{ "architectures": [ "Cosmos3EdgeForConditionalGeneration" ], "dtype": "float32", "image_token_id": 3, "model_type": "cosmos3_edge", "pad_token_id": 11, "projector_hidden_size": 64, "text_config": { "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 1, "eos_token_id": 11, "head_dim": 8, "hidden_act": "relu2", "hidden_size": 32, "initializer_range": 0.02, "intermediate_size": 64, "max_position_embeddings": 200, "mlp_bias": false, "model_type": "cosmos3_edge_text", "num_attention_heads": 4, "num_hidden_layers": 2, "num_key_value_heads": 2, "pad_token_id": 11, "pretraining_tp": 1, "rms_norm_eps": 1e-05, "rope_parameters": { "mrope_section": [ 2, 1, 1 ], "rope_theta": 100000000, "rope_type": "default" }, "tie_word_embeddings": false, "use_cache": true, "vocab_size": 131072 }, "tie_word_embeddings": false, "transformers_version": "5.16.0.dev0", "video_token_id": 4, "vision_config": { "attention_dropout": 0.0, "hidden_act": "gelu_pytorch_tanh", "hidden_size": 32, "intermediate_size": 64, "layer_norm_eps": 1e-06, "model_type": "cosmos3_edge_vision", "num_attention_heads": 4, "num_channels": 3, "num_hidden_layers": 2, "num_patches": 4, "patch_size": 2, "spatial_merge_size": 2 }, "vision_end_token_id": 6, "vision_start_token_id": 5 }