update config
Browse files- config.json +117 -117
config.json
CHANGED
|
@@ -1,118 +1,118 @@
|
|
| 1 |
-
{
|
| 2 |
-
"_name_or_path": "deepseek-ai/DeepSeek-OCR",
|
| 3 |
-
"candidate_resolutions": [
|
| 4 |
-
[
|
| 5 |
-
1024,
|
| 6 |
-
1024
|
| 7 |
-
]
|
| 8 |
-
],
|
| 9 |
-
"global_view_pos": "head",
|
| 10 |
-
"architectures": [
|
| 11 |
-
"DeepseekOCRForCausalLM"
|
| 12 |
-
],
|
| 13 |
-
"auto_map": {
|
| 14 |
-
"AutoConfig": "modeling_deepseekocr.DeepseekOCRConfig",
|
| 15 |
-
"AutoModel": "modeling_deepseekocr.DeepseekOCRForCausalLM"
|
| 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": 2048,
|
| 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": "deeplip_b_l",
|
| 67 |
-
"model_type": "vision",
|
| 68 |
-
"width": {
|
| 69 |
-
"clip-l-14-224": {
|
| 70 |
-
"heads": 16,
|
| 71 |
-
"image_size": 224,
|
| 72 |
-
"layers": 24,
|
| 73 |
-
"patch_size": 14,
|
| 74 |
-
"width": 1024
|
| 75 |
-
},
|
| 76 |
-
"sam_vit_b": {
|
| 77 |
-
"downsample_channels": [
|
| 78 |
-
512,
|
| 79 |
-
1024
|
| 80 |
-
],
|
| 81 |
-
"global_attn_indexes": [
|
| 82 |
-
2,
|
| 83 |
-
5,
|
| 84 |
-
8,
|
| 85 |
-
11
|
| 86 |
-
],
|
| 87 |
-
"heads": 12,
|
| 88 |
-
"layers": 12,
|
| 89 |
-
"width": 768
|
| 90 |
-
}
|
| 91 |
-
}
|
| 92 |
-
},
|
| 93 |
-
"bos_token_id": 0,
|
| 94 |
-
"eos_token_id": 1,
|
| 95 |
-
"first_k_dense_replace": 1,
|
| 96 |
-
"hidden_size": 1280,
|
| 97 |
-
"intermediate_size": 6848,
|
| 98 |
-
"kv_lora_rank": null,
|
| 99 |
-
"lm_head": true,
|
| 100 |
-
"max_position_embeddings": 8192,
|
| 101 |
-
"moe_intermediate_size": 896,
|
| 102 |
-
"n_group": 1,
|
| 103 |
-
"n_routed_experts": 64,
|
| 104 |
-
"n_shared_experts": 2,
|
| 105 |
-
"num_attention_heads": 10,
|
| 106 |
-
"num_experts_per_tok": 6,
|
| 107 |
-
"num_hidden_layers": 12,
|
| 108 |
-
"num_key_value_heads": 10,
|
| 109 |
-
"q_lora_rank": null,
|
| 110 |
-
"qk_nope_head_dim": 0,
|
| 111 |
-
"qk_rope_head_dim": 0,
|
| 112 |
-
"rm_head": false,
|
| 113 |
-
"topk_group": 1,
|
| 114 |
-
"topk_method": "greedy",
|
| 115 |
-
"use_mla": false,
|
| 116 |
-
"v_head_dim": 0,
|
| 117 |
-
"vocab_size": 129280
|
| 118 |
}
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_name_or_path": "deepseek-ai/DeepSeek-OCR",
|
| 3 |
+
"candidate_resolutions": [
|
| 4 |
+
[
|
| 5 |
+
1024,
|
| 6 |
+
1024
|
| 7 |
+
]
|
| 8 |
+
],
|
| 9 |
+
"global_view_pos": "head",
|
| 10 |
+
"architectures": [
|
| 11 |
+
"DeepseekOCRForCausalLM"
|
| 12 |
+
],
|
| 13 |
+
"auto_map": {
|
| 14 |
+
"AutoConfig": "modeling_deepseekocr.DeepseekOCRConfig",
|
| 15 |
+
"AutoModel": "modeling_deepseekocr.DeepseekOCRForCausalLM"
|
| 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": 2048,
|
| 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": "deeplip_b_l",
|
| 67 |
+
"model_type": "vision",
|
| 68 |
+
"width": {
|
| 69 |
+
"clip-l-14-224": {
|
| 70 |
+
"heads": 16,
|
| 71 |
+
"image_size": 224,
|
| 72 |
+
"layers": 24,
|
| 73 |
+
"patch_size": 14,
|
| 74 |
+
"width": 1024
|
| 75 |
+
},
|
| 76 |
+
"sam_vit_b": {
|
| 77 |
+
"downsample_channels": [
|
| 78 |
+
512,
|
| 79 |
+
1024
|
| 80 |
+
],
|
| 81 |
+
"global_attn_indexes": [
|
| 82 |
+
2,
|
| 83 |
+
5,
|
| 84 |
+
8,
|
| 85 |
+
11
|
| 86 |
+
],
|
| 87 |
+
"heads": 12,
|
| 88 |
+
"layers": 12,
|
| 89 |
+
"width": 768
|
| 90 |
+
}
|
| 91 |
+
}
|
| 92 |
+
},
|
| 93 |
+
"bos_token_id": 0,
|
| 94 |
+
"eos_token_id": 1,
|
| 95 |
+
"first_k_dense_replace": 1,
|
| 96 |
+
"hidden_size": 1280,
|
| 97 |
+
"intermediate_size": 6848,
|
| 98 |
+
"kv_lora_rank": null,
|
| 99 |
+
"lm_head": true,
|
| 100 |
+
"max_position_embeddings": 8192,
|
| 101 |
+
"moe_intermediate_size": 896,
|
| 102 |
+
"n_group": 1,
|
| 103 |
+
"n_routed_experts": 64,
|
| 104 |
+
"n_shared_experts": 2,
|
| 105 |
+
"num_attention_heads": 10,
|
| 106 |
+
"num_experts_per_tok": 6,
|
| 107 |
+
"num_hidden_layers": 12,
|
| 108 |
+
"num_key_value_heads": 10,
|
| 109 |
+
"q_lora_rank": null,
|
| 110 |
+
"qk_nope_head_dim": 0,
|
| 111 |
+
"qk_rope_head_dim": 0,
|
| 112 |
+
"rm_head": false,
|
| 113 |
+
"topk_group": 1,
|
| 114 |
+
"topk_method": "greedy",
|
| 115 |
+
"use_mla": false,
|
| 116 |
+
"v_head_dim": 0,
|
| 117 |
+
"vocab_size": 129280
|
| 118 |
}
|