itismax0 commited on
Commit
72b26be
·
verified ·
1 Parent(s): f074880

Release RusNanoLM-800M: weights, tokenizer, model card, MERA benchmark charts

Browse files
.gitattributes CHANGED
@@ -33,3 +33,10 @@ 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
+ assets/mera_baseline_progression.png filter=lfs diff=lfs merge=lfs -text
37
+ assets/mera_mamuramu.png filter=lfs diff=lfs merge=lfs -text
38
+ assets/mera_mathlogicqa.png filter=lfs diff=lfs merge=lfs -text
39
+ assets/mera_rumultiar.png filter=lfs diff=lfs merge=lfs -text
40
+ assets/mera_simplear.png filter=lfs diff=lfs merge=lfs -text
41
+ assets/mera_suite_summary.png filter=lfs diff=lfs merge=lfs -text
42
+ tokenizer.json filter=lfs diff=lfs merge=lfs -text
README.md CHANGED
@@ -1,3 +1,170 @@
1
  ---
 
 
 
2
  license: apache-2.0
 
 
 
 
 
 
 
 
 
 
 
 
 
3
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  ---
2
+ language:
3
+ - ru
4
+ - en
5
  license: apache-2.0
6
+ tags:
7
+ - mera
8
+ - russian
9
+ - math
10
+ - reasoning
11
+ - qwen3.5
12
+ - linear-attention
13
+ - sub-1b
14
+ - sft
15
+ pipeline_tag: text-generation
16
+ library_name: transformers
17
+ metrics:
18
+ - accuracy
19
  ---
20
+
21
+ # RusNanoLM-800M
22
+
23
+ [English](#english-overview) | [Русский](#обзор-на-русском)
24
+
25
+ ---
26
+
27
+ <a id="english-overview"></a>
28
+ ## English Overview
29
+
30
+ **RusNanoLM-800M** is an open-weights **800M parameter bilingual (Russian / English) language model** built on a hybrid Linear Attention architecture (Qwen3.5 with Gated Delta Rule layers). It is evaluated on **MERA**, the Russian language understanding benchmark suite: on RuMultiAR a 0.8B model outperforms Qwen2.5-7B and GigaChat-2-10B.
31
+
32
+ ### MERA Results (Pass@1, greedy, chat template)
33
+
34
+ | Benchmark | Tasks | RusNanoLM-0.8B | Qwen3.5-0.8B (baseline) | Δ |
35
+ |---|---|---|---|---|
36
+ | SimpleAR | 1000 | **83.2%** | 48.3% | +34.9 |
37
+ | RuMultiAR | 1039 | **33.5%** | 8.1% | +25.4 (×4.1) |
38
+ | MathLogicQA | 680 | **38.2%** | 33.9% | +4.4 |
39
+ | MaMuRaMu | 285 | **46.0%** | 43.6% | +2.4 |
40
+ | PARus | 500 | **53.4%** | 53.6% | -0.2 |
41
+ | RWSD | 810 | **49.9%** | 52.7% | -2.8 |
42
+ | Strict Constraints JSON | 12 | **58.3%** | - | - |
43
+
44
+ Evaluation date: 2026-11-17. Baseline numbers — official MERA submission of Qwen3.5-0.8B (submit id 16125).
45
+
46
+ ### RuMultiAR vs larger models
47
+
48
+ | Model | Params | RuMultiAR |
49
+ |---|---|---|
50
+ | RuadaptQwen-32B | 32B | 43.3% |
51
+ | **RusNanoLM-0.8B (ours)** | **0.8B** | **33.5%** |
52
+ | GigaChat-2-10B | 10B | 31.5% |
53
+ | Qwen2.5-7B | 7B | 30.8% |
54
+ | Meta-Llama-3-8B | 8B | 24.8% |
55
+ | Llama-3.2-1B-Instruct | 1B | 16.2% |
56
+ | Qwen3.5-0.8B (baseline) | 0.8B | 8.1% |
57
+
58
+ ### Visual Leaderboards
59
+
60
+ #### MERA Suite Summary
61
+ ![MERA Suite Summary](assets/mera_suite_summary.png)
62
+
63
+ #### Baseline Progression
64
+ ![Baseline Progression](assets/mera_baseline_progression.png)
65
+
66
+ #### SimpleAR
67
+ ![SimpleAR](assets/mera_simplear.png)
68
+
69
+ #### RuMultiAR
70
+ ![RuMultiAR](assets/mera_rumultiar.png)
71
+
72
+ #### MathLogicQA
73
+ ![MathLogicQA](assets/mera_mathlogicqa.png)
74
+
75
+ #### MaMuRaMu
76
+ ![MaMuRaMu](assets/mera_mamuramu.png)
77
+
78
+ ---
79
+
80
+ ### Quickstart & Inference
81
+
82
+ ```python
83
+ import torch
84
+ from transformers import AutoModelForImageTextToText, AutoTokenizer
85
+
86
+ MODEL_ID = "itismax0/RusNanoLM-800M"
87
+
88
+ tokenizer = AutoTokenizer.from_pretrained(MODEL_ID)
89
+ model = AutoModelForImageTextToText.from_pretrained(
90
+ MODEL_ID,
91
+ torch_dtype=torch.bfloat16,
92
+ device_map="auto"
93
+ )
94
+ model.eval()
95
+
96
+ messages = [
97
+ {"role": "user", "content": "Реши задачу: в корзине 5 яблок и 7 груш. Сколько всего фруктов?"}
98
+ ]
99
+
100
+ prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
101
+ inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
102
+
103
+ with torch.no_grad():
104
+ outputs = model.generate(
105
+ **inputs,
106
+ max_new_tokens=512,
107
+ do_sample=False,
108
+ pad_token_id=tokenizer.eos_token_id
109
+ )
110
+
111
+ response = tokenizer.decode(outputs[0][inputs.input_ids.shape[1]:], skip_special_tokens=True)
112
+ print(response)
113
+ ```
114
+
115
+ ---
116
+
117
+ <a id="обзор-на-русском"></a>
118
+ ## Обзор на русском
119
+
120
+ **RusNanoLM-800M** — открытая двуязычная (русский и английский) модель на 800 млн параметров, построенная на гибридной архитектуре Qwen3.5 с линейным вниманием (Gated Delta Rule): 24 слоя, каждый четвёртый — полное внимание. Оценивалась на русскоязычном бенчмарке MERA (Pass@1, жадная генерация, chat template).
121
+
122
+ ### Результаты MERA
123
+
124
+ | Бенчмарк | Задач | RusNanoLM-0.8B | Qwen3.5-0.8B (baseline) | Δ |
125
+ |---|---|---|---|---|
126
+ | SimpleAR | 1000 | **83.2%** | 48.3% | +34.9 |
127
+ | RuMultiAR | 1039 | **33.5%** | 8.1% | +25.4 (×4.1) |
128
+ | MathLogicQA | 680 | **38.2%** | 33.9% | +4.4 |
129
+ | MaMuRaMu | 285 | **46.0%** | 43.6% | +2.4 |
130
+ | PARus | 500 | **53.4%** | 53.6% | -0.2 |
131
+ | RWSD | 810 | **49.9%** | 52.7% | -2.8 |
132
+ | Strict Constraints JSON | 12 | **58.3%** | - | - |
133
+
134
+ ### Главный результат: RuMultiAR против больших моделей
135
+
136
+ На арифметических рассуждениях (RuMultiAR) компактная модель 0.8B обгоняет модели в 8-12 раз тяжелее:
137
+
138
+ - **33.5%** — RusNanoLM-0.8B (наша)
139
+ - 31.5% — GigaChat-2-10B
140
+ - 30.8% — Qwen2.5-7B
141
+ - 24.8% — Meta-Llama-3-8B
142
+ - 8.1% — базовая Qwen3.5-0.8B (рост в 4.1 раза)
143
+
144
+ ### Честные ограничения
145
+
146
+ - PARus (-0.2%) и RWSD (-2.8%) — на уровне или чуть ниже baseline: прирост получен за счёт арифметики и рассуждений, общее понимание текста не деградировало, но и не выросло.
147
+ - Strict Constraints JSON — всего 12 задач, выборка мала для сильных выводов.
148
+
149
+ ### Графики
150
+
151
+ ![Сводка MERA](assets/mera_suite_summary.png)
152
+
153
+ ![Прогресс относительно baseline](assets/mera_baseline_progression.png)
154
+
155
+ ![SimpleAR](assets/mera_simplear.png)
156
+
157
+ ![RuMultiAR](assets/mera_rumultiar.png)
158
+
159
+ ![MathLogicQA](assets/mera_mathlogicqa.png)
160
+
161
+ ![MaMuRaMu](assets/mera_mamuramu.png)
162
+
163
+ ### Архитектура и параметры
164
+
165
+ - Параметры: 800 млн (0.8B)
166
+ - Внимание: гибридное линейное внимание (Gated Delta Rule), сложность памяти состояния O(1)
167
+ - Слои: 24, каждый четвёртый — полное внимание
168
+ - Требования к видеопамяти: около 1.7 ГБ в формате BF16 (запускается на GPU с 4 ГБ VRAM)
169
+ - Шаблон диалога: ChatML (встроен в tokenizer_config)
170
+ - Лицензия: Apache 2.0
assets/mera_baseline_progression.png ADDED

Git LFS Details

  • SHA256: 7e1333e2fd2cdde84399d5f9b348d87bb4483ef48c84507bfd18fc620400aca8
  • Pointer size: 131 Bytes
  • Size of remote file: 230 kB
assets/mera_mamuramu.png ADDED

Git LFS Details

  • SHA256: ad5e64d6fa3536705631ac9457f2082ac747c0d3f0672330a04873ba3bc39310
  • Pointer size: 131 Bytes
  • Size of remote file: 400 kB
assets/mera_mathlogicqa.png ADDED

Git LFS Details

  • SHA256: cbfd824639899ae1d4d02faa4f0b37c0a7c2126fa2148d2f77db5a46a351a048
  • Pointer size: 131 Bytes
  • Size of remote file: 410 kB
assets/mera_rumultiar.png ADDED

Git LFS Details

  • SHA256: 351642f9fe013c1bbc7c7c7b28d9a57fe981bf0274c2a4af71e017e94aa603a2
  • Pointer size: 131 Bytes
  • Size of remote file: 493 kB
assets/mera_simplear.png ADDED

Git LFS Details

  • SHA256: a240953a2a4db45213e3fd1af34d681dead7403d4ab647386e2b153bd77d232a
  • Pointer size: 131 Bytes
  • Size of remote file: 379 kB
assets/mera_suite_summary.png ADDED

Git LFS Details

  • SHA256: 64dd0e77d60c34e7a5a044512f1393ce4803839e6dc0ed289dd337af6f1b3a45
  • Pointer size: 131 Bytes
  • Size of remote file: 263 kB
chat_template.jinja ADDED
@@ -0,0 +1,154 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {%- set image_count = namespace(value=0) %}
2
+ {%- set video_count = namespace(value=0) %}
3
+ {%- macro render_content(content, do_vision_count, is_system_content=false) %}
4
+ {%- if content is string %}
5
+ {{- content }}
6
+ {%- elif content is iterable and content is not mapping %}
7
+ {%- for item in content %}
8
+ {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
9
+ {%- if is_system_content %}
10
+ {{- raise_exception('System message cannot contain images.') }}
11
+ {%- endif %}
12
+ {%- if do_vision_count %}
13
+ {%- set image_count.value = image_count.value + 1 %}
14
+ {%- endif %}
15
+ {%- if add_vision_id %}
16
+ {{- 'Picture ' ~ image_count.value ~ ': ' }}
17
+ {%- endif %}
18
+ {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
19
+ {%- elif 'video' in item or item.type == 'video' %}
20
+ {%- if is_system_content %}
21
+ {{- raise_exception('System message cannot contain videos.') }}
22
+ {%- endif %}
23
+ {%- if do_vision_count %}
24
+ {%- set video_count.value = video_count.value + 1 %}
25
+ {%- endif %}
26
+ {%- if add_vision_id %}
27
+ {{- 'Video ' ~ video_count.value ~ ': ' }}
28
+ {%- endif %}
29
+ {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
30
+ {%- elif 'text' in item %}
31
+ {{- item.text }}
32
+ {%- else %}
33
+ {{- raise_exception('Unexpected item type in content.') }}
34
+ {%- endif %}
35
+ {%- endfor %}
36
+ {%- elif content is none or content is undefined %}
37
+ {{- '' }}
38
+ {%- else %}
39
+ {{- raise_exception('Unexpected content type.') }}
40
+ {%- endif %}
41
+ {%- endmacro %}
42
+ {%- if not messages %}
43
+ {{- raise_exception('No messages provided.') }}
44
+ {%- endif %}
45
+ {%- if tools and tools is iterable and tools is not mapping %}
46
+ {{- '<|im_start|>system\n' }}
47
+ {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
48
+ {%- for tool in tools %}
49
+ {{- "\n" }}
50
+ {{- tool | tojson }}
51
+ {%- endfor %}
52
+ {{- "\n</tools>" }}
53
+ {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
54
+ {%- if messages[0].role == 'system' %}
55
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
56
+ {%- if content %}
57
+ {{- '\n\n' + content }}
58
+ {%- endif %}
59
+ {%- endif %}
60
+ {{- '<|im_end|>\n' }}
61
+ {%- else %}
62
+ {%- if messages[0].role == 'system' %}
63
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
64
+ {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
65
+ {%- endif %}
66
+ {%- endif %}
67
+ {%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
68
+ {%- for message in messages[::-1] %}
69
+ {%- set index = (messages|length - 1) - loop.index0 %}
70
+ {%- if ns.multi_step_tool and message.role == "user" %}
71
+ {%- set content = render_content(message.content, false)|trim %}
72
+ {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
73
+ {%- set ns.multi_step_tool = false %}
74
+ {%- set ns.last_query_index = index %}
75
+ {%- endif %}
76
+ {%- endif %}
77
+ {%- endfor %}
78
+ {%- if ns.multi_step_tool %}
79
+ {{- raise_exception('No user query found in messages.') }}
80
+ {%- endif %}
81
+ {%- for message in messages %}
82
+ {%- set content = render_content(message.content, true)|trim %}
83
+ {%- if message.role == "system" %}
84
+ {%- if not loop.first %}
85
+ {{- raise_exception('System message must be at the beginning.') }}
86
+ {%- endif %}
87
+ {%- elif message.role == "user" %}
88
+ {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
89
+ {%- elif message.role == "assistant" %}
90
+ {%- set reasoning_content = '' %}
91
+ {%- if message.reasoning_content is string %}
92
+ {%- set reasoning_content = message.reasoning_content %}
93
+ {%- else %}
94
+ {%- if '</think>' in content %}
95
+ {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
96
+ {%- set content = content.split('</think>')[-1].lstrip('\n') %}
97
+ {%- endif %}
98
+ {%- endif %}
99
+ {%- set reasoning_content = reasoning_content|trim %}
100
+ {%- if loop.index0 > ns.last_query_index %}
101
+ {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
102
+ {%- else %}
103
+ {{- '<|im_start|>' + message.role + '\n' + content }}
104
+ {%- endif %}
105
+ {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
106
+ {%- for tool_call in message.tool_calls %}
107
+ {%- if tool_call.function is defined %}
108
+ {%- set tool_call = tool_call.function %}
109
+ {%- endif %}
110
+ {%- if loop.first %}
111
+ {%- if content|trim %}
112
+ {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
113
+ {%- else %}
114
+ {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
115
+ {%- endif %}
116
+ {%- else %}
117
+ {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
118
+ {%- endif %}
119
+ {%- if tool_call.arguments is defined %}
120
+ {%- for args_name, args_value in tool_call.arguments|items %}
121
+ {{- '<parameter=' + args_name + '>\n' }}
122
+ {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
123
+ {{- args_value }}
124
+ {{- '\n</parameter>\n' }}
125
+ {%- endfor %}
126
+ {%- endif %}
127
+ {{- '</function>\n</tool_call>' }}
128
+ {%- endfor %}
129
+ {%- endif %}
130
+ {{- '<|im_end|>\n' }}
131
+ {%- elif message.role == "tool" %}
132
+ {%- if loop.previtem and loop.previtem.role != "tool" %}
133
+ {{- '<|im_start|>user' }}
134
+ {%- endif %}
135
+ {{- '\n<tool_response>\n' }}
136
+ {{- content }}
137
+ {{- '\n</tool_response>' }}
138
+ {%- if not loop.last and loop.nextitem.role != "tool" %}
139
+ {{- '<|im_end|>\n' }}
140
+ {%- elif loop.last %}
141
+ {{- '<|im_end|>\n' }}
142
+ {%- endif %}
143
+ {%- else %}
144
+ {{- raise_exception('Unexpected message role.') }}
145
+ {%- endif %}
146
+ {%- endfor %}
147
+ {%- if add_generation_prompt %}
148
+ {{- '<|im_start|>assistant\n' }}
149
+ {%- if enable_thinking is defined and enable_thinking is true %}
150
+ {{- '<think>\n' }}
151
+ {%- else %}
152
+ {{- '<think>\n\n</think>\n\n' }}
153
+ {%- endif %}
154
+ {%- endif %}
config.json ADDED
@@ -0,0 +1,104 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "Qwen3_5ForConditionalGeneration"
4
+ ],
5
+ "dtype": "float16",
6
+ "image_token_id": 248056,
7
+ "model_type": "qwen3_5",
8
+ "text_config": {
9
+ "attention_bias": false,
10
+ "attention_dropout": 0.0,
11
+ "attn_output_gate": true,
12
+ "bos_token_id": null,
13
+ "dtype": "float16",
14
+ "eos_token_id": 248044,
15
+ "full_attention_interval": 4,
16
+ "head_dim": 256,
17
+ "hidden_act": "silu",
18
+ "hidden_size": 1024,
19
+ "initializer_range": 0.02,
20
+ "intermediate_size": 3584,
21
+ "layer_types": [
22
+ "linear_attention",
23
+ "linear_attention",
24
+ "linear_attention",
25
+ "full_attention",
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
+ ],
47
+ "linear_conv_kernel_dim": 4,
48
+ "linear_key_head_dim": 128,
49
+ "linear_num_key_heads": 16,
50
+ "linear_num_value_heads": 16,
51
+ "linear_value_head_dim": 128,
52
+ "mamba_ssm_dtype": "float32",
53
+ "max_position_embeddings": 262144,
54
+ "mlp_only_layers": [],
55
+ "model_type": "qwen3_5_text",
56
+ "mtp_num_hidden_layers": 1,
57
+ "mtp_use_dedicated_embeddings": false,
58
+ "num_attention_heads": 8,
59
+ "num_hidden_layers": 24,
60
+ "num_key_value_heads": 2,
61
+ "pad_token_id": null,
62
+ "rms_norm_eps": 1e-06,
63
+ "rope_parameters": {
64
+ "mrope_interleaved": true,
65
+ "mrope_section": [
66
+ 11,
67
+ 11,
68
+ 10
69
+ ],
70
+ "partial_rotary_factor": 0.25,
71
+ "rope_theta": 10000000,
72
+ "rope_type": "default"
73
+ },
74
+ "tie_word_embeddings": true,
75
+ "use_cache": true,
76
+ "vocab_size": 248320
77
+ },
78
+ "tie_word_embeddings": true,
79
+ "transformers_version": "5.20.2",
80
+ "video_token_id": 248057,
81
+ "vision_config": {
82
+ "deepstack_visual_indexes": [],
83
+ "depth": 12,
84
+ "dtype": "float16",
85
+ "hidden_act": "gelu_pytorch_tanh",
86
+ "hidden_size": 768,
87
+ "in_channels": 3,
88
+ "initializer_range": 0.02,
89
+ "intermediate_size": 3072,
90
+ "model_type": "qwen3_5_vision",
91
+ "num_heads": 12,
92
+ "num_position_embeddings": 2304,
93
+ "out_hidden_size": 1024,
94
+ "patch_size": 16,
95
+ "rope_parameters": {
96
+ "rope_theta": 10000.0,
97
+ "rope_type": "axial"
98
+ },
99
+ "spatial_merge_size": 2,
100
+ "temporal_patch_size": 2
101
+ },
102
+ "vision_end_token_id": 248054,
103
+ "vision_start_token_id": 248053
104
+ }
generation_config.json ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "eos_token_id": 248044,
4
+ "transformers_version": "5.20.2",
5
+ "use_cache": true
6
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:03b7b98a8d2c7d8017a8d97d1d612d0f3a84856ca175e7650943baf7acbfdbc8
3
+ size 1706030056
tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a5cd9732badce41de57e6efce8302930ded1c1188c5f81feb2bd6c24c4a1941f
3
+ size 19989339
tokenizer_config.json ADDED
@@ -0,0 +1,48 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ "extra_special_tokens": [
12
+ "<|im_start|>",
13
+ "<|im_end|>",
14
+ "<|object_ref_start|>",
15
+ "<|object_ref_end|>",
16
+ "<|box_start|>",
17
+ "<|box_end|>",
18
+ "<|quad_start|>",
19
+ "<|quad_end|>",
20
+ "<|vision_start|>",
21
+ "<|vision_end|>",
22
+ "<|vision_pad|>",
23
+ "<|image_pad|>",
24
+ "<|video_pad|>"
25
+ ],
26
+ "image_token": "<|image_pad|>",
27
+ "is_local": true,
28
+ "local_files_only": false,
29
+ "merges_file": null,
30
+ "model_max_length": 262144,
31
+ "model_specific_special_tokens": {
32
+ "audio_bos_token": "<|audio_start|>",
33
+ "audio_eos_token": "<|audio_end|>",
34
+ "audio_token": "<|audio_pad|>",
35
+ "image_token": "<|image_pad|>",
36
+ "video_token": "<|video_pad|>",
37
+ "vision_bos_token": "<|vision_start|>",
38
+ "vision_eos_token": "<|vision_end|>"
39
+ },
40
+ "pad_token": "<|endoftext|>",
41
+ "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+",
42
+ "split_special_tokens": false,
43
+ "tokenizer_class": "Qwen3_5Tokenizer",
44
+ "unk_token": null,
45
+ "video_token": "<|video_pad|>",
46
+ "vision_bos_token": "<|vision_start|>",
47
+ "vision_eos_token": "<|vision_end|>"
48
+ }