config.json
2.3 KB · 83 lines · json Raw
1 {
2 "_attn_implementation": "magi",
3 "_commit_hash": null,
4 "architectures": [
5 "LocateAnythingForConditionalGeneration"
6 ],
7 "auto_map": {
8 "AutoConfig": "configuration_locateanything.LocateAnythingConfig",
9 "AutoModel": "modeling_locateanything.LocateAnythingForConditionalGeneration"
10 },
11 "box_end_token_id": 151669,
12 "box_start_token_id": 151668,
13 "coord_end_token_id": 152677,
14 "coord_start_token_id": 151677,
15 "image_token_index": 151665,
16 "mlp_checkpoint": false,
17 "mlp_connector_layers": 2,
18 "model_type": "locateanything",
19 "none_token_id": 4064,
20 "ref_end_token_id": 151673,
21 "ref_start_token_id": 151672,
22 "template": null,
23 "text_config": {
24 "_attn_implementation_autoset": true,
25 "_name_or_path": "Qwen/Qwen2.5-3B-Instruct",
26 "architectures": [
27 "Qwen2ForCausalLM"
28 ],
29 "attention_dropout": 0.0,
30 "block_size": 6,
31 "bos_token_id": 151643,
32 "causal_attn": false,
33 "eos_token_id": 151645,
34 "hidden_act": "silu",
35 "hidden_size": 2048,
36 "initializer_range": 0.02,
37 "intermediate_size": 11008,
38 "max_position_embeddings": 32768,
39 "max_window_layers": 70,
40 "model_type": "qwen2",
41 "null_token_id": 152678,
42 "num_attention_heads": 16,
43 "num_hidden_layers": 36,
44 "num_key_value_heads": 2,
45 "rms_norm_eps": 1e-06,
46 "rope_scaling": null,
47 "rope_theta": 1000000.0,
48 "sliding_window": 32768,
49 "switch_token_id": 152679,
50 "text_mask_token_id": 151676,
51 "tie_word_embeddings": true,
52 "torch_dtype": "bfloat16",
53 "use_cache": false,
54 "use_sliding_window": false,
55 "vocab_size": 152681
56 },
57 "torch_dtype": "bfloat16",
58 "transformers_version": null,
59 "use_backbone_lora": 0,
60 "use_llm_lora": 0,
61 "vision_config": {
62 "_attn_implementation_autoset": true,
63 "_name_or_path": "moonshotai/MoonViT-SO-400M",
64 "auto_map": {
65 "AutoConfig": "moonshotai/MoonViT-SO-400M--configuration_moonvit.MoonViTConfig",
66 "AutoModel": "moonshotai/MoonViT-SO-400M--modeling_moonvit.MoonVitPretrainedModel"
67 },
68 "hidden_size": 1152,
69 "init_pos_emb_height": 64,
70 "init_pos_emb_width": 64,
71 "intermediate_size": 4304,
72 "merge_kernel_size": [
73 2,
74 2
75 ],
76 "model_type": "moonvit",
77 "num_attention_heads": 16,
78 "num_hidden_layers": 27,
79 "patch_size": 14,
80 "torch_dtype": "bfloat16"
81 }
82 }
83