Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Fixed some errors to make it easier for first-time users. #72

Open
wants to merge 2 commits into
base: main
Choose a base branch
from
Open

Fixed some errors to make it easier for first-time users. #72

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
30a6b89
Fix some errors
MingHui-Fang Jun 14, 2024
d29b2c1
Fix bugs
MingHui-Fang Jun 17, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 5 additions & 1 deletion helpers/model_init_scripts/init_dummy_model.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -57,7 +57,11 @@
model.generation_config.eos_token_id = encodec_vocab_size

# set other default generation config params
model.generation_config.max_length = int(30 * model.audio_encoder.config.frame_rate)
'''
In DAC, the 'model.audio_encoder.config.frame_rate' is 86
'model.generation_config.max_length' should be smaller than 'max_position_embeddings' in the decoder.
'''
model.generation_config.max_length = int(20 * model.audio_encoder.config.frame_rate)
model.generation_config.do_sample = True # True
model.generation_config.guidance_scale = 1 # 3.0

Expand Down
6 changes: 5 additions & 1 deletion helpers/model_init_scripts/init_dummy_model_with_encodec.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -54,7 +54,11 @@
model.generation_config.eos_token_id = encodec_vocab_size

# set other default generation config params
model.generation_config.max_length = int(30 * model.audio_encoder.config.frame_rate)
'''
In encodec, the 'model.audio_encoder.config.frame_rate' is 75
'model.generation_config.max_length' should be smaller than 'max_position_embeddings' in the decoder.
'''
model.generation_config.max_length = int(25 * model.audio_encoder.config.frame_rate)
model.generation_config.do_sample = True # True
model.generation_config.guidance_scale = 1 # 3.0

Expand Down
9 changes: 5 additions & 4 deletions training/run_parler_tts_training.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -43,7 +43,7 @@


from accelerate import Accelerator
from accelerate.utils import set_seed, AutocastKwargs, InitProcessGroupKwargs, TorchDynamoPlugin
from accelerate.utils import set_seed, AutocastKwargs, InitProcessGroupKwargs, TorchDynamoPlugin, DistributedDataParallelKwargs
from accelerate.utils.memory import release_memory

from parler_tts import (
Expand Down Expand Up @@ -97,7 +97,8 @@ def main():
padding = "max_length" if data_args.pad_to_max_length else "longest"

####### A. Preparation
kwargs_handlers = [InitProcessGroupKwargs(timeout=timedelta(minutes=60))]
kwargs_handlers = [InitProcessGroupKwargs(timeout=timedelta(minutes=60)),
DistributedDataParallelKwargs(find_unused_parameters=True)]

accelerator = Accelerator(
gradient_accumulation_steps=training_args.gradient_accumulation_steps,
Expand Down Expand Up @@ -432,8 +433,8 @@ def apply_audio_decoder(batch):

if accelerator.is_main_process:
lab = generate_labels["labels"].cpu().transpose(1, 2).to(torch.int16)
rat = generate_labels["ratio"].cpu().squeeze()
lens = generate_labels["len_audio"].cpu().squeeze()
rat = generate_labels["ratio"].cpu().reshape(-1)
lens = generate_labels["len_audio"].cpu().reshape(-1)
lab = [l[:, : int(ratio * length)] for (l, ratio, length) in zip(lab, rat, lens)]

all_generated_labels.extend(lab)
Expand Down