Hugging Face's logo

amd
/
GLM-5.1-NVFP4

Safetensors
glm_moe_dsa
8-bit precision
quark
Model card Files
xet
 Community
4

upload nvfp4model

#1
by linzhao-amd - opened
base: refs/heads/main
←
from: refs/pr/1
Discussion Files changed
+1289
-0
Files changed (17) hide show
  1. .gitattributes +2 -0
  2. chat_template.jinja +119 -0
  3. config.json +874 -0
  4. generation_config.json +13 -0
  5. model-00001-of-00009.safetensors +3 -0
  6. model-00002-of-00009.safetensors +3 -0
  7. model-00003-of-00009.safetensors +3 -0
  8. model-00004-of-00009.safetensors +3 -0
  9. model-00005-of-00009.safetensors +3 -0
  10. model-00006-of-00009.safetensors +3 -0
  11. model-00007-of-00009.safetensors +3 -0
  12. model-00008-of-00009.safetensors +3 -0
  13. model-00009-of-00009.safetensors +3 -0
  14. model.safetensors.index.json +3 -0
  15. quantize_20260507_114729.log +0 -0
  16. tokenizer.json +3 -0
  17. tokenizer_config.json +33 -0
.gitattributes CHANGED
@@ -33,3 +33,5 @@ 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
+ model.safetensors.index.json filter=lfs diff=lfs merge=lfs -text
37
+ tokenizer.json filter=lfs diff=lfs merge=lfs -text
chat_template.jinja ADDED
@@ -0,0 +1,119 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [gMASK]<sop>
2
+ {%- if tools -%}
3
+ {%- macro tool_to_json(tool) -%}
4
+ {%- set ns_tool = namespace(first=true) -%}
5
+ {{ '{' -}}
6
+ {%- for k, v in tool.items() -%}
7
+ {%- if k != 'defer_loading' and k != 'strict' -%}
8
+ {%- if not ns_tool.first -%}{{- ', ' -}}{%- endif -%}
9
+ {%- set ns_tool.first = false -%}
10
+ "{{ k }}": {{ v | tojson(ensure_ascii=False) }}
11
+ {%- endif -%}
12
+ {%- endfor -%}
13
+ {{- '}' -}}
14
+ {%- endmacro -%}
15
+ <|system|>
16
+ # Tools
17
+
18
+ You may call one or more functions to assist with the user query.
19
+
20
+ You are provided with function signatures within <tools></tools> XML tags:
21
+ <tools>
22
+ {% for tool in tools %}
23
+ {%- if 'function' in tool -%}
24
+ {%- set tool = tool['function'] -%}
25
+ {%- endif -%}
26
+ {% if tool.defer_loading is not defined or not tool.defer_loading %}
27
+ {{ tool_to_json(tool) }}
28
+ {% endif %}
29
+ {% endfor %}
30
+ </tools>
31
+
32
+ For each function call, output the function name and arguments within the following XML format:
33
+ <tool_call>{function-name}<arg_key>{arg-key-1}</arg_key><arg_value>{arg-value-1}</arg_value><arg_key>{arg-key-2}</arg_key><arg_value>{arg-value-2}</arg_value>...</tool_call>{%- endif -%}
34
+ {%- macro visible_text(content) -%}
35
+ {%- if content is string -%}
36
+ {{- content }}
37
+ {%- elif content is iterable and content is not mapping -%}
38
+ {%- for item in content -%}
39
+ {%- if item is mapping and item.type == 'text' -%}
40
+ {{- item.text }}
41
+ {%- elif item is string -%}
42
+ {{- item }}
43
+ {%- endif -%}
44
+ {%- endfor -%}
45
+ {%- else -%}
46
+ {{- content }}
47
+ {%- endif -%}
48
+ {%- endmacro -%}
49
+ {%- set ns = namespace(last_user_index=-1, thinking_indices='') -%}
50
+ {%- for m in messages %}
51
+ {%- if m.role == 'user' %}
52
+ {%- set ns.last_user_index = loop.index0 -%}
53
+ {%- elif m.role == 'assistant' %}
54
+ {%- if m.reasoning_content is string %}
55
+ {%- set ns.thinking_indices = ns.thinking_indices ~ ',' ~ ns.last_user_index ~ ',' -%}
56
+ {%- endif %}
57
+ {%- endif %}
58
+ {%- endfor %}
59
+ {%- set ns.has_thinking = false -%}
60
+ {%- for m in messages -%}
61
+ {%- if m.role == 'user' -%}<|user|>{{ visible_text(m.content) }}{% set ns.has_thinking = (',' ~ loop.index0 ~ ',') in ns.thinking_indices -%}
62
+ {%- elif m.role == 'assistant' -%}
63
+ <|assistant|>
64
+ {%- set content = visible_text(m.content) %}
65
+ {%- if m.reasoning_content is string %}
66
+ {%- set reasoning_content = m.reasoning_content %}
67
+ {%- elif '</think>' in content %}
68
+ {%- set reasoning_content = content.split('</think>')[0].split('<think>')[-1] %}
69
+ {%- set content = content.split('</think>')[-1] %}
70
+ {%- elif loop.index0 > ns.last_user_index and not (enable_thinking is defined and not enable_thinking) %}
71
+ {%- set reasoning_content = '' %}
72
+ {%- elif loop.index0 < ns.last_user_index and ns.has_thinking %}
73
+ {%- set reasoning_content = '' %}
74
+ {%- endif %}
75
+ {%- if ((clear_thinking is defined and not clear_thinking) or loop.index0 > ns.last_user_index) and reasoning_content is defined -%}
76
+ {{ '<think>' + reasoning_content + '</think>'}}
77
+ {%- else -%}
78
+ {{ '</think>' }}
79
+ {%- endif -%}
80
+ {%- if content.strip() -%}
81
+ {{ content.strip() }}
82
+ {%- endif -%}
83
+ {% if m.tool_calls %}
84
+ {% for tc in m.tool_calls %}
85
+ {%- if tc.function %}
86
+ {%- set tc = tc.function %}
87
+ {%- endif %}
88
+ {{- '<tool_call>' + tc.name -}}
89
+ {% set _args = tc.arguments %}{% for k, v in _args.items() %}<arg_key>{{ k }}</arg_key><arg_value>{{ v | tojson(ensure_ascii=False) if v is not string else v }}</arg_value>{% endfor %}</tool_call>{% endfor %}
90
+ {% endif %}
91
+ {%- elif m.role == 'tool' -%}
92
+ {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
93
+ {{- '<|observation|>' -}}
94
+ {%- endif %}
95
+ {%- if m.content is string -%}
96
+ {{- '<tool_response>' + m.content + '</tool_response>' -}}
97
+ {%- elif m.content is iterable and m.content is not mapping and m.content and m.content.0.type == "tool_reference" -%}
98
+ {{- '<tool_response><tools>\n' -}}
99
+ {% for tr in m.content %}
100
+ {%- for tool in tools -%}
101
+ {%- if 'function' in tool -%}
102
+ {%- set tool = tool['function'] -%}
103
+ {%- endif -%}
104
+ {%- if tool.name == tr.name -%}
105
+ {{- tool_to_json(tool) + '\n' -}}
106
+ {%- endif -%}
107
+ {%- endfor -%}
108
+ {%- endfor -%}
109
+ {{- '</tools></tool_response>' -}}
110
+ {%- else -%}
111
+ {{- '<tool_response>' + visible_text(m.content) + '</tool_response>' -}}
112
+ {% endif -%}
113
+ {%- elif m.role == 'system' -%}
114
+ <|system|>{{ visible_text(m.content) }}
115
+ {%- endif -%}
116
+ {%- endfor -%}
117
+ {%- if add_generation_prompt -%}
118
+ <|assistant|>{{- '</think>' if (enable_thinking is defined and not enable_thinking) else '<think>' -}}
119
+ {%- endif -%}
config.json ADDED
@@ -0,0 +1,874 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "GlmMoeDsaForCausalLM"
4
+ ],
5
+ "attention_bias": false,
6
+ "attention_dropout": 0.0,
7
+ "bos_token_id": 0,
8
+ "dtype": "bfloat16",
9
+ "eos_token_id": [
10
+ 154820,
11
+ 154827,
12
+ 154829
13
+ ],
14
+ "ep_size": 1,
15
+ "first_k_dense_replace": 3,
16
+ "hidden_act": "silu",
17
+ "hidden_size": 6144,
18
+ "index_head_dim": 128,
19
+ "index_n_heads": 32,
20
+ "index_topk": 2048,
21
+ "indexer_rope_interleave": true,
22
+ "initializer_range": 0.02,
23
+ "intermediate_size": 12288,
24
+ "kv_lora_rank": 512,
25
+ "max_position_embeddings": 202752,
26
+ "mlp_layer_types": [
27
+ "dense",
28
+ "dense",
29
+ "dense",
30
+ "sparse",
31
+ "sparse",
32
+ "sparse",
33
+ "sparse",
34
+ "sparse",
35
+ "sparse",
36
+ "sparse",
37
+ "sparse",
38
+ "sparse",
39
+ "sparse",
40
+ "sparse",
41
+ "sparse",
42
+ "sparse",
43
+ "sparse",
44
+ "sparse",
45
+ "sparse",
46
+ "sparse",
47
+ "sparse",
48
+ "sparse",
49
+ "sparse",
50
+ "sparse",
51
+ "sparse",
52
+ "sparse",
53
+ "sparse",
54
+ "sparse",
55
+ "sparse",
56
+ "sparse",
57
+ "sparse",
58
+ "sparse",
59
+ "sparse",
60
+ "sparse",
61
+ "sparse",
62
+ "sparse",
63
+ "sparse",
64
+ "sparse",
65
+ "sparse",
66
+ "sparse",
67
+ "sparse",
68
+ "sparse",
69
+ "sparse",
70
+ "sparse",
71
+ "sparse",
72
+ "sparse",
73
+ "sparse",
74
+ "sparse",
75
+ "sparse",
76
+ "sparse",
77
+ "sparse",
78
+ "sparse",
79
+ "sparse",
80
+ "sparse",
81
+ "sparse",
82
+ "sparse",
83
+ "sparse",
84
+ "sparse",
85
+ "sparse",
86
+ "sparse",
87
+ "sparse",
88
+ "sparse",
89
+ "sparse",
90
+ "sparse",
91
+ "sparse",
92
+ "sparse",
93
+ "sparse",
94
+ "sparse",
95
+ "sparse",
96
+ "sparse",
97
+ "sparse",
98
+ "sparse",
99
+ "sparse",
100
+ "sparse",
101
+ "sparse",
102
+ "sparse",
103
+ "sparse",
104
+ "sparse"
105
+ ],
106
+ "model_type": "glm_moe_dsa",
107
+ "moe_intermediate_size": 2048,
108
+ "moe_layer_freq": 1,
109
+ "n_group": 1,
110
+ "n_routed_experts": 256,
111
+ "n_shared_experts": 1,
112
+ "norm_topk_prob": true,
113
+ "num_attention_heads": 64,
114
+ "num_experts_per_tok": 8,
115
+ "num_hidden_layers": 78,
116
+ "num_key_value_heads": 64,
117
+ "num_nextn_predict_layers": 1,
118
+ "pad_token_id": 154820,
119
+ "pretraining_tp": 1,
120
+ "q_lora_rank": 2048,
121
+ "qk_head_dim": 256,
122
+ "qk_nope_head_dim": 192,
123
+ "qk_rope_head_dim": 64,
124
+ "quantization_config": {
125
+ "algo_config": null,
126
+ "exclude": [
127
+ "model.layers.0.self_attn.q_a_proj",
128
+ "model.layers.0.self_attn.q_b_proj",
129
+ "model.layers.0.self_attn.kv_a_proj_with_mqa",
130
+ "model.layers.0.self_attn.kv_b_proj",
131
+ "model.layers.0.self_attn.o_proj",
132
+ "model.layers.0.self_attn.indexer.wq_b",
133
+ "model.layers.0.self_attn.indexer.wk",
134
+ "model.layers.0.self_attn.indexer.weights_proj",
135
+ "model.layers.0.mlp.gate_proj",
136
+ "model.layers.0.mlp.up_proj",
137
+ "model.layers.0.mlp.down_proj",
138
+ "model.layers.1.self_attn.q_a_proj",
139
+ "model.layers.1.self_attn.q_b_proj",
140
+ "model.layers.1.self_attn.kv_a_proj_with_mqa",
141
+ "model.layers.1.self_attn.kv_b_proj",
142
+ "model.layers.1.self_attn.o_proj",
143
+ "model.layers.1.self_attn.indexer.wq_b",
144
+ "model.layers.1.self_attn.indexer.wk",
145
+ "model.layers.1.self_attn.indexer.weights_proj",
146
+ "model.layers.1.mlp.gate_proj",
147
+ "model.layers.1.mlp.up_proj",
148
+ "model.layers.1.mlp.down_proj",
149
+ "model.layers.2.self_attn.q_a_proj",
150
+ "model.layers.2.self_attn.q_b_proj",
151
+ "model.layers.2.self_attn.kv_a_proj_with_mqa",
152
+ "model.layers.2.self_attn.kv_b_proj",
153
+ "model.layers.2.self_attn.o_proj",
154
+ "model.layers.2.self_attn.indexer.wq_b",
155
+ "model.layers.2.self_attn.indexer.wk",
156
+ "model.layers.2.self_attn.indexer.weights_proj",
157
+ "model.layers.2.mlp.gate_proj",
158
+ "model.layers.2.mlp.up_proj",
159
+ "model.layers.2.mlp.down_proj",
160
+ "model.layers.3.self_attn.q_a_proj",
161
+ "model.layers.3.self_attn.q_b_proj",
162
+ "model.layers.3.self_attn.kv_a_proj_with_mqa",
163
+ "model.layers.3.self_attn.kv_b_proj",
164
+ "model.layers.3.self_attn.o_proj",
165
+ "model.layers.3.self_attn.indexer.wq_b",
166
+ "model.layers.3.self_attn.indexer.wk",
167
+ "model.layers.3.self_attn.indexer.weights_proj",
168
+ "model.layers.4.self_attn.q_a_proj",
169
+ "model.layers.4.self_attn.q_b_proj",
170
+ "model.layers.4.self_attn.kv_a_proj_with_mqa",
171
+ "model.layers.4.self_attn.kv_b_proj",
172
+ "model.layers.4.self_attn.o_proj",
173
+ "model.layers.4.self_attn.indexer.wq_b",
174
+ "model.layers.4.self_attn.indexer.wk",
175
+ "model.layers.4.self_attn.indexer.weights_proj",
176
+ "model.layers.5.self_attn.q_a_proj",
177
+ "model.layers.5.self_attn.q_b_proj",
178
+ "model.layers.5.self_attn.kv_a_proj_with_mqa",
179
+ "model.layers.5.self_attn.kv_b_proj",
180
+ "model.layers.5.self_attn.o_proj",
181
+ "model.layers.5.self_attn.indexer.wq_b",
182
+ "model.layers.5.self_attn.indexer.wk",
183
+ "model.layers.5.self_attn.indexer.weights_proj",
184
+ "model.layers.6.self_attn.q_a_proj",
185
+ "model.layers.6.self_attn.q_b_proj",
186
+ "model.layers.6.self_attn.kv_a_proj_with_mqa",
187
+ "model.layers.6.self_attn.kv_b_proj",
188
+ "model.layers.6.self_attn.o_proj",
189
+ "model.layers.6.self_attn.indexer.wq_b",
190
+ "model.layers.6.self_attn.indexer.wk",
191
+ "model.layers.6.self_attn.indexer.weights_proj",
192
+ "model.layers.7.self_attn.q_a_proj",
193
+ "model.layers.7.self_attn.q_b_proj",
194
+ "model.layers.7.self_attn.kv_a_proj_with_mqa",
195
+ "model.layers.7.self_attn.kv_b_proj",
196
+ "model.layers.7.self_attn.o_proj",
197
+ "model.layers.7.self_attn.indexer.wq_b",
198
+ "model.layers.7.self_attn.indexer.wk",
199
+ "model.layers.7.self_attn.indexer.weights_proj",
200
+ "model.layers.8.self_attn.q_a_proj",
201
+ "model.layers.8.self_attn.q_b_proj",
202
+ "model.layers.8.self_attn.kv_a_proj_with_mqa",
203
+ "model.layers.8.self_attn.kv_b_proj",
204
+ "model.layers.8.self_attn.o_proj",
205
+ "model.layers.8.self_attn.indexer.wq_b",
206
+ "model.layers.8.self_attn.indexer.wk",
207
+ "model.layers.8.self_attn.indexer.weights_proj",
208
+ "model.layers.9.self_attn.q_a_proj",
209
+ "model.layers.9.self_attn.q_b_proj",
210
+ "model.layers.9.self_attn.kv_a_proj_with_mqa",
211
+ "model.layers.9.self_attn.kv_b_proj",
212
+ "model.layers.9.self_attn.o_proj",
213
+ "model.layers.9.self_attn.indexer.wq_b",
214
+ "model.layers.9.self_attn.indexer.wk",
215
+ "model.layers.9.self_attn.indexer.weights_proj",
216
+ "model.layers.10.self_attn.q_a_proj",
217
+ "model.layers.10.self_attn.q_b_proj",
218
+ "model.layers.10.self_attn.kv_a_proj_with_mqa",
219
+ "model.layers.10.self_attn.kv_b_proj",
220
+ "model.layers.10.self_attn.o_proj",
221
+ "model.layers.10.self_attn.indexer.wq_b",
222
+ "model.layers.10.self_attn.indexer.wk",
223
+ "model.layers.10.self_attn.indexer.weights_proj",
224
+ "model.layers.11.self_attn.q_a_proj",
225
+ "model.layers.11.self_attn.q_b_proj",
226
+ "model.layers.11.self_attn.kv_a_proj_with_mqa",
227
+ "model.layers.11.self_attn.kv_b_proj",
228
+ "model.layers.11.self_attn.o_proj",
229
+ "model.layers.11.self_attn.indexer.wq_b",
230
+ "model.layers.11.self_attn.indexer.wk",
231
+ "model.layers.11.self_attn.indexer.weights_proj",
232
+ "model.layers.12.self_attn.q_a_proj",
233
+ "model.layers.12.self_attn.q_b_proj",
234
+ "model.layers.12.self_attn.kv_a_proj_with_mqa",
235
+ "model.layers.12.self_attn.kv_b_proj",
236
+ "model.layers.12.self_attn.o_proj",
237
+ "model.layers.12.self_attn.indexer.wq_b",
238
+ "model.layers.12.self_attn.indexer.wk",
239
+ "model.layers.12.self_attn.indexer.weights_proj",
240
+ "model.layers.13.self_attn.q_a_proj",
241
+ "model.layers.13.self_attn.q_b_proj",
242
+ "model.layers.13.self_attn.kv_a_proj_with_mqa",
243
+ "model.layers.13.self_attn.kv_b_proj",
244
+ "model.layers.13.self_attn.o_proj",
245
+ "model.layers.13.self_attn.indexer.wq_b",
246
+ "model.layers.13.self_attn.indexer.wk",
247
+ "model.layers.13.self_attn.indexer.weights_proj",
248
+ "model.layers.14.self_attn.q_a_proj",
249
+ "model.layers.14.self_attn.q_b_proj",
250
+ "model.layers.14.self_attn.kv_a_proj_with_mqa",
251
+ "model.layers.14.self_attn.kv_b_proj",
252
+ "model.layers.14.self_attn.o_proj",
253
+ "model.layers.14.self_attn.indexer.wq_b",
254
+ "model.layers.14.self_attn.indexer.wk",
255
+ "model.layers.14.self_attn.indexer.weights_proj",
256
+ "model.layers.15.self_attn.q_a_proj",
257
+ "model.layers.15.self_attn.q_b_proj",
258
+ "model.layers.15.self_attn.kv_a_proj_with_mqa",
259
+ "model.layers.15.self_attn.kv_b_proj",
260
+ "model.layers.15.self_attn.o_proj",
261
+ "model.layers.15.self_attn.indexer.wq_b",
262
+ "model.layers.15.self_attn.indexer.wk",
263
+ "model.layers.15.self_attn.indexer.weights_proj",
264
+ "model.layers.16.self_attn.q_a_proj",
265
+ "model.layers.16.self_attn.q_b_proj",
266
+ "model.layers.16.self_attn.kv_a_proj_with_mqa",
267
+ "model.layers.16.self_attn.kv_b_proj",
268
+ "model.layers.16.self_attn.o_proj",
269
+ "model.layers.16.self_attn.indexer.wq_b",
270
+ "model.layers.16.self_attn.indexer.wk",
271
+ "model.layers.16.self_attn.indexer.weights_proj",
272
+ "model.layers.17.self_attn.q_a_proj",
273
+ "model.layers.17.self_attn.q_b_proj",
274
+ "model.layers.17.self_attn.kv_a_proj_with_mqa",
275
+ "model.layers.17.self_attn.kv_b_proj",
276
+ "model.layers.17.self_attn.o_proj",
277
+ "model.layers.17.self_attn.indexer.wq_b",
278
+ "model.layers.17.self_attn.indexer.wk",
279
+ "model.layers.17.self_attn.indexer.weights_proj",
280
+ "model.layers.18.self_attn.q_a_proj",
281
+ "model.layers.18.self_attn.q_b_proj",
282
+ "model.layers.18.self_attn.kv_a_proj_with_mqa",
283
+ "model.layers.18.self_attn.kv_b_proj",
284
+ "model.layers.18.self_attn.o_proj",
285
+ "model.layers.18.self_attn.indexer.wq_b",
286
+ "model.layers.18.self_attn.indexer.wk",
287
+ "model.layers.18.self_attn.indexer.weights_proj",
288
+ "model.layers.19.self_attn.q_a_proj",
289
+ "model.layers.19.self_attn.q_b_proj",
290
+ "model.layers.19.self_attn.kv_a_proj_with_mqa",
291
+ "model.layers.19.self_attn.kv_b_proj",
292
+ "model.layers.19.self_attn.o_proj",
293
+ "model.layers.19.self_attn.indexer.wq_b",
294
+ "model.layers.19.self_attn.indexer.wk",
295
+ "model.layers.19.self_attn.indexer.weights_proj",
296
+ "model.layers.20.self_attn.q_a_proj",
297
+ "model.layers.20.self_attn.q_b_proj",
298
+ "model.layers.20.self_attn.kv_a_proj_with_mqa",
299
+ "model.layers.20.self_attn.kv_b_proj",
300
+ "model.layers.20.self_attn.o_proj",
301
+ "model.layers.20.self_attn.indexer.wq_b",
302
+ "model.layers.20.self_attn.indexer.wk",
303
+ "model.layers.20.self_attn.indexer.weights_proj",
304
+ "model.layers.21.self_attn.q_a_proj",
305
+ "model.layers.21.self_attn.q_b_proj",
306
+ "model.layers.21.self_attn.kv_a_proj_with_mqa",
307
+ "model.layers.21.self_attn.kv_b_proj",
308
+ "model.layers.21.self_attn.o_proj",
309
+ "model.layers.21.self_attn.indexer.wq_b",
310
+ "model.layers.21.self_attn.indexer.wk",
311
+ "model.layers.21.self_attn.indexer.weights_proj",
312
+ "model.layers.22.self_attn.q_a_proj",
313
+ "model.layers.22.self_attn.q_b_proj",
314
+ "model.layers.22.self_attn.kv_a_proj_with_mqa",
315
+ "model.layers.22.self_attn.kv_b_proj",
316
+ "model.layers.22.self_attn.o_proj",
317
+ "model.layers.22.self_attn.indexer.wq_b",
318
+ "model.layers.22.self_attn.indexer.wk",
319
+ "model.layers.22.self_attn.indexer.weights_proj",
320
+ "model.layers.23.self_attn.q_a_proj",
321
+ "model.layers.23.self_attn.q_b_proj",
322
+ "model.layers.23.self_attn.kv_a_proj_with_mqa",
323
+ "model.layers.23.self_attn.kv_b_proj",
324
+ "model.layers.23.self_attn.o_proj",
325
+ "model.layers.23.self_attn.indexer.wq_b",
326
+ "model.layers.23.self_attn.indexer.wk",
327
+ "model.layers.23.self_attn.indexer.weights_proj",
328
+ "model.layers.24.self_attn.q_a_proj",
329
+ "model.layers.24.self_attn.q_b_proj",
330
+ "model.layers.24.self_attn.kv_a_proj_with_mqa",
331
+ "model.layers.24.self_attn.kv_b_proj",
332
+ "model.layers.24.self_attn.o_proj",
333
+ "model.layers.24.self_attn.indexer.wq_b",
334
+ "model.layers.24.self_attn.indexer.wk",
335
+ "model.layers.24.self_attn.indexer.weights_proj",
336
+ "model.layers.25.self_attn.q_a_proj",
337
+ "model.layers.25.self_attn.q_b_proj",
338
+ "model.layers.25.self_attn.kv_a_proj_with_mqa",
339
+ "model.layers.25.self_attn.kv_b_proj",
340
+ "model.layers.25.self_attn.o_proj",
341
+ "model.layers.25.self_attn.indexer.wq_b",
342
+ "model.layers.25.self_attn.indexer.wk",
343
+ "model.layers.25.self_attn.indexer.weights_proj",
344
+ "model.layers.26.self_attn.q_a_proj",
345
+ "model.layers.26.self_attn.q_b_proj",
346
+ "model.layers.26.self_attn.kv_a_proj_with_mqa",
347
+ "model.layers.26.self_attn.kv_b_proj",
348
+ "model.layers.26.self_attn.o_proj",
349
+ "model.layers.26.self_attn.indexer.wq_b",
350
+ "model.layers.26.self_attn.indexer.wk",
351
+ "model.layers.26.self_attn.indexer.weights_proj",
352
+ "model.layers.27.self_attn.q_a_proj",
353
+ "model.layers.27.self_attn.q_b_proj",
354
+ "model.layers.27.self_attn.kv_a_proj_with_mqa",
355
+ "model.layers.27.self_attn.kv_b_proj",
356
+ "model.layers.27.self_attn.o_proj",
357
+ "model.layers.27.self_attn.indexer.wq_b",
358
+ "model.layers.27.self_attn.indexer.wk",
359
+ "model.layers.27.self_attn.indexer.weights_proj",
360
+ "model.layers.28.self_attn.q_a_proj",
361
+ "model.layers.28.self_attn.q_b_proj",
362
+ "model.layers.28.self_attn.kv_a_proj_with_mqa",
363
+ "model.layers.28.self_attn.kv_b_proj",
364
+ "model.layers.28.self_attn.o_proj",
365
+ "model.layers.28.self_attn.indexer.wq_b",
366
+ "model.layers.28.self_attn.indexer.wk",
367
+ "model.layers.28.self_attn.indexer.weights_proj",
368
+ "model.layers.29.self_attn.q_a_proj",
369
+ "model.layers.29.self_attn.q_b_proj",
370
+ "model.layers.29.self_attn.kv_a_proj_with_mqa",
371
+ "model.layers.29.self_attn.kv_b_proj",
372
+ "model.layers.29.self_attn.o_proj",
373
+ "model.layers.29.self_attn.indexer.wq_b",
374
+ "model.layers.29.self_attn.indexer.wk",
375
+ "model.layers.29.self_attn.indexer.weights_proj",
376
+ "model.layers.30.self_attn.q_a_proj",
377
+ "model.layers.30.self_attn.q_b_proj",
378
+ "model.layers.30.self_attn.kv_a_proj_with_mqa",
379
+ "model.layers.30.self_attn.kv_b_proj",
380
+ "model.layers.30.self_attn.o_proj",
381
+ "model.layers.30.self_attn.indexer.wq_b",
382
+ "model.layers.30.self_attn.indexer.wk",
383
+ "model.layers.30.self_attn.indexer.weights_proj",
384
+ "model.layers.31.self_attn.q_a_proj",
385
+ "model.layers.31.self_attn.q_b_proj",
386
+ "model.layers.31.self_attn.kv_a_proj_with_mqa",
387
+ "model.layers.31.self_attn.kv_b_proj",
388
+ "model.layers.31.self_attn.o_proj",
389
+ "model.layers.31.self_attn.indexer.wq_b",
390
+ "model.layers.31.self_attn.indexer.wk",
391
+ "model.layers.31.self_attn.indexer.weights_proj",
392
+ "model.layers.32.self_attn.q_a_proj",
393
+ "model.layers.32.self_attn.q_b_proj",
394
+ "model.layers.32.self_attn.kv_a_proj_with_mqa",
395
+ "model.layers.32.self_attn.kv_b_proj",
396
+ "model.layers.32.self_attn.o_proj",
397
+ "model.layers.32.self_attn.indexer.wq_b",
398
+ "model.layers.32.self_attn.indexer.wk",
399
+ "model.layers.32.self_attn.indexer.weights_proj",
400
+ "model.layers.33.self_attn.q_a_proj",
401
+ "model.layers.33.self_attn.q_b_proj",
402
+ "model.layers.33.self_attn.kv_a_proj_with_mqa",
403
+ "model.layers.33.self_attn.kv_b_proj",
404
+ "model.layers.33.self_attn.o_proj",
405
+ "model.layers.33.self_attn.indexer.wq_b",
406
+ "model.layers.33.self_attn.indexer.wk",
407
+ "model.layers.33.self_attn.indexer.weights_proj",
408
+ "model.layers.34.self_attn.q_a_proj",
409
+ "model.layers.34.self_attn.q_b_proj",
410
+ "model.layers.34.self_attn.kv_a_proj_with_mqa",
411
+ "model.layers.34.self_attn.kv_b_proj",
412
+ "model.layers.34.self_attn.o_proj",
413
+ "model.layers.34.self_attn.indexer.wq_b",
414
+ "model.layers.34.self_attn.indexer.wk",
415
+ "model.layers.34.self_attn.indexer.weights_proj",
416
+ "model.layers.35.self_attn.q_a_proj",
417
+ "model.layers.35.self_attn.q_b_proj",
418
+ "model.layers.35.self_attn.kv_a_proj_with_mqa",
419
+ "model.layers.35.self_attn.kv_b_proj",
420
+ "model.layers.35.self_attn.o_proj",
421
+ "model.layers.35.self_attn.indexer.wq_b",
422
+ "model.layers.35.self_attn.indexer.wk",
423
+ "model.layers.35.self_attn.indexer.weights_proj",
424
+ "model.layers.36.self_attn.q_a_proj",
425
+ "model.layers.36.self_attn.q_b_proj",
426
+ "model.layers.36.self_attn.kv_a_proj_with_mqa",
427
+ "model.layers.36.self_attn.kv_b_proj",
428
+ "model.layers.36.self_attn.o_proj",
429
+ "model.layers.36.self_attn.indexer.wq_b",
430
+ "model.layers.36.self_attn.indexer.wk",
431
+ "model.layers.36.self_attn.indexer.weights_proj",
432
+ "model.layers.37.self_attn.q_a_proj",
433
+ "model.layers.37.self_attn.q_b_proj",
434
+ "model.layers.37.self_attn.kv_a_proj_with_mqa",
435
+ "model.layers.37.self_attn.kv_b_proj",
436
+ "model.layers.37.self_attn.o_proj",
437
+ "model.layers.37.self_attn.indexer.wq_b",
438
+ "model.layers.37.self_attn.indexer.wk",
439
+ "model.layers.37.self_attn.indexer.weights_proj",
440
+ "model.layers.38.self_attn.q_a_proj",
441
+ "model.layers.38.self_attn.q_b_proj",
442
+ "model.layers.38.self_attn.kv_a_proj_with_mqa",
443
+ "model.layers.38.self_attn.kv_b_proj",
444
+ "model.layers.38.self_attn.o_proj",
445
+ "model.layers.38.self_attn.indexer.wq_b",
446
+ "model.layers.38.self_attn.indexer.wk",
447
+ "model.layers.38.self_attn.indexer.weights_proj",
448
+ "model.layers.39.self_attn.q_a_proj",
449
+ "model.layers.39.self_attn.q_b_proj",
450
+ "model.layers.39.self_attn.kv_a_proj_with_mqa",
451
+ "model.layers.39.self_attn.kv_b_proj",
452
+ "model.layers.39.self_attn.o_proj",
453
+ "model.layers.39.self_attn.indexer.wq_b",
454
+ "model.layers.39.self_attn.indexer.wk",
455
+ "model.layers.39.self_attn.indexer.weights_proj",
456
+ "model.layers.40.self_attn.q_a_proj",
457
+ "model.layers.40.self_attn.q_b_proj",
458
+ "model.layers.40.self_attn.kv_a_proj_with_mqa",
459
+ "model.layers.40.self_attn.kv_b_proj",
460
+ "model.layers.40.self_attn.o_proj",
461
+ "model.layers.40.self_attn.indexer.wq_b",
462
+ "model.layers.40.self_attn.indexer.wk",
463
+ "model.layers.40.self_attn.indexer.weights_proj",
464
+ "model.layers.41.self_attn.q_a_proj",
465
+ "model.layers.41.self_attn.q_b_proj",
466
+ "model.layers.41.self_attn.kv_a_proj_with_mqa",
467
+ "model.layers.41.self_attn.kv_b_proj",
468
+ "model.layers.41.self_attn.o_proj",
469
+ "model.layers.41.self_attn.indexer.wq_b",
470
+ "model.layers.41.self_attn.indexer.wk",
471
+ "model.layers.41.self_attn.indexer.weights_proj",
472
+ "model.layers.42.self_attn.q_a_proj",
473
+ "model.layers.42.self_attn.q_b_proj",
474
+ "model.layers.42.self_attn.kv_a_proj_with_mqa",
475
+ "model.layers.42.self_attn.kv_b_proj",
476
+ "model.layers.42.self_attn.o_proj",
477
+ "model.layers.42.self_attn.indexer.wq_b",
478
+ "model.layers.42.self_attn.indexer.wk",
479
+ "model.layers.42.self_attn.indexer.weights_proj",
480
+ "model.layers.43.self_attn.q_a_proj",
481
+ "model.layers.43.self_attn.q_b_proj",
482
+ "model.layers.43.self_attn.kv_a_proj_with_mqa",
483
+ "model.layers.43.self_attn.kv_b_proj",
484
+ "model.layers.43.self_attn.o_proj",
485
+ "model.layers.43.self_attn.indexer.wq_b",
486
+ "model.layers.43.self_attn.indexer.wk",
487
+ "model.layers.43.self_attn.indexer.weights_proj",
488
+ "model.layers.44.self_attn.q_a_proj",
489
+ "model.layers.44.self_attn.q_b_proj",
490
+ "model.layers.44.self_attn.kv_a_proj_with_mqa",
491
+ "model.layers.44.self_attn.kv_b_proj",
492
+ "model.layers.44.self_attn.o_proj",
493
+ "model.layers.44.self_attn.indexer.wq_b",
494
+ "model.layers.44.self_attn.indexer.wk",
495
+ "model.layers.44.self_attn.indexer.weights_proj",
496
+ "model.layers.45.self_attn.q_a_proj",
497
+ "model.layers.45.self_attn.q_b_proj",
498
+ "model.layers.45.self_attn.kv_a_proj_with_mqa",
499
+ "model.layers.45.self_attn.kv_b_proj",
500
+ "model.layers.45.self_attn.o_proj",
501
+ "model.layers.45.self_attn.indexer.wq_b",
502
+ "model.layers.45.self_attn.indexer.wk",
503
+ "model.layers.45.self_attn.indexer.weights_proj",
504
+ "model.layers.46.self_attn.q_a_proj",
505
+ "model.layers.46.self_attn.q_b_proj",
506
+ "model.layers.46.self_attn.kv_a_proj_with_mqa",
507
+ "model.layers.46.self_attn.kv_b_proj",
508
+ "model.layers.46.self_attn.o_proj",
509
+ "model.layers.46.self_attn.indexer.wq_b",
510
+ "model.layers.46.self_attn.indexer.wk",
511
+ "model.layers.46.self_attn.indexer.weights_proj",
512
+ "model.layers.47.self_attn.q_a_proj",
513
+ "model.layers.47.self_attn.q_b_proj",
514
+ "model.layers.47.self_attn.kv_a_proj_with_mqa",
515
+ "model.layers.47.self_attn.kv_b_proj",
516
+ "model.layers.47.self_attn.o_proj",
517
+ "model.layers.47.self_attn.indexer.wq_b",
518
+ "model.layers.47.self_attn.indexer.wk",
519
+ "model.layers.47.self_attn.indexer.weights_proj",
520
+ "model.layers.48.self_attn.q_a_proj",
521
+ "model.layers.48.self_attn.q_b_proj",
522
+ "model.layers.48.self_attn.kv_a_proj_with_mqa",
523
+ "model.layers.48.self_attn.kv_b_proj",
524
+ "model.layers.48.self_attn.o_proj",
525
+ "model.layers.48.self_attn.indexer.wq_b",
526
+ "model.layers.48.self_attn.indexer.wk",
527
+ "model.layers.48.self_attn.indexer.weights_proj",
528
+ "model.layers.49.self_attn.q_a_proj",
529
+ "model.layers.49.self_attn.q_b_proj",
530
+ "model.layers.49.self_attn.kv_a_proj_with_mqa",
531
+ "model.layers.49.self_attn.kv_b_proj",
532
+ "model.layers.49.self_attn.o_proj",
533
+ "model.layers.49.self_attn.indexer.wq_b",
534
+ "model.layers.49.self_attn.indexer.wk",
535
+ "model.layers.49.self_attn.indexer.weights_proj",
536
+ "model.layers.50.self_attn.q_a_proj",
537
+ "model.layers.50.self_attn.q_b_proj",
538
+ "model.layers.50.self_attn.kv_a_proj_with_mqa",
539
+ "model.layers.50.self_attn.kv_b_proj",
540
+ "model.layers.50.self_attn.o_proj",
541
+ "model.layers.50.self_attn.indexer.wq_b",
542
+ "model.layers.50.self_attn.indexer.wk",
543
+ "model.layers.50.self_attn.indexer.weights_proj",
544
+ "model.layers.51.self_attn.q_a_proj",
545
+ "model.layers.51.self_attn.q_b_proj",
546
+ "model.layers.51.self_attn.kv_a_proj_with_mqa",
547
+ "model.layers.51.self_attn.kv_b_proj",
548
+ "model.layers.51.self_attn.o_proj",
549
+ "model.layers.51.self_attn.indexer.wq_b",
550
+ "model.layers.51.self_attn.indexer.wk",
551
+ "model.layers.51.self_attn.indexer.weights_proj",
552
+ "model.layers.52.self_attn.q_a_proj",
553
+ "model.layers.52.self_attn.q_b_proj",
554
+ "model.layers.52.self_attn.kv_a_proj_with_mqa",
555
+ "model.layers.52.self_attn.kv_b_proj",
556
+ "model.layers.52.self_attn.o_proj",
557
+ "model.layers.52.self_attn.indexer.wq_b",
558
+ "model.layers.52.self_attn.indexer.wk",
559
+ "model.layers.52.self_attn.indexer.weights_proj",
560
+ "model.layers.53.self_attn.q_a_proj",
561
+ "model.layers.53.self_attn.q_b_proj",
562
+ "model.layers.53.self_attn.kv_a_proj_with_mqa",
563
+ "model.layers.53.self_attn.kv_b_proj",
564
+ "model.layers.53.self_attn.o_proj",
565
+ "model.layers.53.self_attn.indexer.wq_b",
566
+ "model.layers.53.self_attn.indexer.wk",
567
+ "model.layers.53.self_attn.indexer.weights_proj",
568
+ "model.layers.54.self_attn.q_a_proj",
569
+ "model.layers.54.self_attn.q_b_proj",
570
+ "model.layers.54.self_attn.kv_a_proj_with_mqa",
571
+ "model.layers.54.self_attn.kv_b_proj",
572
+ "model.layers.54.self_attn.o_proj",
573
+ "model.layers.54.self_attn.indexer.wq_b",
574
+ "model.layers.54.self_attn.indexer.wk",
575
+ "model.layers.54.self_attn.indexer.weights_proj",
576
+ "model.layers.55.self_attn.q_a_proj",
577
+ "model.layers.55.self_attn.q_b_proj",
578
+ "model.layers.55.self_attn.kv_a_proj_with_mqa",
579
+ "model.layers.55.self_attn.kv_b_proj",
580
+ "model.layers.55.self_attn.o_proj",
581
+ "model.layers.55.self_attn.indexer.wq_b",
582
+ "model.layers.55.self_attn.indexer.wk",
583
+ "model.layers.55.self_attn.indexer.weights_proj",
584
+ "model.layers.56.self_attn.q_a_proj",
585
+ "model.layers.56.self_attn.q_b_proj",
586
+ "model.layers.56.self_attn.kv_a_proj_with_mqa",
587
+ "model.layers.56.self_attn.kv_b_proj",
588
+ "model.layers.56.self_attn.o_proj",
589
+ "model.layers.56.self_attn.indexer.wq_b",
590
+ "model.layers.56.self_attn.indexer.wk",
591
+ "model.layers.56.self_attn.indexer.weights_proj",
592
+ "model.layers.57.self_attn.q_a_proj",
593
+ "model.layers.57.self_attn.q_b_proj",
594
+ "model.layers.57.self_attn.kv_a_proj_with_mqa",
595
+ "model.layers.57.self_attn.kv_b_proj",
596
+ "model.layers.57.self_attn.o_proj",
597
+ "model.layers.57.self_attn.indexer.wq_b",
598
+ "model.layers.57.self_attn.indexer.wk",
599
+ "model.layers.57.self_attn.indexer.weights_proj",
600
+ "model.layers.58.self_attn.q_a_proj",
601
+ "model.layers.58.self_attn.q_b_proj",
602
+ "model.layers.58.self_attn.kv_a_proj_with_mqa",
603
+ "model.layers.58.self_attn.kv_b_proj",
604
+ "model.layers.58.self_attn.o_proj",
605
+ "model.layers.58.self_attn.indexer.wq_b",
606
+ "model.layers.58.self_attn.indexer.wk",
607
+ "model.layers.58.self_attn.indexer.weights_proj",
608
+ "model.layers.59.self_attn.q_a_proj",
609
+ "model.layers.59.self_attn.q_b_proj",
610
+ "model.layers.59.self_attn.kv_a_proj_with_mqa",
611
+ "model.layers.59.self_attn.kv_b_proj",
612
+ "model.layers.59.self_attn.o_proj",
613
+ "model.layers.59.self_attn.indexer.wq_b",
614
+ "model.layers.59.self_attn.indexer.wk",
615
+ "model.layers.59.self_attn.indexer.weights_proj",
616
+ "model.layers.60.self_attn.q_a_proj",
617
+ "model.layers.60.self_attn.q_b_proj",
618
+ "model.layers.60.self_attn.kv_a_proj_with_mqa",
619
+ "model.layers.60.self_attn.kv_b_proj",
620
+ "model.layers.60.self_attn.o_proj",
621
+ "model.layers.60.self_attn.indexer.wq_b",
622
+ "model.layers.60.self_attn.indexer.wk",
623
+ "model.layers.60.self_attn.indexer.weights_proj",
624
+ "model.layers.61.self_attn.q_a_proj",
625
+ "model.layers.61.self_attn.q_b_proj",
626
+ "model.layers.61.self_attn.kv_a_proj_with_mqa",
627
+ "model.layers.61.self_attn.kv_b_proj",
628
+ "model.layers.61.self_attn.o_proj",
629
+ "model.layers.61.self_attn.indexer.wq_b",
630
+ "model.layers.61.self_attn.indexer.wk",
631
+ "model.layers.61.self_attn.indexer.weights_proj",
632
+ "model.layers.62.self_attn.q_a_proj",
633
+ "model.layers.62.self_attn.q_b_proj",
634
+ "model.layers.62.self_attn.kv_a_proj_with_mqa",
635
+ "model.layers.62.self_attn.kv_b_proj",
636
+ "model.layers.62.self_attn.o_proj",
637
+ "model.layers.62.self_attn.indexer.wq_b",
638
+ "model.layers.62.self_attn.indexer.wk",
639
+ "model.layers.62.self_attn.indexer.weights_proj",
640
+ "model.layers.63.self_attn.q_a_proj",
641
+ "model.layers.63.self_attn.q_b_proj",
642
+ "model.layers.63.self_attn.kv_a_proj_with_mqa",
643
+ "model.layers.63.self_attn.kv_b_proj",
644
+ "model.layers.63.self_attn.o_proj",
645
+ "model.layers.63.self_attn.indexer.wq_b",
646
+ "model.layers.63.self_attn.indexer.wk",
647
+ "model.layers.63.self_attn.indexer.weights_proj",
648
+ "model.layers.64.self_attn.q_a_proj",
649
+ "model.layers.64.self_attn.q_b_proj",
650
+ "model.layers.64.self_attn.kv_a_proj_with_mqa",
651
+ "model.layers.64.self_attn.kv_b_proj",
652
+ "model.layers.64.self_attn.o_proj",
653
+ "model.layers.64.self_attn.indexer.wq_b",
654
+ "model.layers.64.self_attn.indexer.wk",
655
+ "model.layers.64.self_attn.indexer.weights_proj",
656
+ "model.layers.65.self_attn.q_a_proj",
657
+ "model.layers.65.self_attn.q_b_proj",
658
+ "model.layers.65.self_attn.kv_a_proj_with_mqa",
659
+ "model.layers.65.self_attn.kv_b_proj",
660
+ "model.layers.65.self_attn.o_proj",
661
+ "model.layers.65.self_attn.indexer.wq_b",
662
+ "model.layers.65.self_attn.indexer.wk",
663
+ "model.layers.65.self_attn.indexer.weights_proj",
664
+ "model.layers.66.self_attn.q_a_proj",
665
+ "model.layers.66.self_attn.q_b_proj",
666
+ "model.layers.66.self_attn.kv_a_proj_with_mqa",
667
+ "model.layers.66.self_attn.kv_b_proj",
668
+ "model.layers.66.self_attn.o_proj",
669
+ "model.layers.66.self_attn.indexer.wq_b",
670
+ "model.layers.66.self_attn.indexer.wk",
671
+ "model.layers.66.self_attn.indexer.weights_proj",
672
+ "model.layers.67.self_attn.q_a_proj",
673
+ "model.layers.67.self_attn.q_b_proj",
674
+ "model.layers.67.self_attn.kv_a_proj_with_mqa",
675
+ "model.layers.67.self_attn.kv_b_proj",
676
+ "model.layers.67.self_attn.o_proj",
677
+ "model.layers.67.self_attn.indexer.wq_b",
678
+ "model.layers.67.self_attn.indexer.wk",
679
+ "model.layers.67.self_attn.indexer.weights_proj",
680
+ "model.layers.68.self_attn.q_a_proj",
681
+ "model.layers.68.self_attn.q_b_proj",
682
+ "model.layers.68.self_attn.kv_a_proj_with_mqa",
683
+ "model.layers.68.self_attn.kv_b_proj",
684
+ "model.layers.68.self_attn.o_proj",
685
+ "model.layers.68.self_attn.indexer.wq_b",
686
+ "model.layers.68.self_attn.indexer.wk",
687
+ "model.layers.68.self_attn.indexer.weights_proj",
688
+ "model.layers.69.self_attn.q_a_proj",
689
+ "model.layers.69.self_attn.q_b_proj",
690
+ "model.layers.69.self_attn.kv_a_proj_with_mqa",
691
+ "model.layers.69.self_attn.kv_b_proj",
692
+ "model.layers.69.self_attn.o_proj",
693
+ "model.layers.69.self_attn.indexer.wq_b",
694
+ "model.layers.69.self_attn.indexer.wk",
695
+ "model.layers.69.self_attn.indexer.weights_proj",
696
+ "model.layers.70.self_attn.q_a_proj",
697
+ "model.layers.70.self_attn.q_b_proj",
698
+ "model.layers.70.self_attn.kv_a_proj_with_mqa",
699
+ "model.layers.70.self_attn.kv_b_proj",
700
+ "model.layers.70.self_attn.o_proj",
701
+ "model.layers.70.self_attn.indexer.wq_b",
702
+ "model.layers.70.self_attn.indexer.wk",
703
+ "model.layers.70.self_attn.indexer.weights_proj",
704
+ "model.layers.71.self_attn.q_a_proj",
705
+ "model.layers.71.self_attn.q_b_proj",
706
+ "model.layers.71.self_attn.kv_a_proj_with_mqa",
707
+ "model.layers.71.self_attn.kv_b_proj",
708
+ "model.layers.71.self_attn.o_proj",
709
+ "model.layers.71.self_attn.indexer.wq_b",
710
+ "model.layers.71.self_attn.indexer.wk",
711
+ "model.layers.71.self_attn.indexer.weights_proj",
712
+ "model.layers.72.self_attn.q_a_proj",
713
+ "model.layers.72.self_attn.q_b_proj",
714
+ "model.layers.72.self_attn.kv_a_proj_with_mqa",
715
+ "model.layers.72.self_attn.kv_b_proj",
716
+ "model.layers.72.self_attn.o_proj",
717
+ "model.layers.72.self_attn.indexer.wq_b",
718
+ "model.layers.72.self_attn.indexer.wk",
719
+ "model.layers.72.self_attn.indexer.weights_proj",
720
+ "model.layers.73.self_attn.q_a_proj",
721
+ "model.layers.73.self_attn.q_b_proj",
722
+ "model.layers.73.self_attn.kv_a_proj_with_mqa",
723
+ "model.layers.73.self_attn.kv_b_proj",
724
+ "model.layers.73.self_attn.o_proj",
725
+ "model.layers.73.self_attn.indexer.wq_b",
726
+ "model.layers.73.self_attn.indexer.wk",
727
+ "model.layers.73.self_attn.indexer.weights_proj",
728
+ "model.layers.74.self_attn.q_a_proj",
729
+ "model.layers.74.self_attn.q_b_proj",
730
+ "model.layers.74.self_attn.kv_a_proj_with_mqa",
731
+ "model.layers.74.self_attn.kv_b_proj",
732
+ "model.layers.74.self_attn.o_proj",
733
+ "model.layers.74.self_attn.indexer.wq_b",
734
+ "model.layers.74.self_attn.indexer.wk",
735
+ "model.layers.74.self_attn.indexer.weights_proj",
736
+ "model.layers.75.self_attn.q_a_proj",
737
+ "model.layers.75.self_attn.q_b_proj",
738
+ "model.layers.75.self_attn.kv_a_proj_with_mqa",
739
+ "model.layers.75.self_attn.kv_b_proj",
740
+ "model.layers.75.self_attn.o_proj",
741
+ "model.layers.75.self_attn.indexer.wq_b",
742
+ "model.layers.75.self_attn.indexer.wk",
743
+ "model.layers.75.self_attn.indexer.weights_proj",
744
+ "model.layers.76.self_attn.q_a_proj",
745
+ "model.layers.76.self_attn.q_b_proj",
746
+ "model.layers.76.self_attn.kv_a_proj_with_mqa",
747
+ "model.layers.76.self_attn.kv_b_proj",
748
+ "model.layers.76.self_attn.o_proj",
749
+ "model.layers.76.self_attn.indexer.wq_b",
750
+ "model.layers.76.self_attn.indexer.wk",
751
+ "model.layers.76.self_attn.indexer.weights_proj",
752
+ "model.layers.77.self_attn.q_a_proj",
753
+ "model.layers.77.self_attn.q_b_proj",
754
+ "model.layers.77.self_attn.kv_a_proj_with_mqa",
755
+ "model.layers.77.self_attn.kv_b_proj",
756
+ "model.layers.77.self_attn.o_proj",
757
+ "model.layers.77.self_attn.indexer.wq_b",
758
+ "model.layers.77.self_attn.indexer.wk",
759
+ "model.layers.77.self_attn.indexer.weights_proj",
760
+ "lm_head"
761
+ ],
762
+ "export": {
763
+ "kv_cache_group": [],
764
+ "min_kv_scale": 0.0,
765
+ "pack_method": "reorder",
766
+ "weight_format": "real_quantized",
767
+ "weight_merge_groups": null
768
+ },
769
+ "global_quant_config": {
770
+ "bias": null,
771
+ "input_tensors": [
772
+ {
773
+ "block_size": null,
774
+ "ch_axis": -1,
775
+ "dtype": "fp4",
776
+ "enable_buffer_reuse": false,
777
+ "group_size": 16,
778
+ "is_dynamic": true,
779
+ "is_scale_quant": false,
780
+ "max_input_numel": 4194304,
781
+ "mx_element_dtype": null,
782
+ "observer_cls": "PerBlockMXObserver",
783
+ "qscheme": "per_group",
784
+ "round_method": "half_even",
785
+ "scale_calculation_mode": null,
786
+ "scale_format": "float32",
787
+ "scale_type": "float32",
788
+ "symmetric": null
789
+ },
790
+ {
791
+ "block_size": null,
792
+ "ch_axis": null,
793
+ "dtype": "fp8_e4m3",
794
+ "enable_buffer_reuse": false,
795
+ "group_size": null,
796
+ "is_dynamic": false,
797
+ "is_scale_quant": true,
798
+ "max_input_numel": 4194304,
799
+ "mx_element_dtype": null,
800
+ "observer_cls": "PerTensorMinMaxObserver",
801
+ "qscheme": "per_tensor",
802
+ "round_method": "half_even",
803
+ "scale_calculation_mode": null,
804
+ "scale_format": null,
805
+ "scale_type": "float32",
806
+ "symmetric": true
807
+ }
808
+ ],
809
+ "output_tensors": null,
810
+ "target_device": null,
811
+ "weight": [
812
+ {
813
+ "block_size": null,
814
+ "ch_axis": -1,
815
+ "dtype": "fp4",
816
+ "enable_buffer_reuse": false,
817
+ "group_size": 16,
818
+ "is_dynamic": false,
819
+ "is_scale_quant": false,
820
+ "max_input_numel": 4194304,
821
+ "mx_element_dtype": null,
822
+ "observer_cls": "PerBlockMXObserver",
823
+ "qscheme": "per_group",
824
+ "round_method": "half_even",
825
+ "scale_calculation_mode": null,
826
+ "scale_format": "float32",
827
+ "scale_type": "float32",
828
+ "symmetric": null
829
+ },
830
+ {
831
+ "block_size": null,
832
+ "ch_axis": null,
833
+ "dtype": "fp8_e4m3",
834
+ "enable_buffer_reuse": false,
835
+ "group_size": null,
836
+ "is_dynamic": false,
837
+ "is_scale_quant": true,
838
+ "max_input_numel": 4194304,
839
+ "mx_element_dtype": null,
840
+ "observer_cls": "PerTensorMinMaxObserver",
841
+ "qscheme": "per_tensor",
842
+ "round_method": "half_even",
843
+ "scale_calculation_mode": null,
844
+ "scale_format": null,
845
+ "scale_type": "float32",
846
+ "symmetric": true
847
+ }
848
+ ]
849
+ },
850
+ "kv_cache_post_rope": false,
851
+ "kv_cache_quant_config": {},
852
+ "layer_quant_config": {},
853
+ "layer_type_quant_config": {},
854
+ "quant_method": "quark",
855
+ "quant_mode": "eager_mode",
856
+ "softmax_quant_spec": null,
857
+ "version": "0.12+bfe174cea91"
858
+ },
859
+ "rms_norm_eps": 1e-05,
860
+ "rope_interleave": true,
861
+ "rope_parameters": {
862
+ "rope_theta": 1000000,
863
+ "rope_type": "default"
864
+ },
865
+ "routed_scaling_factor": 2.5,
866
+ "scoring_func": "sigmoid",
867
+ "tie_word_embeddings": false,
868
+ "topk_group": 1,
869
+ "topk_method": "noaux_tc",
870
+ "transformers_version": "5.4.0",
871
+ "use_cache": true,
872
+ "v_head_dim": 256,
873
+ "vocab_size": 154880
874
+ }
generation_config.json ADDED
@@ -0,0 +1,13 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "do_sample": true,
4
+ "eos_token_id": [
5
+ 154820,
6
+ 154827,
7
+ 154829
8
+ ],
9
+ "pad_token_id": 154820,
10
+ "temperature": 1.0,
11
+ "top_p": 0.95,
12
+ "transformers_version": "5.4.0"
13
+ }
model-00001-of-00009.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7eb7ff455bc396cb93f3f0006d0b893bca2ab24a488936489c7ed2b9a437b6bc
3
+ size 49999051524
model-00002-of-00009.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:37b540c98269297f48afb26badef4b29d9d0517206b8daf2543103f46db3ddfd
3
+ size 49998227584
model-00003-of-00009.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8a38862fa733644d473e6b96d9ffbdd2c1938befa71c0eb79fbd6223e0bdec75
3
+ size 50000184968
model-00004-of-00009.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:acd19130767acadebb47b7c0bb2c5063f17d65dab7a5eebeb584a2b3d9b0f4cf
3
+ size 49998227296
model-00005-of-00009.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f13eda43abc898fa9adb9727b0bb470fe3326b42fd32da59d5338422cfe64d4e
3
+ size 49916961116
model-00006-of-00009.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5d62da9cf4f135006074ad75dd952f9888e42648181855b388873c210d815d51
3
+ size 50002802316
model-00007-of-00009.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:75fc66a1007b11486180c2efd68243c6ab615e21fceb5b971bbcfcc0e279af9a
3
+ size 49999014384
model-00008-of-00009.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6c5fa30f6e744f74c25655958d3c284bd4018b869df00ce13e752e297307fb63
3
+ size 50000185032
model-00009-of-00009.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e680692d2efe667ef366badbe661962909ab605e5fb5e0c36620e49fa0a1c11e
3
+ size 42001612804
model.safetensors.index.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c7553414afe173015b77b566daae8680f2f1ed8f6840dc79f09e28d436d82770
3
+ size 22332014
quantize_20260507_114729.log 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 sha256:47757b9678da19e468edb3ae37a853996599945b5006914e5b088aff30002386
3
+ size 20217707
tokenizer_config.json ADDED
@@ -0,0 +1,33 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "backend": "tokenizers",
3
+ "clean_up_tokenization_spaces": false,
4
+ "do_lower_case": false,
5
+ "eos_token": "<|endoftext|>",
6
+ "extra_special_tokens": [
7
+ "<|endoftext|>",
8
+ "[MASK]",
9
+ "[gMASK]",
10
+ "[sMASK]",
11
+ "<sop>",
12
+ "<eop>",
13
+ "<|system|>",
14
+ "<|user|>",
15
+ "<|assistant|>",
16
+ "<|observation|>",
17
+ "<|begin_of_image|>",
18
+ "<|end_of_image|>",
19
+ "<|begin_of_video|>",
20
+ "<|end_of_video|>",
21
+ "<|begin_of_audio|>",
22
+ "<|end_of_audio|>",
23
+ "<|begin_of_transcription|>",
24
+ "<|end_of_transcription|>"
25
+ ],
26
+ "is_local": true,
27
+ "model_max_length": 202752,
28
+ "model_specific_special_tokens": {},
29
+ "pad_token": "<|endoftext|>",
30
+ "padding_side": "left",
31
+ "remove_space": false,
32
+ "tokenizer_class": "TokenizersBackend"
33
+ }