| {"train_file": "data/metadata_train_makam.json", "validation_file": "data/metadata_val_makam.json", "validation_file2": "data/metadata_test_makam.json", "test_file": "data/metadata_test_makam.json", "num_examples": -1, "text_encoder_name": "google/flan-t5-large", "scheduler_name": "stabilityai/stable-diffusion-2-1", "unet_model_name": null, "unet_model_config": "configs/diffusion_model_config_munet.json", "hf_model": null, "snr_gamma": 5.0, "freeze_text_encoder": true, "text_column": "main_caption", "text2_column": "alt_caption", "audio_column": "location", "beats_column": "beats", "chords_column": "chords", "chords_time_column": "chords_time", "uncondition": false, "uncondition_all": true, "uncondition_single": false, "drop_sentences": true, "random_pick_text_column": true, "model_type": "Mustango", "prefix": null, "per_device_train_batch_size": 2, "per_device_eval_batch_size": 4, "learning_rate": 4.5e-05, "weight_decay": 1e-08, "num_train_epochs": 100, "max_train_steps": null, "gradient_accumulation_steps": 4, "lr_scheduler_type": "linear", "num_warmup_steps": 0, "adam_beta1": 0.9, "adam_beta2": 0.999, "adam_weight_decay": 0.0001, "adam_epsilon": 1e-08, "output_dir": "saved/1743793947", "seed": 1234, "checkpointing_steps": "best", "save_every": 10, "resume_from_checkpoint": null, "with_tracking": false, "report_to": "all"} | |
| {"epoch": 1, "step": 2356, "train_loss": 0.7477, "val_loss": 0.6585, "val_loss2": 0.657} | |
| {"epoch": 2, "step": 4712, "train_loss": 0.7414, "val_loss": 0.6555, "val_loss2": 0.6554} | |
| {"epoch": 3, "step": 7068, "train_loss": 0.7397, "val_loss": 0.6555, "val_loss2": 0.6541} | |
| {"epoch": 4, "step": 9424, "train_loss": 0.7382, "val_loss": 0.6536, "val_loss2": 0.6532} | |
| {"epoch": 5, "step": 11780, "train_loss": 0.7384, "val_loss": 0.6538, "val_loss2": 0.6533} | |
| {"epoch": 6, "step": 14136, "train_loss": 0.7358, "val_loss": 0.6533, "val_loss2": 0.6524} | |
| {"epoch": 7, "step": 16492, "train_loss": 0.738, "val_loss": 0.6523, "val_loss2": 0.6522} | |
| {"epoch": 8, "step": 18848, "train_loss": 0.7383, "val_loss": 0.6533, "val_loss2": 0.6516} | |
| {"epoch": 9, "step": 21204, "train_loss": 0.7368, "val_loss": 0.6519, "val_loss2": 0.6515} | |
| {"epoch": 10, "step": 23560, "train_loss": 0.734, "val_loss": 0.6517, "val_loss2": 0.6511} | |
| {"epoch": 11, "step": 25916, "train_loss": 0.7363, "val_loss": 0.652, "val_loss2": 0.6507} | |
| {"epoch": 12, "step": 28272, "train_loss": 0.734, "val_loss": 0.6503, "val_loss2": 0.6502} | |