| { | |
| "config": { | |
| "run": { | |
| "name": "synthetic_alpha_stable_mixture", | |
| "seed": 0, | |
| "dtype": "float64" | |
| }, | |
| "dataset": { | |
| "kind": "synthetic", | |
| "name": "unbalanced_highdim_alpha_stable_mixture", | |
| "params": { | |
| "n_samples": 36000, | |
| "alpha": 1.7, | |
| "dim": 150, | |
| "n_modes": 16, | |
| "rank": 6, | |
| "imbalance_tau": 1.2, | |
| "mean_scale": 7.5, | |
| "base_scale": 0.55, | |
| "anisotropy": 1.0, | |
| "structure_seed": 0 | |
| }, | |
| "split": { | |
| "val_size": 0.027777777777777776, | |
| "test_size": 0.2777777777777778, | |
| "random_state": 0, | |
| "standardize": false | |
| } | |
| }, | |
| "network": { | |
| "name": "mlp", | |
| "params": { | |
| "width": 240, | |
| "depth": 4, | |
| "time_dim": 128, | |
| "dropout": 0.0, | |
| "use_norm": true | |
| } | |
| }, | |
| "models": { | |
| "ddpm_v": { | |
| "params": { | |
| "n_steps": 128, | |
| "sigma_max": 5.0, | |
| "sampler": "ddpm" | |
| } | |
| }, | |
| "dlpm_eps": { | |
| "params": { | |
| "n_steps": 128, | |
| "alpha": 1.8, | |
| "sampler": "native" | |
| } | |
| }, | |
| "tedm_origin": { | |
| "params": { | |
| "n_steps": 64, | |
| "nu": 2.2, | |
| "sigma_min": 0.005, | |
| "sigma_max": 5.0, | |
| "sigma_data": 1.0, | |
| "solver": "edm_stochastic_heun" | |
| } | |
| } | |
| }, | |
| "pilot": { | |
| "n_trials": 3, | |
| "train": { | |
| "n_epochs": 96, | |
| "early_stopping_patience": 32, | |
| "freq_logging": 16, | |
| "stats_freq_epochs": 1 | |
| }, | |
| "sweep": { | |
| "ddpm_v": { | |
| "model": { | |
| "sigma_max": [ | |
| 2.0, | |
| 5.0, | |
| 8.0 | |
| ] | |
| }, | |
| "train": { | |
| "lr": [ | |
| 0.0001, | |
| 0.0002, | |
| 0.0005 | |
| ] | |
| } | |
| }, | |
| "dlpm_eps": { | |
| "train": { | |
| "lr": [ | |
| 0.0001, | |
| 0.0002, | |
| 0.0005 | |
| ] | |
| } | |
| }, | |
| "tedm_origin": { | |
| "train": { | |
| "lr": [ | |
| 0.0001, | |
| 0.0002, | |
| 0.0005 | |
| ] | |
| } | |
| } | |
| } | |
| }, | |
| "train": { | |
| "device": "cuda", | |
| "data_device": "cpu", | |
| "batch_size": 1024, | |
| "n_epochs": 640, | |
| "early_stopping_patience": 100, | |
| "lr": 0.0005, | |
| "num_workers": 0, | |
| "use_adamw": true, | |
| "weight_decay": 1e-06, | |
| "grad_clip_norm": 10.0, | |
| "lr_schedule": "cosine", | |
| "warmup_steps": 400, | |
| "cosine_eta_min_ratio": 0.05, | |
| "freq_logging": 50, | |
| "stats_freq_epochs": 1, | |
| "log_grad_norm": true, | |
| "ckpt_freq_epochs": 64, | |
| "ckpt_keep_last": 8 | |
| }, | |
| "test": { | |
| "batch_size": 1024, | |
| "loss_repeats": 4, | |
| "n_samples": 32 | |
| }, | |
| "huggingface": { | |
| "namespace": "Diffusion-Research-Lab", | |
| "repository": "alpha-stable-mixture-1p7-150d-diffusion" | |
| } | |
| }, | |
| "model_name": "tedm_origin", | |
| "model_params": { | |
| "n_steps": 64, | |
| "nu": 2.2, | |
| "sigma_min": 0.005, | |
| "sigma_max": 5.0, | |
| "sigma_data": 1.0, | |
| "solver": "edm_stochastic_heun" | |
| }, | |
| "sample_shape": [ | |
| 150 | |
| ], | |
| "data_scale": 0.6885142523782495 | |
| } | |