mirror of
https://github.com/davesarmoury/GLaDOS.git
synced 2025-11-04 14:49:45 +08:00
100 lines
2.2 KiB
YAML
100 lines
2.2 KiB
YAML
# This config contains the default values for training HiFi-GAN model on LJSpeech dataset.
|
|
# If you want to train model on other dataset, you can change config values according to your dataset.
|
|
# Most dataset-specific arguments are in the head of the config file, see below.
|
|
|
|
name: "HifiGan"
|
|
|
|
train_dataset: ???
|
|
validation_datasets: ???
|
|
|
|
# Default values for dataset with sample_rate=22050
|
|
sample_rate: 22050
|
|
n_mel_channels: 80
|
|
n_window_size: 1024
|
|
n_window_stride: 256
|
|
n_fft: 1024
|
|
lowfreq: 0
|
|
highfreq: 8000
|
|
window: hann
|
|
|
|
train_n_segments: 8192
|
|
train_max_duration: null
|
|
train_min_duration: 0.75
|
|
|
|
val_n_segments: 66048
|
|
val_max_duration: null
|
|
val_min_duration: 3
|
|
|
|
defaults:
|
|
- model/generator: v1
|
|
- model/train_ds: train_ds
|
|
- model/validation_ds: val_ds
|
|
|
|
model:
|
|
preprocessor:
|
|
_target_: nemo.collections.asr.parts.preprocessing.features.FilterbankFeatures
|
|
nfilt: ${n_mel_channels}
|
|
lowfreq: ${lowfreq}
|
|
highfreq: ${highfreq}
|
|
n_fft: ${n_fft}
|
|
n_window_size: ${n_window_size}
|
|
n_window_stride: ${n_window_stride}
|
|
pad_to: 0
|
|
pad_value: -11.52
|
|
sample_rate: ${sample_rate}
|
|
window: ${window}
|
|
normalize: null
|
|
preemph: null
|
|
dither: 0.0
|
|
frame_splicing: 1
|
|
log: true
|
|
log_zero_guard_type: clamp
|
|
log_zero_guard_value: 1e-05
|
|
mag_power: 1.0
|
|
use_grads: false
|
|
exact_pad: true
|
|
|
|
optim:
|
|
_target_: torch.optim.AdamW
|
|
lr: 0.0002
|
|
betas: [0.8, 0.99]
|
|
|
|
sched:
|
|
name: CosineAnnealing
|
|
min_lr: 1e-5
|
|
warmup_ratio: 0.02
|
|
|
|
max_steps: 2500000
|
|
l1_loss_factor: 45
|
|
denoise_strength: 0.0025
|
|
|
|
trainer:
|
|
num_nodes: 1
|
|
devices: 1
|
|
accelerator: gpu
|
|
strategy: ddp_find_unused_parameters_true
|
|
precision: 32
|
|
max_steps: ${model.max_steps}
|
|
accumulate_grad_batches: 1
|
|
enable_checkpointing: False # Provided by exp_manager
|
|
logger: false # Provided by exp_manager
|
|
log_every_n_steps: 100
|
|
check_val_every_n_epoch: 10
|
|
benchmark: false
|
|
|
|
exp_manager:
|
|
exp_dir: null
|
|
name: ${name}
|
|
create_tensorboard_logger: true
|
|
create_checkpoint_callback: true
|
|
checkpoint_callback_params:
|
|
monitor: val_loss
|
|
mode: min
|
|
create_wandb_logger: false
|
|
wandb_logger_kwargs:
|
|
name: null
|
|
project: null
|
|
entity: null
|
|
resume_if_exists: false
|
|
resume_ignore_no_checkpoint: false
|