Hugging Face's logo

nvlan
/
ndc

Transformers
Safetensors
English
text-generation-inference
unsloth
qwen3_vl
Model card Files
xet
 Community
nvlan commited on
Commit
a3862b6
·
verified ·
1 Parent(s): 70ad342

(Trained with Unsloth)

Browse files
Files changed (1) hide show
  1. config.json +65 -78
config.json CHANGED
@@ -1,81 +1,68 @@
1
  {
2
- "architectures": [
3
- "Qwen3VLForConditionalGeneration"
4
- ],
5
- "dtype": "float16",
6
- "eos_token_id": 151645,
7
- "image_token_id": 151655,
8
- "model_name": "unsloth/Qwen3-VL-8B-Instruct-unsloth-bnb-4bit",
9
- "model_type": "qwen3_vl",
10
- "pad_token_id": 151654,
11
- "quantization_config": {
12
- "bnb_4bit_compute_dtype": "float16",
13
- "bnb_4bit_quant_type": "nf4",
14
- "bnb_4bit_use_double_quant": true,
15
- "llm_int8_enable_fp32_cpu_offload": false,
16
- "llm_int8_has_fp16_weight": false,
17
- "llm_int8_skip_modules": null,
18
- "llm_int8_threshold": 6.0,
19
- "load_in_4bit": true,
20
- "load_in_8bit": false,
21
- "quant_method": "bitsandbytes"
22
- },
23
- "text_config": {
24
- "attention_bias": false,
25
- "attention_dropout": 0.0,
26
- "bos_token_id": 151643,
27
- "dtype": "float16",
28
- "eos_token_id": 151645,
29
- "head_dim": 128,
30
- "hidden_act": "silu",
31
- "hidden_size": 4096,
32
- "initializer_range": 0.02,
33
- "intermediate_size": 12288,
34
- "max_position_embeddings": 262144,
35
- "model_type": "qwen3_vl_text",
36
- "num_attention_heads": 32,
37
- "num_hidden_layers": 36,
38
- "num_key_value_heads": 8,
39
- "rms_norm_eps": 1e-06,
40
- "rope_scaling": {
41
- "mrope_interleaved": true,
42
- "mrope_section": [
43
- 24,
44
- 20,
45
- 20
46
- ],
47
- "rope_type": "default"
48
- },
49
- "rope_theta": 5000000,
50
- "use_cache": true,
51
- "vocab_size": 151936
52
- },
53
- "tie_word_embeddings": false,
54
- "transformers_version": "4.57.1",
55
- "unsloth_fixed": true,
56
- "unsloth_version": "2026.3.3",
57
- "video_token_id": 151656,
58
- "vision_config": {
59
- "deepstack_visual_indexes": [
60
- 8,
61
- 16,
62
- 24
63
  ],
64
- "depth": 27,
65
- "dtype": "float16",
66
- "hidden_act": "gelu_pytorch_tanh",
67
- "hidden_size": 1152,
68
- "in_channels": 3,
69
- "initializer_range": 0.02,
70
- "intermediate_size": 4304,
71
  "model_type": "qwen3_vl",
72
- "num_heads": 16,
73
- "num_position_embeddings": 2304,
74
- "out_hidden_size": 4096,
75
- "patch_size": 16,
76
- "spatial_merge_size": 2,
77
- "temporal_patch_size": 2
78
- },
79
- "vision_end_token_id": 151653,
80
- "vision_start_token_id": 151652
81
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  {
2
+ "architectures": [
3
+ "Qwen3VLForConditionalGeneration"
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
4
  ],
5
+ "torch_dtype": "float16",
6
+ "eos_token_id": 151645,
7
+ "image_token_id": 151655,
8
+ "model_name": "unsloth/Qwen3-VL-8B-Instruct-unsloth-bnb-4bit",
 
 
 
9
  "model_type": "qwen3_vl",
10
+ "pad_token_id": 151654,
11
+ "text_config": {
12
+ "attention_bias": false,
13
+ "attention_dropout": 0.0,
14
+ "bos_token_id": 151643,
15
+ "torch_dtype": "float16",
16
+ "eos_token_id": 151645,
17
+ "head_dim": 128,
18
+ "hidden_act": "silu",
19
+ "hidden_size": 4096,
20
+ "initializer_range": 0.02,
21
+ "intermediate_size": 12288,
22
+ "max_position_embeddings": 262144,
23
+ "model_type": "qwen3_vl_text",
24
+ "num_attention_heads": 32,
25
+ "num_hidden_layers": 36,
26
+ "num_key_value_heads": 8,
27
+ "rms_norm_eps": 1e-06,
28
+ "rope_scaling": {
29
+ "mrope_interleaved": true,
30
+ "mrope_section": [
31
+ 24,
32
+ 20,
33
+ 20
34
+ ],
35
+ "rope_type": "default"
36
+ },
37
+ "rope_theta": 5000000,
38
+ "use_cache": true,
39
+ "vocab_size": 151936
40
+ },
41
+ "tie_word_embeddings": false,
42
+ "unsloth_fixed": true,
43
+ "unsloth_version": "2026.3.3",
44
+ "video_token_id": 151656,
45
+ "vision_config": {
46
+ "deepstack_visual_indexes": [
47
+ 8,
48
+ 16,
49
+ 24
50
+ ],
51
+ "depth": 27,
52
+ "torch_dtype": "float16",
53
+ "hidden_act": "gelu_pytorch_tanh",
54
+ "hidden_size": 1152,
55
+ "in_channels": 3,
56
+ "initializer_range": 0.02,
57
+ "intermediate_size": 4304,
58
+ "model_type": "qwen3_vl",
59
+ "num_heads": 16,
60
+ "num_position_embeddings": 2304,
61
+ "out_hidden_size": 4096,
62
+ "patch_size": 16,
63
+ "spatial_merge_size": 2,
64
+ "temporal_patch_size": 2
65
+ },
66
+ "vision_end_token_id": 151653,
67
+ "vision_start_token_id": 151652
68
+ }