DistilCodec-v1.0 / train_config.json
Ray0323's picture
Upload Model config and Train config
706d2f9 verified
{
"adam": {
"learning_rate": 0.0002,
"adam_b1": 0.5,
"adam_b2": 0.9,
"lr_decay": 0.98,
"weight_decay": 0.001
},
"wandb": {
"use_wandb": true,
"wandb_project": "audio-tokenizer-0927",
"wandb_group": "audio-tokenizer",
"wandb_team": "ccnl-s-team"
},
"mels_path": null,
"training_files_path": {
"replay_training_file_pathes": [
{
"path": "/cognitive_comp/common_data/audio/output/24k",
"replay_rate": 1.0
},
{
"path": "/cognitive_comp/common_data/audio/output/24k_data/24k_1_10s/stage2",
"replay_rate": 0.6
},
{
"path": "/cognitive_comp/common_data/audio/output/24k_data/24k_1_10s/stage3",
"replay_rate": 0.6
},
{
"path": "/cognitive_comp/common_data/audio/output/lam/asr_tts",
"replay_rate": 0.01
}
],
"current_training_file_pathes": [
"/cognitive_comp/common_data/audio/output/24k_data/24k_1_10s/stage4"
]
},
"validation_files_path": [
"/cognitive_comp/common_data/audio/output/dev",
"/cognitive_comp/common_data/audio/output/24k_data/dev/stage2",
"/cognitive_comp/common_data/audio/output/24k_data/dev/stage3",
"/cognitive_comp/common_data/audio/output/24k_data/dev/stage4"
],
"save_path": "/cognitive_comp/wangrui/data/lam0927",
"distill": {
"is_distill": true,
"use_fm_distill": false,
"quantizer_transfer": false,
"teacher_ckpt_path": "/cognitive_comp/wangrui/data/lam0923/saved_ckpt"
},
"training_epochs": 50,
"checkpoint_interval": 1000,
"validation_interval": 2000,
"accumulation_steps": 1,
"summary_interval": 10,
"stdout_interval": 10,
"num_ckpt_keep": 30,
"fine_tuning": false,
"num_gpus": 8,
"batch_size": 152,
"seed": 1234,
"sampling_rate": 24000,
"codebook_loss_lambda": 1.0,
"commitment_loss_lambda": 0.25,
"num_workers": 4
}