| { | |
| "architectures": [ | |
| "MolmoForConditionalGeneration" | |
| ], | |
| "image_token_index": 152069, | |
| "initializer_range": 0.02, | |
| "model_type": "molmo", | |
| "pad_token_id": 100277, | |
| "pooling_config": { | |
| "attention_dropout": 0.0, | |
| "head_dim": 64, | |
| "hidden_size": 2048, | |
| "image_feature_dropout": 0.0, | |
| "image_num_patches": 24, | |
| "image_padding_embed": "pad_and_partial_pad", | |
| "image_pooling_type": "attention_meanq", | |
| "initializer_range": 0.02, | |
| "model_type": "", | |
| "num_attention_heads": 16, | |
| "num_key_value_heads": 16, | |
| "pad_embed_dim": 2048, | |
| "pooling_height": 2, | |
| "pooling_width": 2, | |
| "projector_hidden_act": "silu", | |
| "text_hidden_size": 4096, | |
| "text_intermediate_size": 22016, | |
| "torch_dtype": "float32" | |
| }, | |
| "text_config": { | |
| "additional_embedding_size": 0, | |
| "attention_bias": false, | |
| "attention_dropout": 0.0, | |
| "head_dim": 128, | |
| "hidden_act": "swiglu", | |
| "hidden_size": 4096, | |
| "initializer_range": 0.02, | |
| "intermediate_size": 22016, | |
| "layer_norm_eps": 1e-06, | |
| "max_position_embeddings": 4096, | |
| "model_type": "molmo_text", | |
| "num_attention_heads": 32, | |
| "num_hidden_layers": 32, | |
| "num_key_value_heads": 32, | |
| "pad_token_id": 100277, | |
| "rope_scaling": null, | |
| "rope_theta": 500000.0, | |
| "sliding_window": 4096, | |
| "torch_dtype": "float32", | |
| "use_attention_layer_norm": true, | |
| "use_cache": true, | |
| "use_postnorm": true, | |
| "use_qk_norm": true, | |
| "vocab_size": 100284 | |
| }, | |
| "torch_dtype": "float32", | |
| "transformers_version": "4.52.0.dev0", | |
| "vision_config": { | |
| "attention_dropout": 0.0, | |
| "hidden_act": "quick_gelu", | |
| "hidden_size": 1024, | |
| "image_size": 576, | |
| "initializer_range": 0.02, | |
| "intermediate_size": 4096, | |
| "layer_norm_eps": 1e-05, | |
| "model_type": "molmo_vision_model", | |
| "num_attention_heads": 16, | |
| "num_hidden_layers": 23, | |
| "num_image_positions": 577, | |
| "num_key_value_heads": 16, | |
| "patch_size": 14, | |
| "torch_dtype": "float32" | |
| }, | |
| "vision_feature_layers": [ | |
| -2, | |
| -9 | |
| ], | |
| "vision_feature_select_strategy": "default", | |
| "vocab_size": 100284 | |
| } | |