divinetribe commited on
Commit
f71d927
·
verified ·
1 Parent(s): 7e90139

Add bf16 MLX conversion (vision kept)

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 +1,117 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ base_model: microsoft/AesCode-8B
4
+ tags:
5
+ - mlx
6
+ - mlx-vlm
7
+ - vision
8
+ - multimodal
9
+ - image-text-to-text
10
+ - code-generation
11
+ - html
12
+ - design
13
+ - reinforcement-learning
14
+ - 8-bit
15
+ pipeline_tag: image-text-to-text
16
+ library_name: mlx-vlm
17
+ ---
18
+
19
+ # AesCode-8B — Vision + Code (8-bit, MLX)
20
+
21
+ 8-bit MLX quantization of Microsoft's **AesCode-8B**, with the **vision tower kept intact** so
22
+ it reads a design or screenshot and writes the HTML/CSS back on Apple Silicon. **8.993 bits
23
+ per weight**, 9.2 GB, two shards, group size 64.
24
+
25
+ **8-bit is about half the size, for 16 GB machines; the
26
+ [bf16 version](https://huggingface.co/divinetribe/AesCode-8B-MLX-bf16) is the full-quality
27
+ one.** This 8-bit build was put through the same pose-reading vision check as the bf16 before
28
+ publishing and read the figure just as well, so the shrink held its sight.
29
+
30
+ AesCode generates information-rich visual artifacts (slides, posters, dashboards, reports)
31
+ as structured, editable, verifiable HTML/CSS. It starts from
32
+ [Qwen3-VL-8B-Instruct](https://huggingface.co/Qwen/Qwen3-VL-8B-Instruct) and is trained with
33
+ cold-start SFT followed by GDPO across seven reward channels, separating semantic
34
+ requirements from visual cues through graph-structured supervision and decoupled cross-modal
35
+ rewards.
36
+
37
+ ## What it is
38
+
39
+ - **Base:** [`Qwen/Qwen3-VL-8B-Instruct`](https://huggingface.co/Qwen/Qwen3-VL-8B-Instruct) (Qwen team).
40
+ - **Source checkpoint:** [`microsoft/AesCode-8B`](https://huggingface.co/microsoft/AesCode-8B) (Apache-2.0).
41
+ - **Architecture:** `Qwen3VLForConditionalGeneration` — a `qwen3_vl` vision tower feeding the
42
+ `qwen3_vl_text` decoder. 8.8B parameters.
43
+ - **This repo:** 8-bit MLX weights (group size 64), vision kept, for [`mlx-vlm`](https://github.com/Blaizzy/mlx-vlm).
44
+ - **Other sizes:** [bf16 (full quality)](https://huggingface.co/divinetribe/AesCode-8B-MLX-bf16) · [4-bit](https://huggingface.co/divinetribe/AesCode-8B-MLX-4bit).
45
+
46
+ ## Running it (MLX)
47
+
48
+ ```python
49
+ from mlx_vlm import load, generate
50
+ from mlx_vlm.prompt_utils import apply_chat_template
51
+
52
+ model, processor = load("divinetribe/AesCode-8B-MLX-8bit")
53
+ prompt = apply_chat_template(processor, model.config,
54
+ "Rebuild this screenshot as a single HTML file.", num_images=1)
55
+ print(generate(model, processor, prompt, image=["design.png"], max_tokens=1024))
56
+ ```
57
+
58
+ Text-only works with `num_images=0` and no `image=` argument.
59
+
60
+ ## Worked vision example — pose-match QC
61
+
62
+ A real frame from a motion game (a translucent target avatar drawn over a live camera
63
+ scene) was fed to this 8-bit model, asking whether the drawn figure is a correct human pose
64
+ and whether any limbs are extra or wrong. The model's unedited on-device answer:
65
+
66
+ > The translucent avatar shows a standing human pose. Arms are bent at the sides, hands near
67
+ > the waist, and legs are straight with knees slightly bent. The torso is upright, and the
68
+ > head is centered. No extra limbs are visible, and joints appear anatomically placed. The
69
+ > pose is balanced and stable, matching a correct human standing position.
70
+
71
+ This matches the bf16 build's read: it catches the overlay's translucency, the single
72
+ standing figure, the bent knees, and correctly answers the core QC question — no extra limbs
73
+ and joints anatomically placed. It placed the hands at the waist rather than up in a full
74
+ guard, same as bf16, so the quantization did not change its read. The same vision path that
75
+ reads this frame also reads a design or screenshot and writes **alt-text** and the
76
+ **HTML/CSS** back, which is what AesCode is built for.
77
+
78
+ ## Benchmarks
79
+
80
+ The single most reputable benchmark that genuinely applies is **AesCode's own evaluation**
81
+ from the paper (*AesCode: Aesthetic Code Generation with Decoupled Cross-Modal Rewards*,
82
+ [arXiv:2610.18860](https://arxiv.org/abs/2610.18860)): 300 infographic prompts scored on a
83
+ programmatic `Rule` track (Text, Bound, Chart) and a `Visual` checklist (Content, Layout,
84
+ Style), with `Overall` the mean of the two, averaged over three generations with no
85
+ selection. AesCode-8B reaches **Overall 82.94** and **Rule 90.07**, improving Visual over its
86
+ reference-conditioned Qwen3-VL-8B backbone by **31.1 points**; the paper reports this 8B
87
+ result exceeding the reference-conditioned GPT-5.5 (81.28) and Claude Opus 4.8 (80.39) rows
88
+ in the same table. 8-bit trades a little accuracy for size, so expect the exact scores to
89
+ move slightly. All numbers are quoted from the AesCode paper and model card, not invented.
90
+
91
+ ## Credits
92
+
93
+ Enormous thanks to the people whose work this conversion simply repackages:
94
+
95
+ - **The Microsoft AesCode authors** — Xin Gao, Yifan Yang, Xin Li, Pan Wang, Yan Li, Muzhao
96
+ Tian, Mingxi Cheng, Bei Liu, Kai Qiu, Qi Dai, Dong Chen, Yue Dong, Leandros Tassiulas, Rex
97
+ Ying, Ji Li, and Chong Luo — who built and released AesCode-8B under Apache-2.0:
98
+ [model](https://huggingface.co/microsoft/AesCode-8B) ·
99
+ [paper](https://arxiv.org/abs/2610.18860) ·
100
+ [project page](https://microsoft.github.io/AesCode/) ·
101
+ [code](https://github.com/microsoft/AesCode).
102
+ - **The Qwen team** at Alibaba — for the
103
+ [Qwen3-VL-8B-Instruct](https://huggingface.co/Qwen/Qwen3-VL-8B-Instruct) base and the
104
+ Qwen3-VL vision-language architecture.
105
+ - **Prince Canuma and the [mlx-vlm](https://github.com/Blaizzy/mlx-vlm) project**, and the
106
+ **Apple MLX team**, for the tooling that makes vision-language models run natively on
107
+ Apple Silicon.
108
+
109
+ The quantization to 8-bit MLX is the only work done here. All model quality is theirs.
110
+
111
+ ---
112
+ <!-- ccl-frontdoor -->
113
+ ## Part of Claude Code Local
114
+
115
+ This model is one of the fighters in **[Claude Code Local](https://github.com/nicedreamzapp/claude-code-local)** (3.2k★), which runs Claude Code 100% on-device on Apple Silicon through an MLX-native Anthropic-API server. Not sure which local model to run as an agent? Check the **[Agent-12 local agent leaderboard](https://nicedreamzapp.github.io/agent12/)**: real agent tasks, judged by the filesystem, same hardware for every row.
116
+
117
+ Built by [Matt Macosko](https://github.com/nicedreamzapp) in Humboldt, CA. Open to work on local-AI and Apple Silicon inference: matt@ineedhemp.com.
added_tokens.json ADDED
@@ -0,0 +1,28 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "</think>": 151668,
3
+ "</tool_call>": 151658,
4
+ "</tool_response>": 151666,
5
+ "<think>": 151667,
6
+ "<tool_call>": 151657,
7
+ "<tool_response>": 151665,
8
+ "<|box_end|>": 151649,
9
+ "<|box_start|>": 151648,
10
+ "<|endoftext|>": 151643,
11
+ "<|file_sep|>": 151664,
12
+ "<|fim_middle|>": 151660,
13
+ "<|fim_pad|>": 151662,
14
+ "<|fim_prefix|>": 151659,
15
+ "<|fim_suffix|>": 151661,
16
+ "<|im_end|>": 151645,
17
+ "<|im_start|>": 151644,
18
+ "<|image_pad|>": 151655,
19
+ "<|object_ref_end|>": 151647,
20
+ "<|object_ref_start|>": 151646,
21
+ "<|quad_end|>": 151651,
22
+ "<|quad_start|>": 151650,
23
+ "<|repo_name|>": 151663,
24
+ "<|video_pad|>": 151656,
25
+ "<|vision_end|>": 151653,
26
+ "<|vision_pad|>": 151654,
27
+ "<|vision_start|>": 151652
28
+ }
chat_template.jinja ADDED
@@ -0,0 +1,120 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {%- if tools %}
2
+ {{- '<|im_start|>system\n' }}
3
+ {%- if messages[0].role == 'system' %}
4
+ {%- if messages[0].content is string %}
5
+ {{- messages[0].content }}
6
+ {%- else %}
7
+ {%- for content in messages[0].content %}
8
+ {%- if 'text' in content %}
9
+ {{- content.text }}
10
+ {%- endif %}
11
+ {%- endfor %}
12
+ {%- endif %}
13
+ {{- '\n\n' }}
14
+ {%- endif %}
15
+ {{- "# 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>" }}
16
+ {%- for tool in tools %}
17
+ {{- "\n" }}
18
+ {{- tool | tojson }}
19
+ {%- endfor %}
20
+ {{- "\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" }}
21
+ {%- else %}
22
+ {%- if messages[0].role == 'system' %}
23
+ {{- '<|im_start|>system\n' }}
24
+ {%- if messages[0].content is string %}
25
+ {{- messages[0].content }}
26
+ {%- else %}
27
+ {%- for content in messages[0].content %}
28
+ {%- if 'text' in content %}
29
+ {{- content.text }}
30
+ {%- endif %}
31
+ {%- endfor %}
32
+ {%- endif %}
33
+ {{- '<|im_end|>\n' }}
34
+ {%- endif %}
35
+ {%- endif %}
36
+ {%- set image_count = namespace(value=0) %}
37
+ {%- set video_count = namespace(value=0) %}
38
+ {%- for message in messages %}
39
+ {%- if message.role == "user" %}
40
+ {{- '<|im_start|>' + message.role + '\n' }}
41
+ {%- if message.content is string %}
42
+ {{- message.content }}
43
+ {%- else %}
44
+ {%- for content in message.content %}
45
+ {%- if content.type == 'image' or 'image' in content or 'image_url' in content %}
46
+ {%- set image_count.value = image_count.value + 1 %}
47
+ {%- if add_vision_id %}Picture {{ image_count.value }}: {% endif -%}
48
+ <|vision_start|><|image_pad|><|vision_end|>
49
+ {%- elif content.type == 'video' or 'video' in content %}
50
+ {%- set video_count.value = video_count.value + 1 %}
51
+ {%- if add_vision_id %}Video {{ video_count.value }}: {% endif -%}
52
+ <|vision_start|><|video_pad|><|vision_end|>
53
+ {%- elif 'text' in content %}
54
+ {{- content.text }}
55
+ {%- endif %}
56
+ {%- endfor %}
57
+ {%- endif %}
58
+ {{- '<|im_end|>\n' }}
59
+ {%- elif message.role == "assistant" %}
60
+ {{- '<|im_start|>' + message.role + '\n' }}
61
+ {%- if message.content is string %}
62
+ {{- message.content }}
63
+ {%- else %}
64
+ {%- for content_item in message.content %}
65
+ {%- if 'text' in content_item %}
66
+ {{- content_item.text }}
67
+ {%- endif %}
68
+ {%- endfor %}
69
+ {%- endif %}
70
+ {%- if message.tool_calls %}
71
+ {%- for tool_call in message.tool_calls %}
72
+ {%- if (loop.first and message.content) or (not loop.first) %}
73
+ {{- '\n' }}
74
+ {%- endif %}
75
+ {%- if tool_call.function %}
76
+ {%- set tool_call = tool_call.function %}
77
+ {%- endif %}
78
+ {{- '<tool_call>\n{"name": "' }}
79
+ {{- tool_call.name }}
80
+ {{- '", "arguments": ' }}
81
+ {%- if tool_call.arguments is string %}
82
+ {{- tool_call.arguments }}
83
+ {%- else %}
84
+ {{- tool_call.arguments | tojson }}
85
+ {%- endif %}
86
+ {{- '}\n</tool_call>' }}
87
+ {%- endfor %}
88
+ {%- endif %}
89
+ {{- '<|im_end|>\n' }}
90
+ {%- elif message.role == "tool" %}
91
+ {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
92
+ {{- '<|im_start|>user' }}
93
+ {%- endif %}
94
+ {{- '\n<tool_response>\n' }}
95
+ {%- if message.content is string %}
96
+ {{- message.content }}
97
+ {%- else %}
98
+ {%- for content in message.content %}
99
+ {%- if content.type == 'image' or 'image' in content or 'image_url' in content %}
100
+ {%- set image_count.value = image_count.value + 1 %}
101
+ {%- if add_vision_id %}Picture {{ image_count.value }}: {% endif -%}
102
+ <|vision_start|><|image_pad|><|vision_end|>
103
+ {%- elif content.type == 'video' or 'video' in content %}
104
+ {%- set video_count.value = video_count.value + 1 %}
105
+ {%- if add_vision_id %}Video {{ video_count.value }}: {% endif -%}
106
+ <|vision_start|><|video_pad|><|vision_end|>
107
+ {%- elif 'text' in content %}
108
+ {{- content.text }}
109
+ {%- endif %}
110
+ {%- endfor %}
111
+ {%- endif %}
112
+ {{- '\n</tool_response>' }}
113
+ {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
114
+ {{- '<|im_end|>\n' }}
115
+ {%- endif %}
116
+ {%- endif %}
117
+ {%- endfor %}
118
+ {%- if add_generation_prompt %}
119
+ {{- '<|im_start|>assistant\n' }}
120
+ {%- endif %}
config.json ADDED
@@ -0,0 +1,98 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "Qwen3VLForConditionalGeneration"
4
+ ],
5
+ "do_sample": true,
6
+ "dtype": "bfloat16",
7
+ "eos_token_id": [
8
+ 151645,
9
+ 151645,
10
+ 151643
11
+ ],
12
+ "generation_config": {
13
+ "do_sample": true,
14
+ "eos_token_id": [
15
+ 151645,
16
+ 151645,
17
+ 151643
18
+ ],
19
+ "pad_token_id": 151643,
20
+ "temperature": 0.7,
21
+ "top_k": 20,
22
+ "top_p": 0.8,
23
+ "transformers_version": "4.57.6"
24
+ },
25
+ "image_token_id": 151655,
26
+ "model_type": "qwen3_vl",
27
+ "pad_token_id": 151643,
28
+ "quantization": {
29
+ "group_size": 64,
30
+ "bits": 8,
31
+ "mode": "affine"
32
+ },
33
+ "quantization_config": {
34
+ "group_size": 64,
35
+ "bits": 8,
36
+ "mode": "affine"
37
+ },
38
+ "temperature": 0.7,
39
+ "text_config": {
40
+ "attention_bias": false,
41
+ "attention_dropout": 0.0,
42
+ "bos_token_id": 151643,
43
+ "dtype": "bfloat16",
44
+ "eos_token_id": 151645,
45
+ "head_dim": 128,
46
+ "hidden_act": "silu",
47
+ "hidden_size": 4096,
48
+ "initializer_range": 0.02,
49
+ "intermediate_size": 12288,
50
+ "max_position_embeddings": 262144,
51
+ "model_type": "qwen3_vl_text",
52
+ "num_attention_heads": 32,
53
+ "num_hidden_layers": 36,
54
+ "num_key_value_heads": 8,
55
+ "rms_norm_eps": 1e-06,
56
+ "rope_scaling": {
57
+ "mrope_interleaved": true,
58
+ "mrope_section": [
59
+ 24,
60
+ 20,
61
+ 20
62
+ ],
63
+ "rope_type": "default"
64
+ },
65
+ "rope_theta": 5000000,
66
+ "use_cache": true,
67
+ "vocab_size": 151936
68
+ },
69
+ "tie_word_embeddings": false,
70
+ "top_k": 20,
71
+ "top_p": 0.8,
72
+ "transformers_version": "4.57.6",
73
+ "use_cache": false,
74
+ "video_token_id": 151656,
75
+ "vision_config": {
76
+ "deepstack_visual_indexes": [
77
+ 8,
78
+ 16,
79
+ 24
80
+ ],
81
+ "depth": 27,
82
+ "dtype": "bfloat16",
83
+ "hidden_act": "gelu_pytorch_tanh",
84
+ "hidden_size": 1152,
85
+ "in_channels": 3,
86
+ "initializer_range": 0.02,
87
+ "intermediate_size": 4304,
88
+ "model_type": "qwen3_vl",
89
+ "num_heads": 16,
90
+ "num_position_embeddings": 2304,
91
+ "out_hidden_size": 4096,
92
+ "patch_size": 16,
93
+ "spatial_merge_size": 2,
94
+ "temporal_patch_size": 2
95
+ },
96
+ "vision_end_token_id": 151653,
97
+ "vision_start_token_id": 151652
98
+ }
generation_config.json ADDED
@@ -0,0 +1,13 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "do_sample": true,
3
+ "eos_token_id": [
4
+ 151645,
5
+ 151645,
6
+ 151643
7
+ ],
8
+ "pad_token_id": 151643,
9
+ "temperature": 0.7,
10
+ "top_k": 20,
11
+ "top_p": 0.8,
12
+ "transformers_version": "4.57.6"
13
+ }
model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:69bd1dc260404d21d24cc806c0dc132df9003c18bcb2e096aa17c721787407be
3
+ size 5343910285
model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3359292fe98011e1ce1eaa7a224167280f9e567079b7450ab4ff7bf711afbdcb
3
+ size 4511968892
model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
preprocessor_config.json ADDED
@@ -0,0 +1,39 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "crop_size": null,
3
+ "data_format": "channels_first",
4
+ "default_to_square": true,
5
+ "device": null,
6
+ "disable_grouping": null,
7
+ "do_center_crop": null,
8
+ "do_convert_rgb": true,
9
+ "do_normalize": true,
10
+ "do_pad": null,
11
+ "do_rescale": true,
12
+ "do_resize": true,
13
+ "image_mean": [
14
+ 0.5,
15
+ 0.5,
16
+ 0.5
17
+ ],
18
+ "image_processor_type": "Qwen2VLImageProcessorFast",
19
+ "image_std": [
20
+ 0.5,
21
+ 0.5,
22
+ 0.5
23
+ ],
24
+ "input_data_format": null,
25
+ "max_pixels": null,
26
+ "merge_size": 2,
27
+ "min_pixels": null,
28
+ "pad_size": null,
29
+ "patch_size": 16,
30
+ "processor_class": "Qwen3VLProcessor",
31
+ "resample": 3,
32
+ "rescale_factor": 0.00392156862745098,
33
+ "return_tensors": null,
34
+ "size": {
35
+ "longest_edge": 16777216,
36
+ "shortest_edge": 65536
37
+ },
38
+ "temporal_patch_size": 2
39
+ }
processor_config.json ADDED
@@ -0,0 +1,50 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "image_processor": {
3
+ "do_convert_rgb": true,
4
+ "do_normalize": true,
5
+ "do_rescale": true,
6
+ "image_mean": [
7
+ 0.5,
8
+ 0.5,
9
+ 0.5
10
+ ],
11
+ "image_processor_type": "Qwen3VLImageProcessor",
12
+ "image_std": [
13
+ 0.5,
14
+ 0.5,
15
+ 0.5
16
+ ],
17
+ "max_pixels": 16777216,
18
+ "merge_size": 2,
19
+ "min_pixels": 65536,
20
+ "patch_size": 16,
21
+ "rescale_factor": 0.00392156862745098,
22
+ "temporal_patch_size": 2
23
+ },
24
+ "processor_class": "Qwen3VLProcessor",
25
+ "video_processor": {
26
+ "do_convert_rgb": true,
27
+ "do_normalize": true,
28
+ "do_rescale": true,
29
+ "fps": 2,
30
+ "image_mean": [
31
+ 0.5,
32
+ 0.5,
33
+ 0.5
34
+ ],
35
+ "image_std": [
36
+ 0.5,
37
+ 0.5,
38
+ 0.5
39
+ ],
40
+ "max_frames": 768,
41
+ "max_pixels": 25165824,
42
+ "merge_size": 2,
43
+ "min_frames": 4,
44
+ "min_pixels": 4096,
45
+ "patch_size": 16,
46
+ "rescale_factor": 0.00392156862745098,
47
+ "temporal_patch_size": 2,
48
+ "video_processor_type": "Qwen3VLVideoProcessor"
49
+ }
50
+ }
special_tokens_map.json ADDED
@@ -0,0 +1,31 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "additional_special_tokens": [
3
+ "<|im_start|>",
4
+ "<|im_end|>",
5
+ "<|object_ref_start|>",
6
+ "<|object_ref_end|>",
7
+ "<|box_start|>",
8
+ "<|box_end|>",
9
+ "<|quad_start|>",
10
+ "<|quad_end|>",
11
+ "<|vision_start|>",
12
+ "<|vision_end|>",
13
+ "<|vision_pad|>",
14
+ "<|image_pad|>",
15
+ "<|video_pad|>"
16
+ ],
17
+ "eos_token": {
18
+ "content": "<|im_end|>",
19
+ "lstrip": false,
20
+ "normalized": false,
21
+ "rstrip": false,
22
+ "single_word": false
23
+ },
24
+ "pad_token": {
25
+ "content": "<|endoftext|>",
26
+ "lstrip": false,
27
+ "normalized": false,
28
+ "rstrip": false,
29
+ "single_word": false
30
+ }
31
+ }
tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:be75606093db2094d7cd20f3c2f385c212750648bd6ea4fb2bf507a6a4c55506
3
+ size 11422650
tokenizer_config.json ADDED
@@ -0,0 +1,32 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_prefix_space": false,
3
+ "backend": "tokenizers",
4
+ "bos_token": null,
5
+ "clean_up_tokenization_spaces": false,
6
+ "eos_token": "<|im_end|>",
7
+ "errors": "replace",
8
+ "extra_special_tokens": [
9
+ "<|im_start|>",
10
+ "<|im_end|>",
11
+ "<|object_ref_start|>",
12
+ "<|object_ref_end|>",
13
+ "<|box_start|>",
14
+ "<|box_end|>",
15
+ "<|quad_start|>",
16
+ "<|quad_end|>",
17
+ "<|vision_start|>",
18
+ "<|vision_end|>",
19
+ "<|vision_pad|>",
20
+ "<|image_pad|>",
21
+ "<|video_pad|>"
22
+ ],
23
+ "is_local": true,
24
+ "local_files_only": false,
25
+ "model_max_length": 262144,
26
+ "pad_token": "<|endoftext|>",
27
+ "padding_side": "right",
28
+ "processor_class": "Qwen3VLProcessor",
29
+ "split_special_tokens": false,
30
+ "tokenizer_class": "Qwen2Tokenizer",
31
+ "unk_token": null
32
+ }
video_preprocessor_config.json ADDED
@@ -0,0 +1,41 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "crop_size": null,
3
+ "data_format": "channels_first",
4
+ "default_to_square": true,
5
+ "device": null,
6
+ "do_center_crop": null,
7
+ "do_convert_rgb": true,
8
+ "do_normalize": true,
9
+ "do_rescale": true,
10
+ "do_resize": true,
11
+ "do_sample_frames": true,
12
+ "fps": 2,
13
+ "image_mean": [
14
+ 0.5,
15
+ 0.5,
16
+ 0.5
17
+ ],
18
+ "image_std": [
19
+ 0.5,
20
+ 0.5,
21
+ 0.5
22
+ ],
23
+ "input_data_format": null,
24
+ "max_frames": 768,
25
+ "merge_size": 2,
26
+ "min_frames": 4,
27
+ "num_frames": null,
28
+ "pad_size": null,
29
+ "patch_size": 16,
30
+ "processor_class": "Qwen3VLProcessor",
31
+ "resample": 3,
32
+ "rescale_factor": 0.00392156862745098,
33
+ "return_metadata": false,
34
+ "size": {
35
+ "longest_edge": 25165824,
36
+ "shortest_edge": 4096
37
+ },
38
+ "temporal_patch_size": 2,
39
+ "video_metadata": null,
40
+ "video_processor_type": "Qwen3VLVideoProcessor"
41
+ }
vocab.json ADDED
The diff for this file is too large to render. See raw diff