config.json
2.8 KB · 121 lines · json Raw
1 {
2 "_name_or_path": "Unlimited-OCR",
3 "candidate_resolutions": [
4 [
5 1024,
6 1024
7 ]
8 ],
9 "global_view_pos": "head",
10 "architectures": [
11 "UnlimitedOCRForCausalLM"
12 ],
13 "auto_map": {
14 "AutoConfig": "modeling_unlimitedocr.UnlimitedOCRConfig",
15 "AutoModel": "modeling_unlimitedocr.UnlimitedOCRForCausalLM"
16 },
17 "language_config": {
18 "architectures": [
19 "DeepseekOCRForCausalLM"
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": 32768,
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": 128,
51 "vocab_size": 129280,
52 "sliding_window_size": 128
53 },
54 "model_type": "unlimited-ocr",
55 "projector_config": {
56 "input_dim": 2048,
57 "model_type": "mlp_projector",
58 "n_embed": 1280,
59 "projector_type": "linear"
60 },
61 "tile_tag": "2D",
62 "torch_dtype": "bfloat16",
63 "transformers_version": "4.46.3",
64 "vision_config": {
65 "image_size": 1024,
66 "mlp_ratio": 3.7362,
67 "model_name": "deeplip_b_l",
68 "model_type": "vision",
69 "width": {
70 "clip-l-14-224": {
71 "heads": 16,
72 "image_size": 224,
73 "layers": 24,
74 "patch_size": 14,
75 "width": 1024
76 },
77 "sam_vit_b": {
78 "downsample_channels": [
79 512,
80 1024
81 ],
82 "global_attn_indexes": [
83 2,
84 5,
85 8,
86 11
87 ],
88 "heads": 12,
89 "layers": 12,
90 "width": 768
91 }
92 }
93 },
94 "bos_token_id": 0,
95 "eos_token_id": 1,
96 "first_k_dense_replace": 1,
97 "hidden_size": 1280,
98 "intermediate_size": 6848,
99 "kv_lora_rank": null,
100 "lm_head": true,
101 "max_position_embeddings": 32768,
102 "moe_intermediate_size": 896,
103 "n_group": 1,
104 "n_routed_experts": 64,
105 "n_shared_experts": 2,
106 "num_attention_heads": 10,
107 "num_experts_per_tok": 6,
108 "num_hidden_layers": 12,
109 "num_key_value_heads": 10,
110 "q_lora_rank": null,
111 "qk_nope_head_dim": 0,
112 "qk_rope_head_dim": 0,
113 "rm_head": false,
114 "topk_group": 1,
115 "topk_method": "greedy",
116 "use_mla": false,
117 "v_head_dim": 128,
118 "vocab_size": 129280,
119 "sliding_window_size": 128,
120 "sliding_window": 128
121 }