size": 64,
129
+ "bits": 8
130
+ },
131
+ "language_model.model.layers.14.mlp.shared_expert_gate": {
132
+ "group_size": 64,
133
+ "bits": 8
134
+ },
135
+ "language_model.model.layers.15.mlp.gate": {
136
+ "group_size": 64,
137
+ "bits": 8
138
+ },
139
+ "language_model.model.layers.15.mlp.shared_expert_gate": {
140
+ "group_size": 64,
141
+ "bits": 8
142
+ },
143
+ "language_model.model.layers.16.mlp.gate": {
144
+ "group_size": 64,
145
+ "bits": 8
146
+ },
147
+ "language_model.model.layers.16.mlp.shared_expert_gate": {
148
+ "group_size": 64,
149
+ "bits": 8
150
+ },
151
+ "language_model.model.layers.17.mlp.gate": {
152
+ "group_size": 64,
153
+ "bits": 8
154
+ },
155
+ "language_model.model.layers.17.mlp.shared_expert_gate": {
156
+ "group_size": 64,
157
+ "bits": 8
158
+ },
159
+ "language_model.model.layers.18.mlp.gate": {
160
+ "group_size": 64,
161
+ "bits": 8
162
+ },
163
+ "language_model.model.layers.18.mlp.shared_expert_gate": {
164
+ "group_size": 64,
165
+ "bits": 8
166
+ },
167
+ "language_model.model.layers.19.mlp.gate": {
168
+ "group_size": 64,
169
+ "bits": 8
170
+ },
171
+ "language_model.model.layers.19.mlp.shared_expert_gate": {
172
+ "group_size": 64,
173
+ "bits": 8
174
+ },
175
+ "language_model.model.layers.20.mlp.gate": {
176
+ "group_size": 64,
177
+ "bits": 8
178
+ },
179
+ "language_model.model.layers.20.mlp.shared_expert_gate": {
180
+ "group_size": 64,
181
+ "bits": 8
182
+ },
183
+ "language_model.model.layers.21.mlp.gate": {
184
+ "group_size": 64,
185
+ "bits": 8
186
+ },
187
+ "language_model.model.layers.21.mlp.shared_expert_gate": {
188
+ "group_size": 64,
189
+ "bits": 8
190
+ },
191
+ "language_model.model.layers.22.mlp.gate": {
192
+ "group_size": 64,
193
+ "bits": 8
194
+ },
195
+ "language_model.model.layers.22.mlp.shared_expert_gate": {
196
+ "group_size": 64,
197
+ "bits": 8
198
+ },
199
+ "language_model.model.layers.23.mlp.gate": {
200
+ "group_size": 64,
201
+ "bits": 8
202
+ },
203
+ "language_model.model.layers.23.mlp.shared_expert_gate": {
204
+ "group_size": 64,
205
+ "bits": 8
206
+ },
207
+ "language_model.model.layers.24.mlp.gate": {
208
+ "group_size": 64,
209
+ "bits": 8
210
+ },
211
+ "language_model.model.layers.24.mlp.shared_expert_gate": {
212
+ "group_size": 64,
213
+ "bits": 8
214
+ },
215
+ "language_model.model.layers.25.mlp.gate": {
216
+ "group_size": 64,
217
+ "bits": 8
218
+ },
219
+ "language_model.model.layers.25.mlp.shared_expert_gate": {
220
+ "group_size": 64,
221
+ "bits": 8
222
+ },
223
+ "language_model.model.layers.26.mlp.gate": {
224
+ "group_size": 64,
225
+ "bits": 8
226
+ },
227
+ "language_model.model.layers.26.mlp.shared_expert_gate": {
228
+ "group_size": 64,
229
+ "bits": 8
230
+ },
231
+ "language_model.model.layers.27.mlp.gate": {
232
+ "group_size": 64,
233
+ "bits": 8
234
+ },
235
+ "language_model.model.layers.27.mlp.shared_expert_gate": {
236
+ "group_size": 64,
237
+ "bits": 8
238
+ },
239
+ "language_model.model.layers.28.mlp.gate": {
240
+ "group_size": 64,
241
+ "bits": 8
242
+ },
243
+ "language_model.model.layers.28.mlp.shared_expert_gate": {
244
+ "group_size": 64,
245
+ "bits": 8
246
+ },
247
+ "language_model.model.layers.29.mlp.gate": {
248
+ "group_size": 64,
249
+ "bits": 8
250
+ },
251
+ "language_model.model.layers.29.mlp.shared_expert_gate": {
252
+ "group_size": 64,
253
+ "bits": 8
254
+ },
255
+ "language_model.model.layers.30.mlp.gate": {
256
+ "group_size": 64,
257
+ "bits": 8
258
+ },
259
+ "language_model.model.layers.30.mlp.shared_expert_gate": {
260
+ "group_size": 64,
261
+ "bits": 8
262
+ },
263
+ "language_model.model.layers.31.mlp.gate": {
264
+ "group_size": 64,
265
+ "bits": 8
266
+ },
267
+ "language_model.model.layers.31.mlp.shared_expert_gate": {
268
+ "group_size": 64,
269
+ "bits": 8
270
+ },
271
+ "language_model.model.layers.32.mlp.gate": {
272
+ "group_size": 64,
273
+ "bits": 8
274
+ },
275
+ "language_model.model.layers.32.mlp.shared_expert_gate": {
276
+ "group_size": 64,
277
+ "bits": 8
278
+ },
279
+ "language_model.model.layers.33.mlp.gate": {
280
+ "group_size": 64,
281
+ "bits": 8
282
+ },
283
+ "language_model.model.layers.33.mlp.shared_expert_gate": {
284
+ "group_size": 64,
285
+ "bits": 8
286
+ },
287
+ "language_model.model.layers.34.mlp.gate": {
288
+ "group_size": 64,
289
+ "bits": 8
290
+ },
291
+ "language_model.model.layers.34.mlp.shared_expert_gate": {
292
+ "group_size": 64,
293
+ "bits": 8
294
+ },
295
+ "language_model.model.layers.35.mlp.gate": {
296
+ "group_size": 64,
297
+ "bits": 8
298
+ },
299
+ "language_model.model.layers.35.mlp.shared_expert_gate": {
300
+ "group_size": 64,
301
+ "bits": 8
302
+ },
303
+ "language_model.model.layers.36.mlp.gate": {
304
+ "group_size": 64,
305
+ "bits": 8
306
+ },
307
+ "language_model.model.layers.36.mlp.shared_expert_gate": {
308
+ "group_size": 64,
309
+ "bits": 8
310
+ },
311
+ "language_model.model.layers.37.mlp.gate": {
312
+ "group_size": 64,
313
+ "bits": 8
314
+ },
315
+ "language_model.model.layers.37.mlp.shared_expert_gate": {
316
+ "group_size": 64,
317
+ "bits": 8
318
+ },
319
+ "language_model.model.layers.38.mlp.gate": {
320
+ "group_size": 64,
321
+ "bits": 8
322
+ },
323
+ "language_model.model.layers.38.mlp.shared_expert_gate": {
324
+ "group_size": 64,
325
+ "bits": 8
326
+ },
327
+ "language_model.model.layers.39.mlp.gate": {
328
+ "group_size": 64,
329
+ "bits": 8
330
+ },
331
+ "language_model.model.layers.39.mlp.shared_expert_gate": {
332
+ "group_size": 64,
333
+ "bits": 8
334
+ }
335
+ },
336
+ "quantization_config": {
337
+ "group_size": 64,
338
+ "bits": 5,
339
+ "mode": "affine",
340
+ "language_model.model.layers.0.mlp.gate": {
341
+ "group_size": 64,
342
+ "bits": 8
343
+ },
344
+ "language_model.model.layers.0.mlp.shared_expert_gate": {
345
+ "group_size": 64,
346
+ "bits": 8
347
+ },
348
+ "language_model.model.layers.1.mlp.gate": {
349
+ "group_size": 64,
350
+ "bits": 8
351
+ },
352
+ "language_model.model.layers.1.mlp.shared_expert_gate": {
353
+ "group_size": 64,
354
+ "bits": 8
355
+ },
356
+ "language_model.model.layers.2.mlp.gate": {
357
+ "group_size": 64,
358
+ "bits": 8
359
+ },
360
+ "language_model.model.layers.2.mlp.shared_expert_gate": {
361
+ "group_size": 64,
362
+ "bits": 8
363
+ },
364
+ "language_model.model.layers.3.mlp.gate": {
365
+ "group_size": 64,
366
+ "bits": 8
367
+ },
368
+ "language_model.model.layers.3.mlp.shared_expert_gate": {
369
+ "group_size": 64,
370
+ "bits": 8
371
+ },
372
+ "language_model.model.layers.4.mlp.gate": {
373
+ "group_size": 64,
374
+ "bits": 8
375
+ },
376
+ "language_model.model.layers.4.mlp.shared_expert_gate": {
377
+ "group_size": 64,
378
+ "bits": 8
379
+ },
380
+ "language_model.model.layers.5.mlp.gate": {
381
+ "group_size": 64,
382
+ "bits": 8
383
+ },
384
+ "language_model.model.layers.5.mlp.shared_expert_gate": {
385
+ "group_size": 64,
386
+ "bits": 8
387
+ },
388
+ "language_model.model.layers.6.mlp.gate": {
389
+ "group_size": 64,
390
+ "bits": 8
391
+ },
392
+ "language_model.model.layers.6.mlp.shared_expert_gate": {
393
+ "group_size": 64,
394
+ "bits": 8
395
+ },
396
+ "language_model.model.layers.7.mlp.gate": {
397
+ "group_size": 64,
398
+ "bits": 8
399
+ },
400
+ "language_model.model.layers.7.mlp.shared_expert_gate": {
401
+ "group_size": 64,
402
+ "bits": 8
403
+ },
404
+ "language_model.model.layers.8.mlp.gate": {
405
+ "group_size": 64,
406
+ "bits": 8
407
+ },
408
+ "language_model.model.layers.8.mlp.shared_expert_gate": {
409
+ "group_size": 64,
410
+ "bits": 8
411
+ },
412
+ "language_model.model.layers.9.mlp.gate": {
413
+ "group_size": 64,
414
+ "bits": 8
415
+ },
416
+ "language_model.model.layers.9.mlp.shared_expert_gate": {
417
+ "group_size": 64,
418
+ "bits": 8
419
+ },
420
+ "language_model.model.layers.10.mlp.gate": {
421
+ "group_size": 64,
422
+ "bits": 8
423
+ },
424
+ "language_model.model.layers.10.mlp.shared_expert_gate": {
425
+ "group_size": 64,
426
+ "bits": 8
427
+ },
428
+ "language_model.model.layers.11.mlp.gate": {
429
+ "group_size": 64,
430
+ "bits": 8
431
+ },
432
+ "language_model.model.layers.11.mlp.shared_expert_gate": {
433
+ "group_size": 64,
434
+ "bits": 8
435
+ },
436
+ "language_model.model.layers.12.mlp.gate": {
437
+ "group_size": 64,
438
+ "bits": 8
439
+ },
440
+ "language_model.model.layers.12.mlp.shared_expert_gate": {
441
+ "group_size": 64,
442
+ "bits": 8
443
+ },
444
+ "language_model.model.layers.13.mlp.gate": {
445
+ "group_size": 64,
446
+ "bits": 8
447
+ },
448
+ "language_model.model.layers.13.mlp.shared_expert_gate": {
449
+ "group_size": 64,
450
+ "bits": 8
451
+ },
452
+ "language_model.model.layers.14.mlp.gate": {
453
+ "group_size": 64,
454
+ "bits": 8
455
+ },
456
+ "language_model.model.layers.14.mlp.shared_expert_gate": {
457
+ "group_size": 64,
458
+ "bits": 8
459
+ },
460
+ "language_model.model.layers.15.mlp.gate": {
461
+ "group_size": 64,
462
+ "bits": 8
463
+ },
464
+ "language_model.model.layers.15.mlp.shared_expert_gate": {
465
+ "group_size": 64,
466
+ "bits": 8
467
+ },
468
+ "language_model.model.layers.16.mlp.gate": {
469
+ "group_size": 64,
470
+ "bits": 8
471
+ },
472
+ "language_model.model.layers.16.mlp.shared_expert_gate": {
473
+ "group_size": 64,
474
+ "bits": 8
475
+ },
476
+ "language_model.model.layers.17.mlp.gate": {
477
+ "group_size": 64,
478
+ "bits": 8
479
+ },
480
+ "language_model.model.layers.17.mlp.shared_expert_gate": {
481
+ "group_size": 64,
482
+ "bits": 8
483
+ },
484
+ "language_model.model.layers.18.mlp.gate": {
485
+ "group_size": 64,
486
+ "bits": 8
487
+ },
488
+ "language_model.model.layers.18.mlp.shared_expert_gate": {
489
+ "group_size": 64,
490
+ "bits": 8
491
+ },
492
+ "language_model.model.layers.19.mlp.gate": {
493
+ "group_size": 64,
494
+ "bits": 8
495
+ },
496
+ "language_model.model.layers.19.mlp.shared_expert_gate": {
497
+ "group_size": 64,
498
+ "bits": 8
499
+ },
500
+ "language_model.model.layers.20.mlp.gate": {
501
+ "group_size": 64,
502
+ "bits": 8
503
+ },
504
+ "language_model.model.layers.20.mlp.shared_expert_gate": {
505
+ "group_size": 64,
506
+ "bits": 8
507
+ },
508
+ "language_model.model.layers.21.mlp.gate": {
509
+ "group_size": 64,
510
+ "bits": 8
511
+ },
512
+ "language_model.model.layers.21.mlp.shared_expert_gate": {
513
+ "group_size": 64,
514
+ "bits": 8
515
+ },
516
+ "language_model.model.layers.22.mlp.gate": {
517
+ "group_size": 64,
518
+ "bits": 8
519
+ },
520
+ "language_model.model.layers.22.mlp.shared_expert_gate": {
521
+ "group_size": 64,
522
+ "bits": 8
523
+ },
524
+ "language_model.model.layers.23.mlp.gate": {
525
+ "group_size": 64,
526
+ "bits": 8
527
+ },
528
+ "language_model.model.layers.23.mlp.shared_expert_gate": {
529
+ "group_size": 64,
530
+ "bits": 8
531
+ },
532
+ "language_model.model.layers.24.mlp.gate": {
533
+ "group_size": 64,
534
+ "bits": 8
535
+ },
536
+ "language_model.model.layers.24.mlp.shared_expert_gate": {
537
+ "group_size": 64,
538
+ "bits": 8
539
+ },
540
+ "language_model.model.layers.25.mlp.gate": {
541
+ "group_size": 64,
542
+ "bits": 8
543
+ },
544
+ "language_model.model.layers.25.mlp.shared_expert_gate": {
545
+ "group_size": 64,
546
+ "bits": 8
547
+ },
548
+ "language_model.model.layers.26.mlp.gate": {
549
+ "group_size": 64,
550
+ "bits": 8
551
+ },
552
+ "language_model.model.layers.26.mlp.shared_expert_gate": {
553
+ "group_size": 64,
554
+ "bits": 8
555
+ },
556
+ "language_model.model.layers.27.mlp.gate": {
557
+ "group_size": 64,
558
+ "bits": 8
559
+ },
560
+ "language_model.model.layers.27.mlp.shared_expert_gate": {
561
+ "group_size": 64,
562
+ "bits": 8
563
+ },
564
+ "language_model.model.layers.28.mlp.gate": {
565
+ "group_size": 64,
566
+ "bits": 8
567
+ },
568
+ "language_model.model.layers.28.mlp.shared_expert_gate": {
569
+ "group_size": 64,
570
+ "bits": 8
571
+ },
572
+ "language_model.model.layers.29.mlp.gate": {
573
+ "group_size": 64,
574
+ "bits": 8
575
+ },
576
+ "language_model.model.layers.29.mlp.shared_expert_gate": {
577
+ "group_size": 64,
578
+ "bits": 8
579
+ },
580
+ "language_model.model.layers.30.mlp.gate": {
581
+ "group_size": 64,
582
+ "bits": 8
583
+ },
584
+ "language_model.model.layers.30.mlp.shared_expert_gate": {
585
+ "group_size": 64,
586
+ "bits": 8
587
+ },
588
+ "language_model.model.layers.31.mlp.gate": {
589
+ "group_size": 64,
590
+ "bits": 8
591
+ },
592
+ "language_model.model.layers.31.mlp.shared_expert_gate": {
593
+ "group_size": 64,
594
+ "bits": 8
595
+ },
596
+ "language_model.model.layers.32.mlp.gate": {
597
+ "group_size": 64,
598
+ "bits": 8
599
+ },
600
+ "language_model.model.layers.32.mlp.shared_expert_gate": {
601
+ "group_size": 64,
602
+ "bits": 8
603
+ },
604
+ "language_model.model.layers.33.mlp.gate": {
605
+ "group_size": 64,
606
+ "bits": 8
607
+ },
608
+ "language_model.model.layers.33.mlp.shared_expert_gate": {
609
+ "group_size": 64,
610
+ "bits": 8
611
+ },
612
+ "language_model.model.layers.34.mlp.gate": {
613
+ "group_size": 64,
614
+ "bits": 8
615
+ },
616
+ "language_model.model.layers.34.mlp.shared_expert_gate": {
617
+ "group_size": 64,
618
+ "bits": 8
619
+ },
620
+ "language_model.model.layers.35.mlp.gate": {
621
+ "group_size": 64,
622
+ "bits": 8
623
+ },
624
+ "language_model.model.layers.35.mlp.shared_expert_gate": {
625
+ "group_size": 64,
626
+ "bits": 8
627
+ },
628
+ "language_model.model.layers.36.mlp.gate": {
629
+ "group_size": 64,
630
+ "bits": 8
631
+ },
632
+ "language_model.model.layers.36.mlp.shared_expert_gate": {
633
+ "group_size": 64,
634
+ "bits": 8
635
+ },
636
+ "language_model.model.layers.37.mlp.gate": {
637
+ "group_size": 64,
638
+ "bits": 8
639
+ },
640
+ "language_model.model.layers.37.mlp.shared_expert_gate": {
641
+ "group_size": 64,
642
+ "bits": 8
643
+ },
644
+ "language_model.model.layers.38.mlp.gate": {
645
+ "group_size": 64,
646
+ "bits": 8
647
+ },
648
+ "language_model.model.layers.38.mlp.shared_expert_gate": {
649
+ "group_size": 64,
650
+ "bits": 8
651
+ },
652
+ "language_model.model.layers.39.mlp.gate": {
653
+ "group_size": 64,
654
+ "bits": 8
655
+ },
656
+ "language_model.model.layers.39.mlp.shared_expert_gate": {
657
+ "group_size": 64,
658
+ "bits": 8
659
+ }
660
+ },
661
+ "text_config": {
662
+ "attention_bias": false,
663
+ "attention_dropout": 0.0,
664
+ "attn_output_gate": true,
665
+ "dtype": "bfloat16",
666
+ "eos_token_id": 248044,
667
+ "full_attention_interval": 4,
668
+ "head_dim": 256,
669
+ "hidden_act": "silu",
670
+ "hidden_size": 2048,
671
+ "initializer_range": 0.02,
672
+ "layer_types": [
673
+ "linear_attention",
674
+ "linear_attention",
675
+ "linear_attention",
676
+ "full_attention",
677
+ "linear_attention",
678
+ "linear_attention",
679
+ "linear_attention",
680
+ "full_attention",
681
+ "linear_attention",
682
+ "linear_attention",
683
+ "linear_attention",
684
+ "full_attention",
685
+ "linear_attention",
686
+ "linear_attention",
687
+ "linear_attention",
688
+ "full_attention",
689
+ "linear_attention",
690
+ "linear_attention",
691
+ "linear_attention",
692
+ "full_attention",
693
+ "linear_attention",
694
+ "linear_attention",
695
+ "linear_attention",
696
+ "full_attention",
697
+ "linear_attention",
698
+ "linear_attention",
699
+ "linear_attention",
700
+ "full_attention",
701
+ "linear_attention",
702
+ "linear_attention",
703
+ "linear_attention",
704
+ "full_attention",
705
+ "linear_attention",
706
+ "linear_attention",
707
+ "linear_attention",
708
+ "full_attention",
709
+ "linear_attention",
710
+ "linear_attention",
711
+ "linear_attention",
712
+ "full_attention"
713
+ ],
714
+ "linear_conv_kernel_dim": 4,
715
+ "linear_key_head_dim": 128,
716
+ "linear_num_key_heads": 16,
717
+ "linear_num_value_heads": 32,
718
+ "linear_value_head_dim": 128,
719
+ "max_position_embeddings": 262144,
720
+ "mlp_only_layers": [],
721
+ "model_type": "qwen3_5_moe_text",
722
+ "moe_intermediate_size": 512,
723
+ "mtp_num_hidden_layers": 1,
724
+ "mtp_use_dedicated_embeddings": false,
725
+ "num_attention_heads": 16,
726
+ "num_experts": 256,
727
+ "num_experts_per_tok": 8,
728
+ "num_hidden_layers": 40,
729
+ "num_key_value_heads": 2,
730
+ "rms_norm_eps": 1e-06,
731
+ "router_aux_loss_coef": 0.001,
732
+ "shared_expert_intermediate_size": 512,
733
+ "use_cache": true,
734
+ "vocab_size": 248320,
735
+ "mamba_ssm_dtype": "float32",
736
+ "rope_parameters": {
737
+ "mrope_interleaved": true,
738
+ "mrope_section": [
739
+ 11,
740
+ 11,
741
+ 10
742
+ ],
743
+ "rope_theta": 10000000,
744
+ "partial_rotary_factor": 0.25,
745
+ "type": "default"
746
+ }
747
+ },
748
+ "tie_word_embeddings": false,
749
+ "transformers_version": "4.57.0.dev0",
750
+ "video_token_id": 248057,
751
+ "vision_end_token_id": 248054,
752
+ "vision_start_token_id": 248053
753
+ }
generation_config.json ADDED
@@ -0,0 +1,13 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token_id": 248044,
3
+ "do_sample": true,
4
+ "eos_token_id": [
5
+ 248046,
6
+ 248044
7
+ ],
8
+ "pad_token_id": 248044,
9
+ "temperature": 1.0,
10
+ "top_k": 20,
11
+ "top_p": 0.95,
12
+ "transformers_version": "4.57.0.dev0"
13
+ }
model-00001-of-00005.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:26ee2e747eced8ba0ae63f8ee9b8d6c0e65207936137c31bd4abedb518d11d39
3
+ size 5362048467
model-00002-of-00005.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4934e0bdb97c4ba1928637b228cde813e24814ec15372df61852445229e1a0bc
3
+ size 5215907743
model-00003-of-00005.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cc641f8bb522d4b689ee9ce0dccf25e58ec866bc181ec373b728d287348348b7
3
+ size 5207519082
model-00004-of-00005.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e9180894e565f54126ec240a3d58905f39ef1fe017425d7f33592f3ece798564
3
+ size 5203037976
model-00005-of-00005.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fb5f321d6f228d7c31104d71d01607fd4ecf6bc61769cad6158dc67c7da5ee84
3
+ size 2850310361
model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:06b9509352d2af50381ab2247e083b80d32d5c0aba91c272ca9ff729b6a0e523
3
+ size 19989325
tokenizer_config.json ADDED
@@ -0,0 +1,33 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_prefix_space": false,
3
+ "audio_bos_token": "<|audio_start|>",
4
+ "audio_eos_token": "<|audio_end|>",
5
+ "audio_token": "<|audio_pad|>",
6
+ "backend": "tokenizers",
7
+ "bos_token": null,
8
+ "clean_up_tokenization_spaces": false,
9
+ "eos_token": "<|im_end|>",
10
+ "errors": "replace",
11
+ "image_token": "<|image_pad|>",
12
+ "is_local": true,
13
+ "local_files_only": false,
14
+ "model_max_length": 262144,
15
+ "model_specific_special_tokens": {
16
+ "audio_bos_token": "<|audio_start|>",
17
+ "audio_eos_token": "<|audio_end|>",
18
+ "audio_token": "<|audio_pad|>",
19
+ "image_token": "<|image_pad|>",
20
+ "video_token": "<|video_pad|>",
21
+ "vision_bos_token": "<|vision_start|>",
22
+ "vision_eos_token": "<|vision_end|>"
23
+ },
24
+ "pad_token": "<|endoftext|>",
25
+ "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
26
+ "split_special_tokens": false,
27
+ "tokenizer_class": "Qwen2Tokenizer",
28
+ "tool_parser_type": "qwen3_coder",
29
+ "unk_token": null,
30
+ "video_token": "<|video_pad|>",
31
+ "vision_bos_token": "<|vision_start|>",
32
+ "vision_eos_token": "<|vision_end|>"
33
+ }