Soundboard / config.json
protodotdesign's picture
publish step 150,000 (Soundboard)
6c1cfa3 verified
{
"model": {
"n_updown_levels": 5,
"in_channels": 3,
"out_channels": 3,
"hidden_channels": [
128,
256,
512,
768,
1024,
2048
],
"emb_channels": 128,
"band_embedding_dim": 16,
"n_attn_heads": 8,
"attention_levels": [
3,
4
],
"use_attn_input_norm": true,
"num_res_blocks": 2
},
"diffusion": {
"beta_min": 0.0001,
"beta_max": 1.0,
"use_ot_ode": false,
"n_timestep_channels": 128
},
"data": {
"music_root": "../music",
"profiles_dir": "/root/locutius/corruption-chain/profiles",
"impulses_root": "/root/locutius/corruption-chain/impulses",
"crowd_root": "/root/locutius/corruption-chain/crowd_chunks",
"sounds_root": "/root/locutius/corruption-chain/sounds",
"mode_probabilities": null,
"profile_weight": 0.85,
"skip_unsupported_modes": true,
"val_fraction": 0.02,
"passes": 2,
"exclude_dirs": [],
"stereo_mode": "mid_side",
"s3": {
"bucket": "soundboard-training-data",
"prefix": "",
"endpoint_url": "https://s3.g.s4.mega.io",
"addressing_style": "virtual",
"region": "auto"
}
},
"optim": {
"lr": 1e-05,
"weight_decay": 0.0,
"grad_clip": 0.5,
"betas": [
0.9,
0.999
]
},
"output_dir": "/root/runs/v0.3-midside-mega",
"checkpoint_in": "/root/locutius/training/checkpoints/model.pt",
"t_range": [
0.0,
1.0
],
"seed": 42,
"max_steps": 150000,
"val_every_steps": 2000,
"save_every_steps": 2000,
"log_every_steps": 10,
"keep_last_checkpoints": 2,
"wandb_project": "soundboard",
"wandb_run_name": "v0.3-midside-mega",
"batch_size": 2,
"num_workers": 16,
"precision": "bf16",
"device": "cuda"
}