Hugging Face's logo

poolside
/
Laguna-M.1-FP8

Text Generation
Safetensors
vllm
laguna
laguna-m.1
sglang
fp8
Mixture of Experts
conversational
custom_code
compressed-tensors
Model card Files
xet
 Community
5
joerowell commited on
Commit
a7228e8
·
verified ·
1 Parent(s): 12423df

Add files using upload-large-folder tool

Browse files
Files changed (50) hide show
  1. chat_template.jinja +132 -0
  2. config.json +527 -0
  3. configuration_laguna.py +186 -0
  4. generation_config.json +12 -0
  5. model-00001-of-00045.safetensors +3 -0
  6. model-00002-of-00045.safetensors +3 -0
  7. model-00003-of-00045.safetensors +3 -0
  8. model-00004-of-00045.safetensors +3 -0
  9. model-00005-of-00045.safetensors +3 -0
  10. model-00006-of-00045.safetensors +3 -0
  11. model-00007-of-00045.safetensors +3 -0
  12. model-00008-of-00045.safetensors +3 -0
  13. model-00009-of-00045.safetensors +3 -0
  14. model-00010-of-00045.safetensors +3 -0
  15. model-00011-of-00045.safetensors +3 -0
  16. model-00012-of-00045.safetensors +3 -0
  17. model-00013-of-00045.safetensors +3 -0
  18. model-00014-of-00045.safetensors +3 -0
  19. model-00015-of-00045.safetensors +3 -0
  20. model-00016-of-00045.safetensors +3 -0
  21. model-00017-of-00045.safetensors +3 -0
  22. model-00018-of-00045.safetensors +3 -0
  23. model-00019-of-00045.safetensors +3 -0
  24. model-00020-of-00045.safetensors +3 -0
  25. model-00021-of-00045.safetensors +3 -0
  26. model-00022-of-00045.safetensors +3 -0
  27. model-00023-of-00045.safetensors +3 -0
  28. model-00024-of-00045.safetensors +3 -0
  29. model-00025-of-00045.safetensors +3 -0
  30. model-00026-of-00045.safetensors +3 -0
  31. model-00027-of-00045.safetensors +3 -0
  32. model-00028-of-00045.safetensors +3 -0
  33. model-00029-of-00045.safetensors +3 -0
  34. model-00030-of-00045.safetensors +3 -0
  35. model-00031-of-00045.safetensors +3 -0
  36. model-00032-of-00045.safetensors +3 -0
  37. model-00033-of-00045.safetensors +3 -0
  38. model-00034-of-00045.safetensors +3 -0
  39. model-00035-of-00045.safetensors +3 -0
  40. model-00036-of-00045.safetensors +3 -0
  41. model-00037-of-00045.safetensors +3 -0
  42. model-00038-of-00045.safetensors +3 -0
  43. model-00039-of-00045.safetensors +3 -0
  44. model-00040-of-00045.safetensors +3 -0
  45. model-00041-of-00045.safetensors +3 -0
  46. model-00042-of-00045.safetensors +3 -0
  47. model-00043-of-00045.safetensors +3 -0
  48. model-00044-of-00045.safetensors +3 -0
  49. model-00045-of-00045.safetensors +3 -0
  50. model-kv_scales.safetensors +3 -0
chat_template.jinja ADDED
@@ -0,0 +1,132 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {#- Copied from laguna_glm_thinking_v4/chat_template.jinja -#}
2
+ {#- Removes prefix that references <think> token, and replaces message.reasoning_content reference with message.reasoning -#}
3
+ {{- "〈|EOS|〉" -}}
4
+ {%- set enable_thinking = enable_thinking | default(false) -%}
5
+ {%- set render_assistant_messages_raw = render_assistant_messages_raw | default(false) -%}
6
+ {%- set add_generation_prompt = add_generation_prompt | default(false) -%}
7
+
8
+ {#- ───── header (system message) ───── -#}
9
+ {%- set system_message = "" -%}
10
+ {%- if messages and messages[0].role == "system" -%}
11
+ {%- set system_message = messages[0].content -%}
12
+ {%- endif -%}
13
+
14
+ {%- if (system_message and system_message.strip()) or tools -%}
15
+ {{- "<system>\n" -}}
16
+
17
+ {%- if system_message and system_message.strip() -%}
18
+ {{- "\n" -}}
19
+ {{- system_message.rstrip() -}}
20
+ {%- endif -%}
21
+
22
+ {%- if tools -%}
23
+ {{- "\n\n### Tools\n\n" -}}
24
+ {%- set ns = namespace(tool_string="You may call functions to assist with the user query.\n"
25
+ ~ "All available function signatures are listed below:\n"
26
+ ~ "<available_tools>\n") -%}
27
+ {%- for tool in tools -%}
28
+ {%- set ns.tool_string = ns.tool_string ~ (tool | tojson) ~ "\n" -%}
29
+ {%- endfor -%}
30
+ {%- if enable_thinking -%}
31
+ {%- set tool_string = ns.tool_string + "</available_tools>\n\n" ~
32
+ "Wrap your thinking in '<think>', '</think>' tags, followed by a function call. For each function call, return an unescaped XML-like object with function name and arguments within '<tool_call>' and '</tool_call>' tags, like here:\n" ~
33
+ "<think> your thoughts here </think>\n" ~
34
+ "<tool_call>function-name\n<arg_key>argument-key</arg_key>\n<arg_value>value-of-argument-key</arg_value>\n" ~
35
+ "</tool_call>" -%}
36
+ {%- else -%}
37
+ {%- set tool_string = ns.tool_string + "</available_tools>\n\n" ~
38
+ "For each function call, return an unescaped XML-like object " ~
39
+ "with function name and arguments within '<tool_call>' and '</tool_call>' tags, like here:\n" ~
40
+ "<tool_call>function-name\n<arg_key>argument-key</arg_key>\n<arg_value>value-of-argument-key</arg_value>\n" ~
41
+ "</tool_call>" -%}
42
+ {%- endif -%}
43
+ {{- tool_string -}}
44
+ {%- endif -%}
45
+
46
+ {{- "\n</system>\n" -}}
47
+ {%- endif -%}
48
+
49
+ {#- ───── main loop ───── -#}
50
+ {%- for message in messages -%}
51
+ {%- set content = message.content if message.content is string else "" -%}
52
+ {%- if message.role == "user" -%}
53
+ {{- "<user>\n" + content + "\n</user>\n" -}}
54
+ {%- elif message.role == "assistant" -%}
55
+ {%- generation -%}
56
+ {{- "<assistant>\n" -}}
57
+ {%- if render_assistant_messages_raw -%}
58
+ {#- Raw mode: prepend the generation prompt token, then dump content verbatim. -#}
59
+ {#- The generation prompt is <think> when enable_thinking, </think> otherwise. -#}
60
+ {#- Only prepend if content doesn't already start with it. -#}
61
+ {%- if enable_thinking -%}
62
+ {%- if not content.startswith('<think>') -%}
63
+ {{- '<think>' -}}
64
+ {%- endif -%}
65
+ {%- else -%}
66
+ {%- if not content.startswith('</think>') -%}
67
+ {{- '</think>' -}}
68
+ {%- endif -%}
69
+ {%- endif -%}
70
+ {{- content -}}
71
+ {#- Append closing tag if content doesn't already end with it. -#}
72
+ {%- if not content.endswith('</assistant>\n') and not content.endswith('</assistant>') -%}
73
+ {{- '\n</assistant>' -}}
74
+ {%- endif -%}
75
+ {{- "\n" -}}
76
+ {%- else -%}
77
+ {#- Extract reasoning content from message.reasoning (vLLM field name) or message.reasoning_content, or from <think> tags -#}
78
+ {%- set reasoning_content = '' %}
79
+ {%- if message.reasoning is string %}
80
+ {%- set reasoning_content = message.reasoning %}
81
+ {%- elif message.reasoning_content is string %}
82
+ {%- set reasoning_content = message.reasoning_content %}
83
+ {%- endif %}
84
+ {#- Always strip <think> tags from content if present to avoid duplication -#}
85
+ {%- if '</think>' in content %}
86
+ {%- if not reasoning_content %}
87
+ {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
88
+ {%- endif %}
89
+ {%- set content = content.split('</think>')[-1].lstrip('\n') %}
90
+ {%- endif %}
91
+ {#- Display reasoning content for all messages -#}
92
+ {%- if reasoning_content -%}
93
+ {{- '<think>\n' + reasoning_content.strip() + '\n</think>\n' -}}
94
+ {%- else -%}
95
+ {{- '</think>\n' -}}
96
+ {%- endif -%}
97
+ {#- Display main content -#}
98
+ {%- if content.strip() -%}
99
+ {{- content.strip() ~ "\n" -}}
100
+ {%- endif -%}
101
+ {%- if message.tool_calls -%}
102
+ {%- for tool_call in message.tool_calls -%}
103
+ {%- set function_data = tool_call.function -%}
104
+ {{- '<tool_call>' + function_data.name }}
105
+ {% set _args = function_data.arguments %}
106
+ {%- for k, v in _args.items() -%}
107
+ {{- "<arg_key>" ~ k ~ "</arg_key>\n" -}}
108
+ {{- "<arg_value>"}}{{ v | tojson(ensure_ascii=False) if v is not string else v }}{{ "</arg_value>\n" -}}
109
+ {%- endfor -%}
110
+ {{- "</tool_call>\n" -}}
111
+ {%- endfor -%}
112
+ {%- endif -%}
113
+ {{- "</assistant>\n" -}}
114
+ {%- endif -%}
115
+ {%- endgeneration -%}
116
+ {%- elif message.role == "tool" -%}
117
+ {{- "<tool_response>\n" + content + "\n</tool_response>\n" -}}
118
+ {%- elif message.role == "system" and loop.index0 != 0 -%}
119
+ {#- Render additional system messages (skip the first one which is handled separately in the header) -#}
120
+ {{- "<system>\n" + content + "\n</system>\n" -}}
121
+ {%- endif -%}
122
+ {%- endfor -%}
123
+ {#- ───── generation prompt ───── -#}
124
+ {%- if add_generation_prompt -%}
125
+ {{- "<assistant>\n" -}}
126
+ {#- ───── Include reasoning mode directive ───── -#}
127
+ {%- if not enable_thinking %}
128
+ {{- '</think>' -}}
129
+ {%- else %}
130
+ {{- '<think>' -}}
131
+ {%- endif %}
132
+ {%- endif -%}
config.json ADDED
@@ -0,0 +1,527 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "LagunaForCausalLM"
4
+ ],
5
+ "model_type": "laguna",
6
+ "vocab_size": 100352,
7
+ "hidden_size": 4096,
8
+ "intermediate_size": 16384,
9
+ "num_hidden_layers": 70,
10
+ "num_attention_heads": 64,
11
+ "num_key_value_heads": 8,
12
+ "head_dim": 128,
13
+ "max_position_embeddings": 131072,
14
+ "attention_bias": false,
15
+ "attention_dropout": 0.0,
16
+ "rms_norm_eps": 1e-06,
17
+ "num_experts": 256,
18
+ "num_experts_per_tok": 16,
19
+ "moe_intermediate_size": 1024,
20
+ "shared_expert_intermediate_size": 1024,
21
+ "router_aux_loss_coef": 0.0,
22
+ "bos_token_id": 2,
23
+ "eos_token_id": [
24
+ 2,
25
+ 24
26
+ ],
27
+ "pad_token_id": 9,
28
+ "tie_word_embeddings": false,
29
+ "use_cache": true,
30
+ "torch_dtype": "bfloat16",
31
+ "gating": "per-element",
32
+ "sliding_window": null,
33
+ "rope_parameters": {
34
+ "full_attention": {
35
+ "rope_theta": 500000.0,
36
+ "rope_type": "yarn",
37
+ "factor": 32.0,
38
+ "original_max_position_embeddings": 4096,
39
+ "beta_slow": 1.0,
40
+ "beta_fast": 64.0,
41
+ "attention_factor": 1.0
42
+ }
43
+ },
44
+ "moe_apply_router_weight_on_input": false,
45
+ "partial_rotary_factor": 1.0,
46
+ "mlp_layer_types": [
47
+ "dense",
48
+ "dense",
49
+ "dense",
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
+ "sparse",
106
+ "sparse",
107
+ "sparse",
108
+ "sparse",
109
+ "sparse",
110
+ "sparse",
111
+ "sparse",
112
+ "sparse",
113
+ "sparse",
114
+ "sparse",
115
+ "sparse",
116
+ "sparse"
117
+ ],
118
+ "use_bidirectional_attention": false,
119
+ "moe_routed_scaling_factor": 1.0,
120
+ "compression_config": {
121
+ "config_groups": {
122
+ "group_0": {
123
+ "format": "float-quantized",
124
+ "input_activations": {
125
+ "actorder": null,
126
+ "block_structure": null,
127
+ "dynamic": true,
128
+ "group_size": 128,
129
+ "num_bits": 8,
130
+ "observer": null,
131
+ "observer_kwargs": {},
132
+ "strategy": "group",
133
+ "symmetric": true,
134
+ "type": "float"
135
+ },
136
+ "output_activations": null,
137
+ "targets": [
138
+ "Linear"
139
+ ],
140
+ "weights": {
141
+ "actorder": null,
142
+ "block_structure": [
143
+ 128,
144
+ 128
145
+ ],
146
+ "dynamic": false,
147
+ "group_size": null,
148
+ "num_bits": 8,
149
+ "observer": "minmax",
150
+ "observer_kwargs": {},
151
+ "strategy": "block",
152
+ "symmetric": true,
153
+ "type": "float"
154
+ }
155
+ }
156
+ },
157
+ "format": "float-quantized",
158
+ "global_compression_ratio": null,
159
+ "ignore": [
160
+ "model.layers.0.self_attn.o_proj",
161
+ "model.layers.1.self_attn.o_proj",
162
+ "model.layers.2.self_attn.o_proj",
163
+ "model.layers.3.self_attn.o_proj",
164
+ "model.layers.3.mlp.gate",
165
+ "model.layers.4.self_attn.o_proj",
166
+ "model.layers.4.mlp.gate",
167
+ "model.layers.5.self_attn.o_proj",
168
+ "model.layers.5.mlp.gate",
169
+ "model.layers.6.self_attn.o_proj",
170
+ "model.layers.6.mlp.gate",
171
+ "model.layers.7.self_attn.o_proj",
172
+ "model.layers.7.mlp.gate",
173
+ "model.layers.8.self_attn.o_proj",
174
+ "model.layers.8.mlp.gate",
175
+ "model.layers.9.self_attn.o_proj",
176
+ "model.layers.9.mlp.gate",
177
+ "model.layers.10.self_attn.o_proj",
178
+ "model.layers.10.mlp.gate",
179
+ "model.layers.11.self_attn.o_proj",
180
+ "model.layers.11.mlp.gate",
181
+ "model.layers.12.self_attn.o_proj",
182
+ "model.layers.12.mlp.gate",
183
+ "model.layers.13.self_attn.o_proj",
184
+ "model.layers.13.mlp.gate",
185
+ "model.layers.14.self_attn.o_proj",
186
+ "model.layers.14.mlp.gate",
187
+ "model.layers.15.self_attn.o_proj",
188
+ "model.layers.15.mlp.gate",
189
+ "model.layers.16.self_attn.o_proj",
190
+ "model.layers.16.mlp.gate",
191
+ "model.layers.17.self_attn.o_proj",
192
+ "model.layers.17.mlp.gate",
193
+ "model.layers.18.self_attn.o_proj",
194
+ "model.layers.18.mlp.gate",
195
+ "model.layers.19.self_attn.o_proj",
196
+ "model.layers.19.mlp.gate",
197
+ "model.layers.20.self_attn.o_proj",
198
+ "model.layers.20.mlp.gate",
199
+ "model.layers.21.self_attn.o_proj",
200
+ "model.layers.21.mlp.gate",
201
+ "model.layers.22.self_attn.o_proj",
202
+ "model.layers.22.mlp.gate",
203
+ "model.layers.23.self_attn.o_proj",
204
+ "model.layers.23.mlp.gate",
205
+ "model.layers.24.self_attn.o_proj",
206
+ "model.layers.24.mlp.gate",
207
+ "model.layers.25.self_attn.o_proj",
208
+ "model.layers.25.mlp.gate",
209
+ "model.layers.26.self_attn.o_proj",
210
+ "model.layers.26.mlp.gate",
211
+ "model.layers.27.self_attn.o_proj",
212
+ "model.layers.27.mlp.gate",
213
+ "model.layers.28.self_attn.o_proj",
214
+ "model.layers.28.mlp.gate",
215
+ "model.layers.29.self_attn.o_proj",
216
+ "model.layers.29.mlp.gate",
217
+ "model.layers.30.self_attn.o_proj",
218
+ "model.layers.30.mlp.gate",
219
+ "model.layers.31.self_attn.o_proj",
220
+ "model.layers.31.mlp.gate",
221
+ "model.layers.32.self_attn.o_proj",
222
+ "model.layers.32.mlp.gate",
223
+ "model.layers.33.self_attn.o_proj",
224
+ "model.layers.33.mlp.gate",
225
+ "model.layers.34.self_attn.o_proj",
226
+ "model.layers.34.mlp.gate",
227
+ "model.layers.35.self_attn.o_proj",
228
+ "model.layers.35.mlp.gate",
229
+ "model.layers.36.self_attn.o_proj",
230
+ "model.layers.36.mlp.gate",
231
+ "model.layers.37.self_attn.o_proj",
232
+ "model.layers.37.mlp.gate",
233
+ "model.layers.38.self_attn.o_proj",
234
+ "model.layers.38.mlp.gate",
235
+ "model.layers.39.self_attn.o_proj",
236
+ "model.layers.39.mlp.gate",
237
+ "model.layers.40.self_attn.o_proj",
238
+ "model.layers.40.mlp.gate",
239
+ "model.layers.41.self_attn.o_proj",
240
+ "model.layers.41.mlp.gate",
241
+ "model.layers.42.self_attn.o_proj",
242
+ "model.layers.42.mlp.gate",
243
+ "model.layers.43.self_attn.o_proj",
244
+ "model.layers.43.mlp.gate",
245
+ "model.layers.44.self_attn.o_proj",
246
+ "model.layers.44.mlp.gate",
247
+ "model.layers.45.self_attn.o_proj",
248
+ "model.layers.45.mlp.gate",
249
+ "model.layers.46.self_attn.o_proj",
250
+ "model.layers.46.mlp.gate",
251
+ "model.layers.47.self_attn.o_proj",
252
+ "model.layers.47.mlp.gate",
253
+ "model.layers.48.self_attn.o_proj",
254
+ "model.layers.48.mlp.gate",
255
+ "model.layers.49.self_attn.o_proj",
256
+ "model.layers.49.mlp.gate",
257
+ "model.layers.50.self_attn.o_proj",
258
+ "model.layers.50.mlp.gate",
259
+ "model.layers.51.self_attn.o_proj",
260
+ "model.layers.51.mlp.gate",
261
+ "model.layers.52.self_attn.o_proj",
262
+ "model.layers.52.mlp.gate",
263
+ "model.layers.53.self_attn.o_proj",
264
+ "model.layers.53.mlp.gate",
265
+ "model.layers.54.self_attn.o_proj",
266
+ "model.layers.54.mlp.gate",
267
+ "model.layers.55.self_attn.o_proj",
268
+ "model.layers.55.mlp.gate",
269
+ "model.layers.56.self_attn.o_proj",
270
+ "model.layers.56.mlp.gate",
271
+ "model.layers.57.self_attn.o_proj",
272
+ "model.layers.57.mlp.gate",
273
+ "model.layers.58.self_attn.o_proj",
274
+ "model.layers.58.mlp.gate",
275
+ "model.layers.59.self_attn.o_proj",
276
+ "model.layers.59.mlp.gate",
277
+ "model.layers.60.self_attn.o_proj",
278
+ "model.layers.60.mlp.gate",
279
+ "model.layers.61.self_attn.o_proj",
280
+ "model.layers.61.mlp.gate",
281
+ "model.layers.62.self_attn.o_proj",
282
+ "model.layers.62.mlp.gate",
283
+ "model.layers.63.self_attn.o_proj",
284
+ "model.layers.63.mlp.gate",
285
+ "model.layers.64.self_attn.o_proj",
286
+ "model.layers.64.mlp.gate",
287
+ "model.layers.65.self_attn.o_proj",
288
+ "model.layers.65.mlp.gate",
289
+ "model.layers.66.self_attn.o_proj",
290
+ "model.layers.66.mlp.gate",
291
+ "model.layers.67.self_attn.o_proj",
292
+ "model.layers.67.mlp.gate",
293
+ "model.layers.68.self_attn.o_proj",
294
+ "model.layers.68.mlp.gate",
295
+ "model.layers.69.self_attn.o_proj",
296
+ "model.layers.69.mlp.gate",
297
+ "lm_head"
298
+ ],
299
+ "kv_cache_scheme": {
300
+ "actorder": null,
301
+ "block_structure": null,
302
+ "dynamic": false,
303
+ "group_size": null,
304
+ "num_bits": 8,
305
+ "observer": "minmax",
306
+ "observer_kwargs": {},
307
+ "strategy": "tensor",
308
+ "symmetric": true,
309
+ "type": "float"
310
+ },
311
+ "quant_method": "compressed-tensors",
312
+ "quantization_status": "compressed",
313
+ "sparsity_config": {},
314
+ "transform_config": {},
315
+ "version": "0.11.0"
316
+ },
317
+ "quantization_config": {
318
+ "config_groups": {
319
+ "group_0": {
320
+ "format": "float-quantized",
321
+ "input_activations": {
322
+ "actorder": null,
323
+ "block_structure": null,
324
+ "dynamic": true,
325
+ "group_size": 128,
326
+ "num_bits": 8,
327
+ "observer": null,
328
+ "observer_kwargs": {},
329
+ "strategy": "group",
330
+ "symmetric": true,
331
+ "type": "float"
332
+ },
333
+ "output_activations": null,
334
+ "targets": [
335
+ "Linear"
336
+ ],
337
+ "weights": {
338
+ "actorder": null,
339
+ "block_structure": [
340
+ 128,
341
+ 128
342
+ ],
343
+ "dynamic": false,
344
+ "group_size": null,
345
+ "num_bits": 8,
346
+ "observer": "minmax",
347
+ "observer_kwargs": {},
348
+ "strategy": "block",
349
+ "symmetric": true,
350
+ "type": "float"
351
+ }
352
+ }
353
+ },
354
+ "format": "float-quantized",
355
+ "global_compression_ratio": null,
356
+ "ignore": [
357
+ "model.layers.0.self_attn.o_proj",
358
+ "model.layers.1.self_attn.o_proj",
359
+ "model.layers.2.self_attn.o_proj",
360
+ "model.layers.3.self_attn.o_proj",
361
+ "model.layers.3.mlp.gate",
362
+ "model.layers.4.self_attn.o_proj",
363
+ "model.layers.4.mlp.gate",
364
+ "model.layers.5.self_attn.o_proj",
365
+ "model.layers.5.mlp.gate",
366
+ "model.layers.6.self_attn.o_proj",
367
+ "model.layers.6.mlp.gate",
368
+ "model.layers.7.self_attn.o_proj",
369
+ "model.layers.7.mlp.gate",
370
+ "model.layers.8.self_attn.o_proj",
371
+ "model.layers.8.mlp.gate",
372
+ "model.layers.9.self_attn.o_proj",
373
+ "model.layers.9.mlp.gate",
374
+ "model.layers.10.self_attn.o_proj",
375
+ "model.layers.10.mlp.gate",
376
+ "model.layers.11.self_attn.o_proj",
377
+ "model.layers.11.mlp.gate",
378
+ "model.layers.12.self_attn.o_proj",
379
+ "model.layers.12.mlp.gate",
380
+ "model.layers.13.self_attn.o_proj",
381
+ "model.layers.13.mlp.gate",
382
+ "model.layers.14.self_attn.o_proj",
383
+ "model.layers.14.mlp.gate",
384
+ "model.layers.15.self_attn.o_proj",
385
+ "model.layers.15.mlp.gate",
386
+ "model.layers.16.self_attn.o_proj",
387
+ "model.layers.16.mlp.gate",
388
+ "model.layers.17.self_attn.o_proj",
389
+ "model.layers.17.mlp.gate",
390
+ "model.layers.18.self_attn.o_proj",
391
+ "model.layers.18.mlp.gate",
392
+ "model.layers.19.self_attn.o_proj",
393
+ "model.layers.19.mlp.gate",
394
+ "model.layers.20.self_attn.o_proj",
395
+ "model.layers.20.mlp.gate",
396
+ "model.layers.21.self_attn.o_proj",
397
+ "model.layers.21.mlp.gate",
398
+ "model.layers.22.self_attn.o_proj",
399
+ "model.layers.22.mlp.gate",
400
+ "model.layers.23.self_attn.o_proj",
401
+ "model.layers.23.mlp.gate",
402
+ "model.layers.24.self_attn.o_proj",
403
+ "model.layers.24.mlp.gate",
404
+ "model.layers.25.self_attn.o_proj",
405
+ "model.layers.25.mlp.gate",
406
+ "model.layers.26.self_attn.o_proj",
407
+ "model.layers.26.mlp.gate",
408
+ "model.layers.27.self_attn.o_proj",
409
+ "model.layers.27.mlp.gate",
410
+ "model.layers.28.self_attn.o_proj",
411
+ "model.layers.28.mlp.gate",
412
+ "model.layers.29.self_attn.o_proj",
413
+ "model.layers.29.mlp.gate",
414
+ "model.layers.30.self_attn.o_proj",
415
+ "model.layers.30.mlp.gate",
416
+ "model.layers.31.self_attn.o_proj",
417
+ "model.layers.31.mlp.gate",
418
+ "model.layers.32.self_attn.o_proj",
419
+ "model.layers.32.mlp.gate",
420
+ "model.layers.33.self_attn.o_proj",
421
+ "model.layers.33.mlp.gate",
422
+ "model.layers.34.self_attn.o_proj",
423
+ "model.layers.34.mlp.gate",
424
+ "model.layers.35.self_attn.o_proj",
425
+ "model.layers.35.mlp.gate",
426
+ "model.layers.36.self_attn.o_proj",
427
+ "model.layers.36.mlp.gate",
428
+ "model.layers.37.self_attn.o_proj",
429
+ "model.layers.37.mlp.gate",
430
+ "model.layers.38.self_attn.o_proj",
431
+ "model.layers.38.mlp.gate",
432
+ "model.layers.39.self_attn.o_proj",
433
+ "model.layers.39.mlp.gate",
434
+ "model.layers.40.self_attn.o_proj",
435
+ "model.layers.40.mlp.gate",
436
+ "model.layers.41.self_attn.o_proj",
437
+ "model.layers.41.mlp.gate",
438
+ "model.layers.42.self_attn.o_proj",
439
+ "model.layers.42.mlp.gate",
440
+ "model.layers.43.self_attn.o_proj",
441
+ "model.layers.43.mlp.gate",
442
+ "model.layers.44.self_attn.o_proj",
443
+ "model.layers.44.mlp.gate",
444
+ "model.layers.45.self_attn.o_proj",
445
+ "model.layers.45.mlp.gate",
446
+ "model.layers.46.self_attn.o_proj",
447
+ "model.layers.46.mlp.gate",
448
+ "model.layers.47.self_attn.o_proj",
449
+ "model.layers.47.mlp.gate",
450
+ "model.layers.48.self_attn.o_proj",
451
+ "model.layers.48.mlp.gate",
452
+ "model.layers.49.self_attn.o_proj",
453
+ "model.layers.49.mlp.gate",
454
+ "model.layers.50.self_attn.o_proj",
455
+ "model.layers.50.mlp.gate",
456
+ "model.layers.51.self_attn.o_proj",
457
+ "model.layers.51.mlp.gate",
458
+ "model.layers.52.self_attn.o_proj",
459
+ "model.layers.52.mlp.gate",
460
+ "model.layers.53.self_attn.o_proj",
461
+ "model.layers.53.mlp.gate",
462
+ "model.layers.54.self_attn.o_proj",
463
+ "model.layers.54.mlp.gate",
464
+ "model.layers.55.self_attn.o_proj",
465
+ "model.layers.55.mlp.gate",
466
+ "model.layers.56.self_attn.o_proj",
467
+ "model.layers.56.mlp.gate",
468
+ "model.layers.57.self_attn.o_proj",
469
+ "model.layers.57.mlp.gate",
470
+ "model.layers.58.self_attn.o_proj",
471
+ "model.layers.58.mlp.gate",
472
+ "model.layers.59.self_attn.o_proj",
473
+ "model.layers.59.mlp.gate",
474
+ "model.layers.60.self_attn.o_proj",
475
+ "model.layers.60.mlp.gate",
476
+ "model.layers.61.self_attn.o_proj",
477
+ "model.layers.61.mlp.gate",
478
+ "model.layers.62.self_attn.o_proj",
479
+ "model.layers.62.mlp.gate",
480
+ "model.layers.63.self_attn.o_proj",
481
+ "model.layers.63.mlp.gate",
482
+ "model.layers.64.self_attn.o_proj",
483
+ "model.layers.64.mlp.gate",
484
+ "model.layers.65.self_attn.o_proj",
485
+ "model.layers.65.mlp.gate",
486
+ "model.layers.66.self_attn.o_proj",
487
+ "model.layers.66.mlp.gate",
488
+ "model.layers.67.self_attn.o_proj",
489
+ "model.layers.67.mlp.gate",
490
+ "model.layers.68.self_attn.o_proj",
491
+ "model.layers.68.mlp.gate",
492
+ "model.layers.69.self_attn.o_proj",
493
+ "model.layers.69.mlp.gate",
494
+ "lm_head"
495
+ ],
496
+ "kv_cache_scheme": {
497
+ "actorder": null,
498
+ "block_structure": null,
499
+ "dynamic": false,
500
+ "group_size": null,
501
+ "num_bits": 8,
502
+ "observer": "minmax",
503
+ "observer_kwargs": {},
504
+ "strategy": "tensor",
505
+ "symmetric": true,
506
+ "type": "float"
507
+ },
508
+ "quant_method": "compressed-tensors",
509
+ "quantization_status": "compressed",
510
+ "sparsity_config": {},
511
+ "transform_config": {},
512
+ "version": "0.11.0"
513
+ },
514
+ "auto_map": {
515
+ "AutoConfig": "configuration_laguna.LagunaConfig",
516
+ "AutoModelForCausalLM": "modeling_laguna.LagunaForCausalLM"
517
+ },
518
+ "rope_theta": 500000.0,
519
+ "rope_scaling": {
520
+ "rope_type": "yarn",
521
+ "factor": 32.0,
522
+ "original_max_position_embeddings": 4096,
523
+ "beta_slow": 1.0,
524
+ "beta_fast": 64.0,
525
+ "attention_factor": 1.0
526
+ }
527
+ }
configuration_laguna.py ADDED
@@ -0,0 +1,186 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # ruff: noqa
2
+ # Copyright 2025 Poolside and the HuggingFace Inc. team. All rights reserved.
3
+ #
4
+ # Licensed under the Apache License, Version 2.0 (the "License");
5
+ # you may not use this file except in compliance with the License.
6
+ # You may obtain a copy of the License at
7
+ #
8
+ # http://www.apache.org/licenses/LICENSE-2.0
9
+ #
10
+ # Unless required by applicable law or agreed to in writing, software
11
+ # distributed under the License is distributed on an "AS IS" BASIS,
12
+ # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13
+ # See the License for the specific language governing permissions and
14
+ # limitations under the License.
15
+ """
16
+ Laguna configuration for transformers 4.56-4.x (used by vLLM).
17
+
18
+ This uses rope_theta + rope_scaling (legacy format) instead of
19
+ rope_parameters (v5 format).
20
+ """
21
+
22
+ from transformers.configuration_utils import PretrainedConfig
23
+
24
+
25
+ class LagunaConfig(PretrainedConfig):
26
+ r"""
27
+ Configuration class for Laguna model.
28
+
29
+ Laguna is Poolside's MoE architecture with:
30
+ - Attention output gating (softplus gate)
31
+ - Sigmoid routing instead of softmax
32
+ - No QKV bias
33
+ - Explicit head_dim parameter
34
+
35
+ Args:
36
+ head_dim (`int`, *optional*, defaults to 128):
37
+ Dimension of attention heads. Laguna uses explicit head_dim rather than
38
+ computing it from hidden_size // num_attention_heads.
39
+ qkv_bias (`bool`, *optional*, defaults to `False`):
40
+ Whether to add bias to QKV projections. Laguna uses no QKV bias.
41
+ attention_bias (`bool`, *optional*, defaults to `False`):
42
+ Whether to add bias to attention output projection. Laguna uses no attention bias.
43
+ gating (`bool`, *optional*, defaults to `True`):
44
+ Whether to use softplus output gating on attention. When True, a g_proj linear
45
+ layer is added and attn_output = attn_output * softplus(g_proj(x)).
46
+ vocab_size (`int`, *optional*, defaults to 100352):
47
+ Vocabulary size of the Laguna model.
48
+ hidden_size (`int`, *optional*, defaults to 2048):
49
+ Dimension of the hidden representations.
50
+ intermediate_size (`int`, *optional*, defaults to 8192):
51
+ Dimension of the MLP representations for dense layers.
52
+ num_hidden_layers (`int`, *optional*, defaults to 48):
53
+ Number of hidden layers in the Transformer.
54
+ num_attention_heads (`int`, *optional*, defaults to 32):
55
+ Number of attention heads.
56
+ num_key_value_heads (`int`, *optional*, defaults to 8):
57
+ Number of key-value heads for GQA.
58
+ max_position_embeddings (`int`, *optional*, defaults to 4096):
59
+ Maximum sequence length.
60
+ rms_norm_eps (`float`, *optional*, defaults to 1e-6):
61
+ Epsilon for RMSNorm layers.
62
+ rope_theta (`float`, *optional*, defaults to 500000.0):
63
+ Base frequency for RoPE embeddings.
64
+ rope_scaling (`dict`, *optional*):
65
+ RoPE scaling configuration (e.g. YaRN, linear).
66
+ sliding_window (`int`, *optional*):
67
+ Sliding window attention size. Used by layers whose type in ``layer_types``
68
+ is ``"sliding_attention"``. When ``None``, all layers use full attention.
69
+ layer_types (`list[str]`, *optional*):
70
+ Per-layer attention type. Each element should be ``"sliding_attention"`` or
71
+ ``"global_attention"``. Length must equal ``num_hidden_layers``. When ``None``,
72
+ all layers default to global attention.
73
+ swa_attention_sink_enabled (`bool`, *optional*, defaults to `False`):
74
+ Whether to enable learnable attention sinks on sliding-window attention layers.
75
+ num_experts (`int`, *optional*, defaults to 256):
76
+ Number of routed experts.
77
+ num_experts_per_tok (`int`, *optional*, defaults to 16):
78
+ Number of experts selected per token (top-k).
79
+ moe_intermediate_size (`int`, *optional*, defaults to 1024):
80
+ Intermediate size of routed experts.
81
+ shared_expert_intermediate_size (`int`, *optional*, defaults to 1024):
82
+ Intermediate size of the shared expert.
83
+ norm_topk_prob (`bool`, *optional*, defaults to `True`):
84
+ Whether to normalize top-k routing probabilities.
85
+ decoder_sparse_step (`int`, *optional*, defaults to 1):
86
+ Frequency of MoE layers (1 = every layer is MoE after mlp_only_layers).
87
+ mlp_only_layers (`list[int]`, *optional*, defaults to `[0]`):
88
+ Layer indices that use dense MLP instead of MoE.
89
+ router_aux_loss_coef (`float`, *optional*, defaults to 0.001):
90
+ Auxiliary loss coefficient for load balancing.
91
+ """
92
+
93
+ model_type = "laguna"
94
+ keys_to_ignore_at_inference = ["past_key_values"]
95
+ base_model_tp_plan = {
96
+ "layers.*.self_attn.q_proj": "colwise",
97
+ "layers.*.self_attn.k_proj": "colwise",
98
+ "layers.*.self_attn.v_proj": "colwise",
99
+ "layers.*.self_attn.g_proj": "colwise", # Laguna-specific gating projection
100
+ "layers.*.self_attn.o_proj": "rowwise",
101
+ "layers.*.mlp.gate_proj": "colwise",
102
+ "layers.*.mlp.up_proj": "colwise",
103
+ "layers.*.mlp.down_proj": "rowwise",
104
+ }
105
+ base_model_pp_plan = {
106
+ "embed_tokens": (["input_ids"], ["inputs_embeds"]),
107
+ "layers": (["hidden_states", "attention_mask"], ["hidden_states"]),
108
+ "norm": (["hidden_states"], ["hidden_states"]),
109
+ }
110
+
111
+ def __init__(
112
+ self,
113
+ vocab_size: int = 100352,
114
+ hidden_size: int = 2048,
115
+ intermediate_size: int = 8192,
116
+ num_hidden_layers: int = 48,
117
+ num_attention_heads: int = 32,
118
+ num_key_value_heads: int = 8,
119
+ head_dim: int = 128,
120
+ qkv_bias: bool = False,
121
+ attention_bias: bool = False,
122
+ gating: bool = True,
123
+ hidden_act: str = "silu",
124
+ max_position_embeddings: int = 4096,
125
+ initializer_range: float = 0.02,
126
+ rms_norm_eps: float = 1e-6,
127
+ use_cache: bool = True,
128
+ tie_word_embeddings: bool = False,
129
+ rope_theta: float = 500000.0,
130
+ rope_scaling: dict | None = None,
131
+ attention_dropout: float = 0.0,
132
+ sliding_window: int | None = None,
133
+ layer_types: list[str] | None = None,
134
+ swa_attention_sink_enabled: bool = False,
135
+ num_experts: int = 256,
136
+ num_experts_per_tok: int = 16,
137
+ moe_intermediate_size: int = 1024,
138
+ shared_expert_intermediate_size: int = 1024,
139
+ norm_topk_prob: bool = True,
140
+ decoder_sparse_step: int = 1,
141
+ mlp_only_layers: list[int] | None = None,
142
+ router_aux_loss_coef: float = 0.001,
143
+ output_router_logits: bool = False,
144
+ **kwargs,
145
+ ):
146
+ # Default mlp_only_layers: first layer is dense (moe_first_k_dense_replace=1)
147
+ if mlp_only_layers is None:
148
+ mlp_only_layers = [0]
149
+
150
+ self.vocab_size = vocab_size
151
+ self.hidden_size = hidden_size
152
+ self.intermediate_size = intermediate_size
153
+ self.num_hidden_layers = num_hidden_layers
154
+ self.num_attention_heads = num_attention_heads
155
+ self.num_key_value_heads = num_key_value_heads
156
+ self.head_dim = head_dim
157
+ self.qkv_bias = qkv_bias
158
+ self.attention_bias = attention_bias
159
+ self.gating = gating
160
+ self.hidden_act = hidden_act
161
+ self.max_position_embeddings = max_position_embeddings
162
+ self.initializer_range = initializer_range
163
+ self.rms_norm_eps = rms_norm_eps
164
+ self.use_cache = use_cache
165
+ self.rope_theta = rope_theta
166
+ self.rope_scaling = rope_scaling
167
+ self.attention_dropout = attention_dropout
168
+ # Sliding window attention arguments
169
+ self.sliding_window = sliding_window
170
+ self.layer_types = layer_types
171
+ self.swa_attention_sink_enabled = swa_attention_sink_enabled
172
+ # MoE arguments
173
+ self.num_experts = num_experts
174
+ self.num_experts_per_tok = num_experts_per_tok
175
+ self.moe_intermediate_size = moe_intermediate_size
176
+ self.shared_expert_intermediate_size = shared_expert_intermediate_size
177
+ self.norm_topk_prob = norm_topk_prob
178
+ self.decoder_sparse_step = decoder_sparse_step
179
+ self.mlp_only_layers = mlp_only_layers
180
+ self.router_aux_loss_coef = router_aux_loss_coef
181
+ self.output_router_logits = output_router_logits
182
+
183
+ super().__init__(tie_word_embeddings=tie_word_embeddings, **kwargs)
184
+
185
+
186
+ __all__ = ["LagunaConfig"]
generation_config.json ADDED
@@ -0,0 +1,12 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token_id": 2,
3
+ "do_sample": true,
4
+ "eos_token_id": [
5
+ 2,
6
+ 24
7
+ ],
8
+ "max_new_tokens": 2048,
9
+ "pad_token_id": 9,
10
+ "temperature": 0.7,
11
+ "top_p": 0.9
12
+ }
model-00001-of-00045.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:393a5863004d5c402c065129a1f93da263bc00501466fb34fa578cdb6903d627
3
+ size 5118019912
model-00002-of-00045.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4873cb659f6efa610cffb4ff42ff11528ce05c795ff669b7163c86512aea6599
3
+ size 5076112752
model-00003-of-00045.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:968265e5db97a1a82f110587e1fd90eef53e844dfed80bf8e2bd9eff1dc5494a
3
+ size 5119812952
model-00004-of-00045.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:eb355291dc19fbf36d0cea3eac0fc75b1ef50f13f1a248f7cd10b8dacf2576dd
3
+ size 5117975080
model-00005-of-00045.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4df412f87fc9fb72872e390281deb47c13bf2e7833d402d8bc4dd65669affd5e
3
+ size 5117975272
model-00006-of-00045.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7df42927f87d341b248253307df91a51847152f948d1f837daf3e8e51aabfed5
3
+ size 5117972952
model-00007-of-00045.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a507b691db9b6b8902a76cf45b4f67e7b355ac538df167385a0c9efbe8445489
3
+ size 5117975496
model-00008-of-00045.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a0db46e473b3723da63309f7ce342fcfc3679d7e2373e4bcaa890eda033af44a
3
+ size 5117977296
model-00009-of-00045.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1d13bb8520755eedef27b9ef8c11fd09cc437a5016977b53fc5f8e04742936b8
3
+ size 5117975712
model-00010-of-00045.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a3bc781cb8eaa228c8495c7955fa6012ddf7549532041b3817cfda166405c627
3
+ size 5117977624
model-00011-of-00045.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:aa125530b4c72714c8c1a47cbe3c354fb26970fc9bd3995b3100db76e7334298
3
+ size 5117977328
model-00012-of-00045.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:141fed6170c479b11579693229a5cac1015ad54185d7419d96342224e7248b27
3
+ size 5117976000
model-00013-of-00045.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8bc474905b40baf5f211a40f598ad64e262d87a04c9e43029291e9fddbb63b1e
3
+ size 5117977288
model-00014-of-00045.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:45f97cd0e76816fa86ad0624f27e30d14560f032262db12448df5bbc0528ef84
3
+ size 5117977688
model-00015-of-00045.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1d87e883e021bc9135a48cd6a989810c1aea623b7afc3175ef8360999d663189
3
+ size 5117975640
model-00016-of-00045.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7a7762781077993192c1f138f2fcf3515311a52d100c6e9ba7689df6845ee887
3
+ size 5117977424
model-00017-of-00045.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:344cfddb8359935bd779b18a7ab3cc9a81ff8eeefe2774763398c2e5a592a146
3
+ size 5117977808
model-00018-of-00045.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c0a8ad03f4a4eae9f7c3536d773f257f025708866a0c3749cf4799747c1e2e25
3
+ size 5117975392
model-00019-of-00045.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:79531f7918635bf9ccee16dd866dbd3d344ac5e5497478bb77bcb42a9a0ba47d
3
+ size 5117977784
model-00020-of-00045.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:190fa1ec6b12b962d6caa0be6f1dbe651d4ef9a2cc552df56cf0eee223ede72e
3
+ size 5117977448
model-00021-of-00045.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:55ad531b4dc60f10f3bd7087185a4008c91ab104e41daab44e0d131b4b766883
3
+ size 5117975616
model-00022-of-00045.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:683e93e8476675e6d87d7d860fb5b065f2b1b97d0894b58c2567cb53350e10ea
3
+ size 5117977720
model-00023-of-00045.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:64ee6bc48169f2dc0c46538776ed5296dd1001cb9b75b3d4939c4bd164b7ab44
3
+ size 5117977280
model-00024-of-00045.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4f2e8dc4be3997a55a224365c613c1462f2ea0885ca608c3084f4ca163bd3918
3
+ size 5117976024
model-00025-of-00045.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:173a21679839bbc181e9d42838905a6daba0102d9936183e62c186378a1f70f7
3
+ size 5117977312
model-00026-of-00045.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b81b4f24d2a5ae0d1b36da3a05232023fc1ac22648f17ee440368b93c513f43e
3
+ size 5117977592
model-00027-of-00045.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8ed8db746a64b39583460987dcf7837c89121fd5b503d470daa8e382f6ec809e
3
+ size 5117975736
model-00028-of-00045.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:897af0aa64c57ee745e7755cccf44186cacd2c126079a5ed7ba128f779b1b4af
3
+ size 5117977328
model-00029-of-00045.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4547844b8f2459614ec1827a6c6c63a6b475b9985976d6ba8ceff42707712255
3
+ size 5117977904
model-00030-of-00045.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0b319bc30d56166272c24fd940db154abc7d2cfe28114881132dd2cd4c6c371f
3
+ size 5117975392
model-00031-of-00045.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ce98d325af4afca16ee57fbfb00f99f9fe21dc18b737e397c35dca31355f6bb8
3
+ size 5117977688
model-00032-of-00045.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:92be0d5a3c99f5747811535823082d229ab7761b5c37209da7d1376e691e30cb
3
+ size 5117977544
model-00033-of-00045.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3b6864788416250c01dd7ce310be952ced6c7a0879fce037afb5ed5e5a81eaba
3
+ size 5117975520
model-00034-of-00045.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e1871a96b175a0b7cd07634699507ca0c371a963d6ed1bad04605ae659dc3529
3
+ size 5117977816
model-00035-of-00045.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:705d2d0d053ecd855d992a8375ff6a2c174a0a4a6447d8a2d7bed47594644e5e
3
+ size 5117977280
model-00036-of-00045.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9a282fc251fc7af44d6b5740956f4cb260790f16b8d5bb9ab1b7bb6552ecd16d
3
+ size 5117975880
model-00037-of-00045.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:66d9273a4b70096c7b744537003f69b10318766159c1cd1cf8110a0c45bdd39f
3
+ size 5117977456
model-00038-of-00045.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e5dd3b670462fe6541fd3adf3efc418dcd6cc01d08e91c4f71308e431f0e923f
3
+ size 5117977496
model-00039-of-00045.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9535270c4e3b6f44e9a4ce22211a95ad2efb7ef506ebb353a690f743957fe03a
3
+ size 5117975832
model-00040-of-00045.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ab1738642cb5acd0588885fd37744ff2ceccc5a510850b97e94ae968f4844507
3
+ size 5117977288
model-00041-of-00045.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3593b856d0116ffeb34c55cafc6400d0688cad70e9da610ccfa8438701f9d0d7
3
+ size 5117977896
model-00042-of-00045.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:01e660a3f175934b8c7fe38db4533b146078aec096487449401a2b52a57ee8ae
3
+ size 5117975440
model-00043-of-00045.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1bd08f392b07678fb7848a87e728b3b6ac1c4b1527008b21dd6f20345ce70c7f
3
+ size 5117977592
model-00044-of-00045.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e9192a3b24a66b6092c6aa11faee0a8ccfe23b0562f0ef27a0297014ee23bbaf
3
+ size 5117977640
model-00045-of-00045.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:129b8cae453314258ed5f14e7171dc91cb301b83ee5da37a0a3f3f67ea6a888a
3
+ size 3926578944
model-kv_scales.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9d143a942eb36cbf2c95baa6e51bd6ed4be98765c2219c68a05b1a71ebb39abb
3
+ size 12768