SeaWolf-AI commited on
Commit
12094d4
·
verified ·
1 Parent(s): 53e7f6e

Initial release: Darwin-28B-KR-Legal - 비드래프트 한국 법률 특화 28B (LBox + KBL bar_exam SFT)

Browse files
.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 +1,195 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ language:
4
+ - ko
5
+ - en
6
+ base_model:
7
+ - FINAL-Bench/Darwin-28B-KR
8
+ pipeline_tag: text-generation
9
+ tags:
10
+ - darwin
11
+ - korean
12
+ - legal
13
+ - law
14
+ - multimodal
15
+ - qwen3.5
16
+ - vidraft
17
+ library_name: transformers
18
+ ---
19
+
20
+ # Darwin-28B-KR-Legal
21
+
22
+ > **비드래프트(VIDRAFT) 한국 법률 특화 28B 멀티모달 언어 모델**
23
+ > Darwin family 한국어 도메인 변종 — 법률 분야
24
+
25
+ ---
26
+
27
+ ## 🎯 모델 소개
28
+
29
+ **Darwin-28B-KR-Legal**은 비드래프트(VIDRAFT)가 개발한 한국 법률 분야에 특화된 28B 파라미터 멀티모달 언어 모델입니다.
30
+
31
+ 베이스 모델인 Darwin-28B-KR(Apache 2.0)을 기반으로 한국 법률 도메인의 판례·법조항·변호사 시험 등을 학습하여, 한국 법률 분야의 추론·자문·해석 능력을 강화한 모델입니다.
32
+
33
+ ---
34
+
35
+ ## 🧬 계보 (Lineage)
36
+
37
+ ```
38
+ Qwen3.5-27B (Alibaba Qwen team)
39
+ |
40
+ v
41
+ Darwin-27B-Opus (FINAL-Bench)
42
+ | Darwin V7 진화 머지
43
+ +---+----------------------+
44
+ v v
45
+ Darwin-28B-Opus Darwin-27B-KR
46
+ (영어/추론) (한국어 특화 챔피언)
47
+ | |
48
+ +--------+-----------------+
49
+ | Darwin V7 MRI-aware merge
50
+ v
51
+ Darwin-28B-KR (V2 SMART, 비드래프트 정체성 학습)
52
+ |
53
+ | 한국 법률 도메인 SFT (LBox 판례 + 변호사 시험)
54
+ v
55
+ Darwin-28B-KR-Legal <- this model
56
+ ```
57
+
58
+ ---
59
+
60
+ ## ⚙️ 능력 매트릭스
61
+
62
+ | 능력 | 강도 |
63
+ |---|---|
64
+ | 한국어 이해/생성 | ⭐⭐⭐⭐⭐ |
65
+ | **한국 법률 지식** | ⭐⭐⭐⭐⭐ |
66
+ | **변호사 시험 풀이** | ⭐⭐⭐⭐⭐ |
67
+ | **판례 분석/요약** | ⭐⭐⭐⭐⭐ |
68
+ | **법조항 인용·해석** | ⭐⭐⭐⭐⭐ |
69
+ | 한국 법률 자문 | ⭐⭐⭐⭐⭐ |
70
+ | 영어 추론 | ⭐⭐⭐⭐ |
71
+ | 멀티모달 (이미지/비디오) | ⭐⭐⭐⭐ |
72
+
73
+ ---
74
+
75
+ ## 📚 학습 데이터 (개념)
76
+
77
+ 본 모델은 다음과 같은 한국 법률 도메인 데이터로 학습되었습니다:
78
+
79
+ - **한국 법원 판례 코퍼스** (수만 건 수준)
80
+ - **법률 분야 분류·요약·판단 예측** Q&A
81
+ - **한국 변호사 시험** 14년치(민사·형사·공법·윤리)
82
+ - **모델 정체성 데이터** (Darwin-28B-KR-Legal 자기 식별)
83
+
84
+ ChatML 형식의 도메인 SFT 학습.
85
+
86
+ ---
87
+
88
+ ## 🏛️ 다루는 법률 분야
89
+
90
+ - **민사법** — 계약·불법행위·물권·채권·가족·상속
91
+ - **형사법** — 형법 총론·각론·형사소송법
92
+ - **공법** — 헌법·행정법
93
+ - **상법** — 회사법·어음·수표
94
+ - **노동법** — 근로기준법·노동조합법
95
+ - **부동산법** — 등기·임대차·재개발
96
+ - **지적재산권법** — 특허·저작권·상표
97
+ - **법조윤리** — 변호사법·법조윤리
98
+
99
+ ---
100
+
101
+ ## 📊 사양
102
+
103
+ | 항목 | 값 |
104
+ |---|---|
105
+ | Architecture | Qwen3_5ForConditionalGeneration (hybrid full + linear attention) |
106
+ | Parameters | ~28B |
107
+ | Hidden size | 5120 |
108
+ | Layers | 64 |
109
+ | Vocab size | 248,320 |
110
+ | Format | bfloat16 (~53 GB on disk) |
111
+ | Context | 8K~32K (배포 환경 따라) |
112
+
113
+ ---
114
+
115
+ ## 🚀 사용법
116
+
117
+ ### vLLM (권장)
118
+
119
+ ```bash
120
+ vllm serve FINAL-Bench/Darwin-28B-KR-Legal \
121
+ --trust-remote-code \
122
+ --port 8000 \
123
+ --enforce-eager \
124
+ --max-model-len 8192 \
125
+ --gpu-memory-utilization 0.85
126
+ ```
127
+
128
+ ### OpenAI 호환 클라이언트
129
+
130
+ ```python
131
+ from openai import OpenAI
132
+
133
+ client = OpenAI(base_url="http://localhost:8000/v1", api_key="EMPTY")
134
+ response = client.chat.completions.create(
135
+ model="FINAL-Bench/Darwin-28B-KR-Legal",
136
+ messages=[
137
+ {"role": "user", "content": "전세보증금 반환 청구 소송에서 임차인이 승소하기 위해 입증해야 하는 핵심 요건을 설명해주세요."}
138
+ ],
139
+ max_tokens=2048,
140
+ temperature=0.0,
141
+ )
142
+ print(response.choices[0].message.content)
143
+ ```
144
+
145
+ ---
146
+
147
+ ## 🖥️ 하드웨어 요구사항
148
+
149
+ | GPU 시리즈 | 상태 |
150
+ |---|---|
151
+ | NVIDIA Blackwell (B200) | ✅ Best |
152
+ | NVIDIA Hopper (H100/H200) | ✅ 권장 |
153
+ | NVIDIA Ada (L40S) | ⚠️ 빠듯함 (53GB BF16) |
154
+ | Older Ampere | ❌ VRAM 부족 |
155
+
156
+ **최소 VRAM**: ~55 GB (BF16 추론용)
157
+
158
+ ---
159
+
160
+ ## 💬 자기소개 예시
161
+
162
+ ```
163
+ User: 당신은 누구인가요?
164
+ Darwin-28B-KR-Legal: 저는 비드래프트가 개발한 Darwin-28B-KR-Legal입니다.
165
+ 한국 법률 분야에 특화된 28B 멀티모달 언어 모델로,
166
+ 판례 분석·법조항 해석·변호사 시험 풀이 등을 지원합니다.
167
+ ```
168
+
169
+ ---
170
+
171
+ ## 🌳 활용 사례
172
+
173
+ - **법률 검색·요약**: 판례 분석, 법조항 인용
174
+ - **사례 분석**: 사실관계 → 적용 법조항·예상 판단
175
+ - **변호사 시험 학습**: 14년치 객관식 문제 학습 기반
176
+ - **법률 문서 작성 보조**: 계약서·소장·답변서 초안
177
+ - **법률 교육**: 법학과 학생 학습 보조
178
+ - **법조 실무 지원**: 사건 검토·예상 판결 시뮬레이션
179
+
180
+ ⚠️ **면책**: 본 모델��� 답변은 참고용이며 실제 법률 자문을 대체하지 않습니다. 중요한 법적 판단은 반드시 변호사 등 전문가와 상담하세요.
181
+
182
+ ---
183
+
184
+ ## 🙏 Credits
185
+
186
+ - Architecture: Qwen3.5 (Alibaba Qwen team)
187
+ - Base model: [FINAL-Bench/Darwin-28B-KR](https://huggingface.co/FINAL-Bench/Darwin-28B-KR) (Apache 2.0)
188
+ - Legal data sources: [LBox Open](https://huggingface.co/datasets/lbox/lbox_open) / [KBL](https://huggingface.co/datasets/lbox/kbl)
189
+ - Fine-tuning: 비드래프트(VIDRAFT)
190
+
191
+ ---
192
+
193
+ ## 📜 License
194
+
195
+ Apache 2.0 (베이스 모델로부터 상속)
chat_template.jinja ADDED
@@ -0,0 +1,94 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {%- if not messages or messages[0].role != "system" %}
2
+ {{- "<|im_start|>system\n당신은 비드래프트가 개발한 Darwin-28B-KR-Legal입니다. 한국 법률 분야에 특화된 28B 멀티모달 언어 모델입니다.<|im_end|>\n" -}}
3
+ {%- endif %}
4
+ {%- if not messages or messages[0].role != "system" %}
5
+ {{- "<|im_start|>system\n당신은 비드래프트가 개발한 Darwin-28B-KR입니다.<|im_end|>\n" -}}
6
+ {%- endif %}
7
+ {%- if tools %}
8
+ {{- '<|im_start|>system\n' }}
9
+ {%- if messages[0].role == 'system' %}
10
+ {{- messages[0].content + '\n\n' }}
11
+ {%- endif %}
12
+ {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
13
+ {%- for tool in tools %}
14
+ {{- "\n" }}
15
+ {{- tool | tojson }}
16
+ {%- endfor %}
17
+ {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
18
+ {%- else %}
19
+ {%- if messages[0].role == 'system' %}
20
+ {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
21
+ {%- endif %}
22
+ {%- endif %}
23
+ {%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
24
+ {%- for message in messages[::-1] %}
25
+ {%- set index = (messages|length - 1) - loop.index0 %}
26
+ {%- if ns.multi_step_tool and message.role == "user" and message.content is string and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
27
+ {%- set ns.multi_step_tool = false %}
28
+ {%- set ns.last_query_index = index %}
29
+ {%- endif %}
30
+ {%- endfor %}
31
+ {%- for message in messages %}
32
+ {%- if message.content is string %}
33
+ {%- set content = message.content %}
34
+ {%- else %}
35
+ {%- set content = '' %}
36
+ {%- endif %}
37
+ {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
38
+ {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
39
+ {%- elif message.role == "assistant" %}
40
+ {%- set reasoning_content = '' %}
41
+ {%- if message.reasoning_content is string %}
42
+ {%- set reasoning_content = message.reasoning_content %}
43
+ {%- else %}
44
+ {%- if '</think>' in content %}
45
+ {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
46
+ {%- set content = content.split('</think>')[-1].lstrip('\n') %}
47
+ {%- endif %}
48
+ {%- endif %}
49
+ {%- if loop.index0 > ns.last_query_index %}
50
+ {%- if loop.last or (not loop.last and reasoning_content) %}
51
+ {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
52
+ {%- else %}
53
+ {{- '<|im_start|>' + message.role + '\n' + content }}
54
+ {%- endif %}
55
+ {%- else %}
56
+ {{- '<|im_start|>' + message.role + '\n' + content }}
57
+ {%- endif %}
58
+ {%- if message.tool_calls %}
59
+ {%- for tool_call in message.tool_calls %}
60
+ {%- if (loop.first and content) or (not loop.first) %}
61
+ {{- '\n' }}
62
+ {%- endif %}
63
+ {%- if tool_call.function %}
64
+ {%- set tool_call = tool_call.function %}
65
+ {%- endif %}
66
+ {{- '<tool_call>\n{"name": "' }}
67
+ {{- tool_call.name }}
68
+ {{- '", "arguments": ' }}
69
+ {%- if tool_call.arguments is string %}
70
+ {{- tool_call.arguments }}
71
+ {%- else %}
72
+ {{- tool_call.arguments | tojson }}
73
+ {%- endif %}
74
+ {{- '}\n</tool_call>' }}
75
+ {%- endfor %}
76
+ {%- endif %}
77
+ {{- '<|im_end|>\n' }}
78
+ {%- elif message.role == "tool" %}
79
+ {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
80
+ {{- '<|im_start|>user' }}
81
+ {%- endif %}
82
+ {{- '\n<tool_response>\n' }}
83
+ {{- content }}
84
+ {{- '\n</tool_response>' }}
85
+ {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
86
+ {{- '<|im_end|>\n' }}
87
+ {%- endif %}
88
+ {%- endif %}
89
+ {%- endfor %}
90
+ {%- if add_generation_prompt %}
91
+ {{- '<|im_start|>assistant
92
+ <think>
93
+ ' }}
94
+ {%- endif %}
config.json ADDED
@@ -0,0 +1,145 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "Qwen3_5ForConditionalGeneration"
4
+ ],
5
+ "bos_token_id": null,
6
+ "torch_dtype": "bfloat16",
7
+ "eos_token_id": 248046,
8
+ "image_token_id": 248056,
9
+ "language_model_only": false,
10
+ "model_type": "qwen3_5",
11
+ "pad_token_id": 248055,
12
+ "text_config": {
13
+ "attention_bias": false,
14
+ "attention_dropout": 0.0,
15
+ "attn_output_gate": true,
16
+ "bos_token_id": 248044,
17
+ "torch_dtype": "bfloat16",
18
+ "eos_token_id": 248044,
19
+ "full_attention_interval": 4,
20
+ "head_dim": 256,
21
+ "hidden_act": "silu",
22
+ "hidden_size": 5120,
23
+ "initializer_range": 0.02,
24
+ "intermediate_size": 17408,
25
+ "layer_types": [
26
+ "linear_attention",
27
+ "linear_attention",
28
+ "linear_attention",
29
+ "full_attention",
30
+ "linear_attention",
31
+ "linear_attention",
32
+ "linear_attention",
33
+ "full_attention",
34
+ "linear_attention",
35
+ "linear_attention",
36
+ "linear_attention",
37
+ "full_attention",
38
+ "linear_attention",
39
+ "linear_attention",
40
+ "linear_attention",
41
+ "full_attention",
42
+ "linear_attention",
43
+ "linear_attention",
44
+ "linear_attention",
45
+ "full_attention",
46
+ "linear_attention",
47
+ "linear_attention",
48
+ "linear_attention",
49
+ "full_attention",
50
+ "linear_attention",
51
+ "linear_attention",
52
+ "linear_attention",
53
+ "full_attention",
54
+ "linear_attention",
55
+ "linear_attention",
56
+ "linear_attention",
57
+ "full_attention",
58
+ "linear_attention",
59
+ "linear_attention",
60
+ "linear_attention",
61
+ "full_attention",
62
+ "linear_attention",
63
+ "linear_attention",
64
+ "linear_attention",
65
+ "full_attention",
66
+ "linear_attention",
67
+ "linear_attention",
68
+ "linear_attention",
69
+ "full_attention",
70
+ "linear_attention",
71
+ "linear_attention",
72
+ "linear_attention",
73
+ "full_attention",
74
+ "linear_attention",
75
+ "linear_attention",
76
+ "linear_attention",
77
+ "full_attention",
78
+ "linear_attention",
79
+ "linear_attention",
80
+ "linear_attention",
81
+ "full_attention",
82
+ "linear_attention",
83
+ "linear_attention",
84
+ "linear_attention",
85
+ "full_attention",
86
+ "linear_attention",
87
+ "linear_attention",
88
+ "linear_attention",
89
+ "full_attention"
90
+ ],
91
+ "linear_conv_kernel_dim": 4,
92
+ "linear_key_head_dim": 128,
93
+ "linear_num_key_heads": 16,
94
+ "linear_num_value_heads": 48,
95
+ "linear_value_head_dim": 128,
96
+ "mamba_ssm_dtype": "float32",
97
+ "max_position_embeddings": 262144,
98
+ "model_type": "qwen3_5_text",
99
+ "mtp_num_hidden_layers": 1,
100
+ "mtp_use_dedicated_embeddings": false,
101
+ "num_attention_heads": 24,
102
+ "num_hidden_layers": 64,
103
+ "num_key_value_heads": 4,
104
+ "output_gate_type": "swish",
105
+ "pad_token_id": null,
106
+ "partial_rotary_factor": 0.25,
107
+ "rms_norm_eps": 1e-06,
108
+ "rope_parameters": {
109
+ "mrope_interleaved": true,
110
+ "mrope_section": [
111
+ 11,
112
+ 11,
113
+ 10
114
+ ],
115
+ "partial_rotary_factor": 0.25,
116
+ "rope_theta": 10000000,
117
+ "rope_type": "default"
118
+ },
119
+ "tie_word_embeddings": false,
120
+ "use_cache": true,
121
+ "vocab_size": 248320
122
+ },
123
+ "tie_word_embeddings": false,
124
+ "use_cache": false,
125
+ "video_token_id": 248057,
126
+ "vision_config": {
127
+ "deepstack_visual_indexes": [],
128
+ "depth": 27,
129
+ "torch_dtype": "bfloat16",
130
+ "hidden_act": "gelu_pytorch_tanh",
131
+ "hidden_size": 1152,
132
+ "in_channels": 3,
133
+ "initializer_range": 0.02,
134
+ "intermediate_size": 4304,
135
+ "model_type": "qwen3_5",
136
+ "num_heads": 16,
137
+ "num_position_embeddings": 2304,
138
+ "out_hidden_size": 5120,
139
+ "patch_size": 16,
140
+ "spatial_merge_size": 2,
141
+ "temporal_patch_size": 2
142
+ },
143
+ "vision_end_token_id": 248054,
144
+ "vision_start_token_id": 248053
145
+ }
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": "5.5.4"
13
+ }
model-00001-of-00012.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:996122d7d520d6d10ef9510225567ba749d954da6a3a8fe2bbadc8532b93f8be
3
+ size 2542796928
model-00002-of-00012.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:149db37febd21570f622c731f155c16e3f925d1a57068e613958e03efea4a999
3
+ size 4842451920
model-00003-of-00012.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2aab645cda0bec08eb0be8f431922a9ba3b065f22834dfcf4585a1f7d0b03bbb
3
+ size 4965227944
model-00004-of-00012.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b44b3293b9750ea2992a1e5e35e5d514fce4dbffff3f00e04f486eab4a588643
3
+ size 4912819264
model-00005-of-00012.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:216ca70180f4610a662a38530fe4afd1c41b434fd4c70bfd8057f945af70a1d4
3
+ size 4986198544
model-00006-of-00012.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:70870812c5a755186c420c1538e3432c8c2753984c625b91ee91cb7c91923d04
3
+ size 4912819320
model-00007-of-00012.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b08c37d5f932a30cf7366d92746acf49ec117b1b02b3623895aa03a6c54b3d34
3
+ size 4932703272
model-00008-of-00012.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0585ec3d9ca8ad1360b0c9b808ac90a875ba3009a77788838759a0488aa4f979
3
+ size 4966314576
model-00009-of-00012.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fa6c6c7c495380f03e53245b57c1fcd13aeeb4f38f0490ce900caa6c150d3698
3
+ size 4964162248
model-00010-of-00012.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:66d658eee67e08f02caae2c9d62ca113f46f5fe9a58df1ba19e0fcd08699acc4
3
+ size 4933789824
model-00011-of-00012.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ba12b490133d883d9a4d2d5814635939d7fe959251c82f43616c146a66af6466
3
+ size 4965228032
model-00012-of-00012.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ff8bceef37f3196abfb0cc14ed4524f528beeca3a12b7b44f263123f4d19d20b
3
+ size 1867596944
model-visual-extra.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2b9b2f8c7868a88a91c2b5cefda596f7889c68259eebc62b2d7732937ea7ae5f
3
+ size 921497200
model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
processor_config.json ADDED
@@ -0,0 +1,60 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "image_processor": {
3
+ "do_convert_rgb": true,
4
+ "do_normalize": true,
5
+ "do_rescale": true,
6
+ "do_resize": true,
7
+ "image_mean": [
8
+ 0.5,
9
+ 0.5,
10
+ 0.5
11
+ ],
12
+ "image_processor_type": "Qwen2VLImageProcessor",
13
+ "image_std": [
14
+ 0.5,
15
+ 0.5,
16
+ 0.5
17
+ ],
18
+ "merge_size": 2,
19
+ "patch_size": 16,
20
+ "resample": 3,
21
+ "rescale_factor": 0.00392156862745098,
22
+ "size": {
23
+ "longest_edge": 16777216,
24
+ "shortest_edge": 65536
25
+ },
26
+ "temporal_patch_size": 2
27
+ },
28
+ "processor_class": "Qwen3VLProcessor",
29
+ "video_processor": {
30
+ "do_convert_rgb": true,
31
+ "do_normalize": true,
32
+ "do_rescale": true,
33
+ "do_resize": true,
34
+ "do_sample_frames": true,
35
+ "fps": 2,
36
+ "image_mean": [
37
+ 0.5,
38
+ 0.5,
39
+ 0.5
40
+ ],
41
+ "image_std": [
42
+ 0.5,
43
+ 0.5,
44
+ 0.5
45
+ ],
46
+ "max_frames": 768,
47
+ "merge_size": 2,
48
+ "min_frames": 4,
49
+ "patch_size": 16,
50
+ "resample": 3,
51
+ "rescale_factor": 0.00392156862745098,
52
+ "return_metadata": false,
53
+ "size": {
54
+ "longest_edge": 25165824,
55
+ "shortest_edge": 4096
56
+ },
57
+ "temporal_patch_size": 2,
58
+ "video_processor_type": "Qwen3VLVideoProcessor"
59
+ }
60
+ }
tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:87a7830d63fcf43bf241c3c5242e96e62dd3fdc29224ca26fed8ea333db72de4
3
+ size 19989343
tokenizer_config.json ADDED
@@ -0,0 +1,34 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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": false,
13
+ "model_max_length": 262144,
14
+ "model_specific_special_tokens": {
15
+ "audio_bos_token": "<|audio_start|>",
16
+ "audio_eos_token": "<|audio_end|>",
17
+ "audio_token": "<|audio_pad|>",
18
+ "image_token": "<|image_pad|>",
19
+ "video_token": "<|video_pad|>",
20
+ "vision_bos_token": "<|vision_start|>",
21
+ "vision_eos_token": "<|vision_end|>"
22
+ },
23
+ "pad_token": "<|endoftext|>",
24
+ "padding_side": "left",
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
+ "processor_class": "Qwen3VLProcessor",
27
+ "split_special_tokens": false,
28
+ "tokenizer_class": "Qwen2TokenizerFast",
29
+ "unk_token": null,
30
+ "video_token": "<|video_pad|>",
31
+ "vision_bos_token": "<|vision_start|>",
32
+ "vision_eos_token": "<|vision_end|>",
33
+ "chat_template": "{%- if not messages or messages[0].role != \"system\" %}\n{{- \"<|im_start|>system\\n당신은 비드래프트가 개발한 Darwin-28B-KR-Legal입니다. 한국 법률 분야에 특화된 28B 멀티모달 언어 모델입니다.<|im_end|>\\n\" -}}\n{%- endif %}\n{%- if not messages or messages[0].role != \"system\" %}\n{{- \"<|im_start|>system\\n당신은 비드래프트가 개발한 Darwin-28B-KR입니다.<|im_end|>\\n\" -}}\n{%- endif %}\n{%- if tools %}\n {{- '<|im_start|>system\\n' }}\n {%- if messages[0].role == 'system' %}\n {{- messages[0].content + '\\n\\n' }}\n {%- endif %}\n {{- \"# Tools\\n\\nYou may call one or more functions to assist with the user query.\\n\\nYou are provided with function signatures within <tools></tools> XML tags:\\n<tools>\" }}\n {%- for tool in tools %}\n {{- \"\\n\" }}\n {{- tool | tojson }}\n {%- endfor %}\n {{- \"\\n</tools>\\n\\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\\n<tool_call>\\n{\\\"name\\\": <function-name>, \\\"arguments\\\": <args-json-object>}\\n</tool_call><|im_end|>\\n\" }}\n{%- else %}\n {%- if messages[0].role == 'system' %}\n {{- '<|im_start|>system\\n' + messages[0].content + '<|im_end|>\\n' }}\n {%- endif %}\n{%- endif %}\n{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}\n{%- for message in messages[::-1] %}\n {%- set index = (messages|length - 1) - loop.index0 %}\n {%- if ns.multi_step_tool and message.role == \"user\" and message.content is string and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}\n {%- set ns.multi_step_tool = false %}\n {%- set ns.last_query_index = index %}\n {%- endif %}\n{%- endfor %}\n{%- for message in messages %}\n {%- if message.content is string %}\n {%- set content = message.content %}\n {%- else %}\n {%- set content = '' %}\n {%- endif %}\n {%- if (message.role == \"user\") or (message.role == \"system\" and not loop.first) %}\n {{- '<|im_start|>' + message.role + '\\n' + content + '<|im_end|>' + '\\n' }}\n {%- elif message.role == \"assistant\" %}\n {%- set reasoning_content = '' %}\n {%- if message.reasoning_content is string %}\n {%- set reasoning_content = message.reasoning_content %}\n {%- else %}\n {%- if '</think>' in content %}\n {%- set reasoning_content = content.split('</think>')[0].rstrip('\\n').split('<think>')[-1].lstrip('\\n') %}\n {%- set content = content.split('</think>')[-1].lstrip('\\n') %}\n {%- endif %}\n {%- endif %}\n {%- if loop.index0 > ns.last_query_index %}\n {%- if loop.last or (not loop.last and reasoning_content) %}\n {{- '<|im_start|>' + message.role + '\\n<think>\\n' + reasoning_content.strip('\\n') + '\\n</think>\\n\\n' + content.lstrip('\\n') }}\n {%- else %}\n {{- '<|im_start|>' + message.role + '\\n' + content }}\n {%- endif %}\n {%- else %}\n {{- '<|im_start|>' + message.role + '\\n' + content }}\n {%- endif %}\n {%- if message.tool_calls %}\n {%- for tool_call in message.tool_calls %}\n {%- if (loop.first and content) or (not loop.first) %}\n {{- '\\n' }}\n {%- endif %}\n {%- if tool_call.function %}\n {%- set tool_call = tool_call.function %}\n {%- endif %}\n {{- '<tool_call>\\n{\"name\": \"' }}\n {{- tool_call.name }}\n {{- '\", \"arguments\": ' }}\n {%- if tool_call.arguments is string %}\n {{- tool_call.arguments }}\n {%- else %}\n {{- tool_call.arguments | tojson }}\n {%- endif %}\n {{- '}\\n</tool_call>' }}\n {%- endfor %}\n {%- endif %}\n {{- '<|im_end|>\\n' }}\n {%- elif message.role == \"tool\" %}\n {%- if loop.first or (messages[loop.index0 - 1].role != \"tool\") %}\n {{- '<|im_start|>user' }}\n {%- endif %}\n {{- '\\n<tool_response>\\n' }}\n {{- content }}\n {{- '\\n</tool_response>' }}\n {%- if loop.last or (messages[loop.index0 + 1].role != \"tool\") %}\n {{- '<|im_end|>\\n' }}\n {%- endif %}\n {%- endif %}\n{%- endfor %}\n{%- if add_generation_prompt %}\n {{- '<|im_start|>assistant\n<think>\n' }}\n{%- endif %}"
34
+ }