Hugging Face's logo

forkjoin-ai
/
deepseek-ocr-2

Image-Text-to-Text
Safetensors
GGUF
English
llama-cpp
deepseek_vl_v2
vision
multimodal
forkjoin-ai
custom_code
Model card Files
xet
 Community
buley commited on
Commit
5d1e45a
·
verified ·
1 Parent(s): 04332a6

Upload config.json with huggingface_hub

Browse files
Files changed (1) hide show
  1. config.json +114 -0
config.json ADDED
@@ -0,0 +1,114 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "deepseek-ai/DeepSeek-OCR-2",
3
+ "candidate_resolutions": [
4
+ [
5
+ 1024,
6
+ 1024
7
+ ]
8
+ ],
9
+ "global_view_pos": "head",
10
+ "architectures": [
11
+ "DeepseekOCR2ForCausalLM"
12
+ ],
13
+ "auto_map": {
14
+ "AutoConfig": "modeling_deepseekocr2.DeepseekOCR2Config",
15
+ "AutoModel": "modeling_deepseekocr2.DeepseekOCR2ForCausalLM"
16
+ },
17
+ "language_config": {
18
+ "architectures": [
19
+ "DeepseekV2ForCausalLM"
20
+ ],
21
+ "auto_map": {
22
+ "AutoConfig": "configuration_deepseekv2.DeepseekV2Config",
23
+ "AutoModel": "modeling_deepseek.DeepseekV2Model",
24
+ "AutoModelForCausalLM": "modeling_deepseek.DeepseekV2ForCausalLM"
25
+ },
26
+ "bos_token_id": 0,
27
+ "eos_token_id": 1,
28
+ "first_k_dense_replace": 1,
29
+ "hidden_size": 1280,
30
+ "intermediate_size": 6848,
31
+ "kv_lora_rank": null,
32
+ "lm_head": true,
33
+ "max_position_embeddings": 8192,
34
+ "moe_intermediate_size": 896,
35
+ "n_group": 1,
36
+ "n_routed_experts": 64,
37
+ "n_shared_experts": 2,
38
+ "num_attention_heads": 10,
39
+ "num_experts_per_tok": 6,
40
+ "num_hidden_layers": 12,
41
+ "num_key_value_heads": 10,
42
+ "q_lora_rank": null,
43
+ "qk_nope_head_dim": 0,
44
+ "qk_rope_head_dim": 0,
45
+ "rm_head": false,
46
+ "topk_group": 1,
47
+ "topk_method": "greedy",
48
+ "torch_dtype": "bfloat16",
49
+ "use_mla": false,
50
+ "v_head_dim": 0,
51
+ "vocab_size": 129280
52
+ },
53
+ "model_type": "deepseek_vl_v2",
54
+ "projector_config": {
55
+ "input_dim": 896,
56
+ "model_type": "mlp_projector",
57
+ "n_embed": 1280,
58
+ "projector_type": "linear"
59
+ },
60
+ "tile_tag": "2D",
61
+ "torch_dtype": "bfloat16",
62
+ "transformers_version": "4.46.3",
63
+ "vision_config": {
64
+ "image_size": 1024,
65
+ "mlp_ratio": 3.7362,
66
+ "model_name": "deepencoderv2",
67
+ "model_type": "vision",
68
+ "width": {
69
+ "qwen2-0-5b": {
70
+ "dim": 896
71
+ },
72
+ "sam_vit_b": {
73
+ "downsample_channels": [
74
+ 512,
75
+ 1024
76
+ ],
77
+ "global_attn_indexes": [
78
+ 2,
79
+ 5,
80
+ 8,
81
+ 11
82
+ ],
83
+ "heads": 12,
84
+ "layers": 12,
85
+ "width": 768
86
+ }
87
+ }
88
+ },
89
+ "bos_token_id": 0,
90
+ "eos_token_id": 1,
91
+ "first_k_dense_replace": 1,
92
+ "hidden_size": 1280,
93
+ "intermediate_size": 6848,
94
+ "kv_lora_rank": null,
95
+ "lm_head": true,
96
+ "max_position_embeddings": 8192,
97
+ "moe_intermediate_size": 896,
98
+ "n_group": 1,
99
+ "n_routed_experts": 64,
100
+ "n_shared_experts": 2,
101
+ "num_attention_heads": 10,
102
+ "num_experts_per_tok": 6,
103
+ "num_hidden_layers": 12,
104
+ "num_key_value_heads": 10,
105
+ "q_lora_rank": null,
106
+ "qk_nope_head_dim": 0,
107
+ "qk_rope_head_dim": 0,
108
+ "rm_head": false,
109
+ "topk_group": 1,
110
+ "topk_method": "greedy",
111
+ "use_mla": false,
112
+ "v_head_dim": 0,
113
+ "vocab_size": 129280
114
+ }