| { |
| "run": { |
| "runs_root": "/home/vaclav/prog/personal/samuel/.claude/worktrees/rest-pose/runs", |
| "name": "s8-restpose-w01-100k", |
| "seed": 0 |
| }, |
| "data": { |
| "manifest_path": "/home/vaclav/prog/personal/samuel/.claude/worktrees/rest-pose/manifests/librilight_1000h.jsonl", |
| "sample_rate": 44100, |
| "chunk_seconds": 4.0, |
| "num_workers": 4, |
| "pitch_cache_path": "/home/vaclav/prog/personal/samuel/.claude/worktrees/rest-pose/manifests/pitch_cache/librilight_1000h_spf512.npz", |
| "target_rms": 0.05, |
| "val_fraction": 0.05 |
| }, |
| "model": { |
| "encoder": { |
| "channels": 1, |
| "dimension": 128, |
| "n_filters": 32, |
| "n_residual_layers": 3, |
| "ratios": [ |
| 8, |
| 5, |
| 4, |
| 2 |
| ], |
| "kernel_size": 7, |
| "last_kernel_size": 7, |
| "residual_kernel_size": 3, |
| "dilation_base": 2, |
| "pad_mode": "constant", |
| "compress": 2 |
| }, |
| "param_spec": { |
| "voiceness": [ |
| 0.0, |
| 1.0, |
| 0.6 |
| ], |
| "intensity": [ |
| 0.0, |
| 1.0, |
| 1.0 |
| ], |
| "tongueIndex": [ |
| 10.0, |
| 35.0, |
| 20.0 |
| ], |
| "tongueDiameter": [ |
| 1.5, |
| 3.5, |
| 2.4 |
| ], |
| "constrictionIndex": [ |
| 22.0, |
| 39.0, |
| 33.0 |
| ], |
| "constrictionDiameter": [ |
| -2.0, |
| 3.0, |
| 1.25 |
| ], |
| "lipDiameter": [ |
| 0.0, |
| 3.0, |
| 1.25 |
| ] |
| }, |
| "frozen_values": { |
| "vibratoWobble": 0.0, |
| "vibratoFrequency": 6.0, |
| "vibratoGain": 0.0, |
| "tractLength": 44.0 |
| }, |
| "samples_per_frame": 512, |
| "n_buckets": 32, |
| "gumbel_hard": false |
| }, |
| "synth": { |
| "ir_length": 256 |
| }, |
| "optim": { |
| "lr": 0.0003, |
| "betas": [ |
| 0.9, |
| 0.95 |
| ], |
| "weight_decay": 0.0, |
| "grad_clip": 1.0, |
| "max_steps": 100000, |
| "warmup_steps": 800, |
| "tau_start": 2.0, |
| "tau_end": 2.0, |
| "tau_anneal_steps": null |
| }, |
| "log": { |
| "wandb_project": "samuel", |
| "wandb_entity": "moboehle-kyutai", |
| "wandb_mode": "online", |
| "log_every": 200, |
| "eval_every": 1000, |
| "ckpt_every": 2000, |
| "ckpt_wandb_artifact": true, |
| "n_eval_clips": 200, |
| "eval_chunk_seconds": 10.0, |
| "n_audio_samples": 10, |
| "pitch_fmin": 70.0, |
| "pitch_fmax": 500.0, |
| "pitch_voiced_prob_threshold": 0.5, |
| "asr_whisper_size": "base" |
| }, |
| "loss": { |
| "mfcc": 1.0, |
| "mel": 0.0, |
| "stft": 0.0, |
| "entropy": 0.1, |
| "entropy_floor": 1.0, |
| "mfcc_n_fft": 2048, |
| "smooth": 0.3, |
| "smooth_weights": { |
| "tongueIndex": 1.0, |
| "constrictionIndex": 1.0, |
| "tongueDiameter": 0.3, |
| "constrictionDiameter": 0.3, |
| "lipDiameter": 0.3, |
| "voiceness": 0.1, |
| "intensity": 0.1 |
| }, |
| "accel": 0.3, |
| "accel_weights": { |
| "tongueIndex": 1.0, |
| "constrictionIndex": 1.0, |
| "tongueDiameter": 0.3, |
| "constrictionDiameter": 0.3, |
| "lipDiameter": 0.3, |
| "voiceness": 0.1, |
| "intensity": 0.1 |
| }, |
| "rest": 0.01, |
| "rest_targets": { |
| "lipDiameter": 0.3, |
| "tongueDiameter": 3.5, |
| "constrictionDiameter": 3.0 |
| }, |
| "rest_weights": { |
| "lipDiameter": 0.3, |
| "tongueDiameter": 1.0, |
| "constrictionDiameter": 1.0 |
| }, |
| "ssl": 1.0, |
| "ssl_model": "facebook/wav2vec2-base-960h", |
| "ssl_layer": 6, |
| "ssl_distance": "L1" |
| }, |
| "batch_size": 64 |
| } |
