MithrilMan / F5-TTS

Official code for "F5-TTS: A Fairytaler that Fakes Fluent and Faithful Speech with Flow Matching"

https://arxiv.org/abs/2410.06885

MIT License

0 stars 0 forks source link

perf. notes #1

Open MithrilMan opened 20 hours ago

MithrilMan commented 20 hours ago

Checks

[X] This template is only for question, not feature requests or bug reports.
[X] I have thoroughly reviewed the project documentation and read the related paper(s).
[X] I have searched for existing issues, including closed ones, no similar questions.
[X] I confirm that I am using English to submit this report in order to facilitate communication.

Question details

MithrilMan commented 20 hours ago

Memory Usage:

dataset 9h

89 steps/epoch epoch 1 = 02:13 min epoch 2 = 02:15 min

MithrilMan commented 20 hours ago

same without adam

accelerate launch /workspace/F5-TTS/src/f5_tts/train/finetune_cli.py --exp_name F5TTS_Base --learning_rate 1e-05 --batch_size_per_gpu 35000 --batch_size_type frame --max_samples 64 --grad_accumulation_steps 1 --max_grad_norm 1 --epochs 316 --num_warmup_updates 108 --save_per_updates 200 --last_per_steps 120 --dataset_name italian9h --finetune True --tokenizer char --log_samples True --logger wandb

stessi tempi 89 steps/epoch epoch 1: 02:13 epoch 2: 02:16