model_config.json

{"name": "JNFCLIPConfig", "n_modalities": 3, "latent_dim": 128, "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": ["CLIP_networks", "decoders", "joint_encoder"], "warmup": 200, "nb_epochs_clip": 50, "embedding_clip_dim": 64, "weights_clip": null, "beta": 0.5}