[DEFAULTS] #name of the run exp_name = F5 # the batch size batch_size = 8 # the chunk size max_frames = 3000 min_frames = 10 # number of CPU workers for the DataLoader num_workers = 4 # the random seed seed = 42 # Batches for gradient accumulation accum_batches = 1 # Number of steps between checkpoints checkpoint_every = 10000 # trainer checkpoint file to restart training from ckpt_path = '' # model checkpoint file to start a new training run from pretrained_ckpt_path = '' # Checkpoint path for the pretransform model if needed pretransform_ckpt_path = '' # configuration model specifying model hyperparameters model_config = '' # configuration for datasets dataset_config = '' # directory to save the checkpoints in save_dir = '' # grad norm max_grad_norm = 1.0 # grad accu grad_accumulation_steps = 1 # lr learning_rate = 7.5e-5 # epoch epochs = 110 # warmup steps num_warmup_updates = 2000 # save checkpoint per steps save_per_updates = 5000 # save last checkpoint per steps last_per_steps = 5000 prompt_path = "/mnt/sfs/music/lance/style-lance-full|/mnt/sfs/music/lance/style-lance-cnen-music-second" lrc_path = "/mnt/sfs/music/lance/lrc-lance-emb-full|/mnt/sfs/music/lance/lrc-lance-cnen-second" latent_path = "/mnt/sfs/music/lance/latent-lance|/mnt/sfs/music/lance/latent-lance-cnen-music-second-1|/mnt/sfs/music/lance/latent-lance-cnen-music-second-2" audio_drop_prob = 0.3 cond_drop_prob = 0.0 style_drop_prob = 0.1 lrc_drop_prob = 0.1 align_lyrics = 0 lyrics_slice = 0 parse_lyrics = 1 skip_empty_lyrics = 0 lyrics_shift = -1 use_style_prompt = 1 tokenizer_type = gpt2 reset_lr = 0 resumable_with_seed = 666 downsample_rate = 2048 grad_ckpt = 0 dataset_path = "/mnt/sfs/music/hkchen/workspace/F5-TTS-HW/filelists/music123latent_asred_bpmstyle_cnen_pure1" pure_prob = 0.0