storytold/storyteller-ml
2
0
Fork 0
mirror of https://github.com/storytold/storyteller-ml.git synced 2026-10-15 00:03:06 +00:00
Code Issues Packages Projects Releases Wiki Activity

Have to replace base repo because this repo doesn't actually work.

Browse Source
This commit is contained in:
arensc
2024-03-08 21:48:08 -05:00
parent 42a881a3f7
commit 772832eed6
2 changed files with 6 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tts/StyleTTS2Train/README.md
+4
View File
@@ -4,6 +4,10 @@
d0ea889aa827b0c87e50ca44b302b030ccaf95e2
https://dagshub.com/StyleTTS/StyleTTS2/commits/main
To run:
accelerate launch --mixed_precision=fp16 --num_processes=1 train_finetune_accelerate.py --config_path ./Configs/config_harness.yml # use your own config but ensure you use train_finetune_accelerate.py
Should be accessible in our weights and biases account.
With a test harness that will be able easily observe the results to weights and biases.
Use python 3.10.12
tts/StyleTTS2Train/train_finetune_accelerate.py
+2 -1
View File
@@ -741,10 +741,11 @@ def main(config_path):
# writer.add_scalar('eval/mel_loss', loss_test / iters_test, epoch + 1)
# writer.add_scalar('eval/dur_loss', loss_test / iters_test, epoch + 1)
# writer.add_scalar('eval/F0_loss', loss_f / iters_test, epoch + 1)
harness.log_eval(mel_loss=float(loss_test / iters_test),dur_loss=float(loss_test / iters_test),F0_loss=float(loss_f / iters_test),epoch)
harness.log_eval(mel_loss=float(loss_test / iters_test),dur_loss=float(loss_test / iters_test),F0_loss=float(loss_f / iters_test),epoch=epoch)
accelerator.log({'eval/mel_loss': float(loss_test / iters_test),
'eval/dur_loss': float(loss_test / iters_test),
'eval/F0_loss': float(loss_f / iters_test)}, step=epoch + 1)
except ZeroDivisionError:
accelerator.print("Eval loss was divided by zero... skipping eval cycle")