| {"out_dir": "/root/autodl-tmp/OmniSep/omnisep/exp/vggsound/omnisep", "train_list": ["data/vggsound/test.csv"], "val_list": ["data/vggsound/test.csv"], "n_validation": null, "weights": null, "batch_size": 32, "drop_closest": null, "drop_closest_steps": 10000, "repeat": null, "frame_margin": null, "audio_only": false, "audio_len": 65535, "emb_dim": 1024, "audio_rate": 16000, "n_fft": 1024, "hop_len": 256, "win_len": 1024, "img_size": 224, "fps": 1, "train_mode": ["image", "text", "audio"], "n_mix": 2, "channels": 32, "layers": 7, "frames": 3, "stride_frames": 1, "binary_mask": true, "loss": "bce", "weighted_loss": true, "log_freq": true, "n_labels": null, "steps": 500000, "valid_steps": 10000, "lr": 0.001, "lr_warmup_steps": 5000, "lr_decay_steps": 100000, "lr_decay_multiplier": 0.1, "grad_norm_clip": 1.0, "pit_warmup_steps": 0, "seed": 1234, "gpus": 1, "workers": 20, "quiet": false, "is_feature": true, "is_neg": false, "feature_mode": "imagebind"} |