| {"name": "JNFGMCConfig", "n_modalities": 3, "latent_dim": 64, "input_dims": {"image": [3, 28, 28], "audio": [1, 32, 128], "trajectory": [200]}, "uses_likelihood_rescaling": true, "rescale_factors": null, "decoders_dist": {"image": "normal", "audio": "normal", "trajectory": "normal"}, "decoder_dist_params": {}, "custom_architectures": ["gmc_model", "decoders", "joint_encoder"], "warmup": 200, "annealing": false, "alpha": 0.1, "nb_epochs_gmc": 100, "beta": 1.0, "gmc_config": {"name": "GMCConfig", "n_modalities": 3, "input_dims": {"image": [3, 28, 28], "audio": [1, 32, 128], "trajectory": [200]}, "common_dim": 16, "latent_dim": 16, "temperature": 0.07, "custom_architectures": ["processors", "shared_encoder"], "loss": "between_modality_pairs"}, "logits_to_std": "standard"} |