{ "architectures": [ "MolmoForConditionalGeneration" ], "image_token_index": 152069, "initializer_range": 0.02, "model_type": "molmo", "pad_token_id": 152070, "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": 8192, "text_intermediate_size": 59392, "torch_dtype": "bfloat16" }, "text_config": { "additional_embedding_size": 0, "attention_bias": true, "attention_dropout": 0.0, "head_dim": 128, "hidden_act": "swiglu", "hidden_size": 8192, "initializer_range": 0.02, "intermediate_size": 59392, "layer_norm_eps": 1e-05, "max_position_embeddings": 4096, "model_type": "molmo_text", "num_attention_heads": 64, "num_hidden_layers": 80, "num_key_value_heads": 8, "pad_token_id": 152070, "rope_scaling": null, "rope_theta": 1000000.0, "sliding_window": 4096, "torch_dtype": "bfloat16", "use_attention_layer_norm": false, "use_cache": true, "use_postnorm": false, "use_qk_norm": false, "vocab_size": 152071 }, "torch_dtype": "bfloat16", "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": "bfloat16" }, "vision_feature_layers": [ -2, -9 ], "vision_feature_select_strategy": "default", "vocab_size": 152071 }