| { | |
| "model_type": "moondream3", | |
| "region_config": { | |
| "coord_feat_dim": 256, | |
| "coord_out_dim": 1024, | |
| "hidden_size": 2048, | |
| "model_type": "moondream3_region", | |
| "size_feat_dim": 512, | |
| "size_out_dim": 2048 | |
| }, | |
| "text_config": { | |
| "attention_bias": true, | |
| "attention_dropout": 0.0, | |
| "bos_token_id": 0, | |
| "eos_token_id": 0, | |
| "coord_token_id": 5, | |
| "head_dim": 64, | |
| "hidden_act": "gelu_pytorch_tanh", | |
| "hidden_size": 2048, | |
| "initializer_range": 0.02, | |
| "intermediate_size": 8192, | |
| "max_position_embeddings": 4096, | |
| "model_type": "moondream3_text", | |
| "moe_hidden_act": "gelu", | |
| "moe_intermediate_size": 1024, | |
| "moe_start_layer": 4, | |
| "num_attention_heads": 32, | |
| "num_experts": 64, | |
| "num_experts_per_tok": 8, | |
| "num_hidden_layers": 24, | |
| "num_key_value_heads": 32, | |
| "output_router_logits": false, | |
| "rms_norm_eps": 1e-05, | |
| "rope_parameters": { | |
| "rope_theta": 1500000.0, | |
| "rope_type": "default" | |
| }, | |
| "use_cache": false, | |
| "vocab_size": 51200 | |
| }, | |
| "tie_word_embeddings": false, | |
| "transformers_version": "5.0.0.dev0", | |
| "vision_config": { | |
| "attention_bias": true, | |
| "attention_dropout": 0.0, | |
| "crop_size": 378, | |
| "hidden_act": "gelu_pytorch_tanh", | |
| "hidden_size": 1152, | |
| "in_channels": 3, | |
| "initializer_range": 0.02, | |
| "intermediate_size": 4304, | |
| "max_crops": 12, | |
| "model_type": "moondream3_vision", | |
| "num_attention_heads": 16, | |
| "num_hidden_layers": 27, | |
| "overlap_margin": 4, | |
| "patch_size": 14, | |
| "prefix_len": 730, | |
| "proj_inner_dim": 8192, | |
| "proj_out_dim": 2048 | |
| } | |
| } | |