search

neonbjb / tortoise-tts

A multi-voice TTS system trained with an emphasis on quality
Apache License 2.0
12.35k stars 1.73k forks source link

run tortoise-tts on several gpus at once #649

Open MagicChakram opened 8 months ago

MagicChakram commented 8 months ago

For users with 2+ dedicated gpus in a system (nvidia sli) - is there a way to speed up the process by loading both gpus?

kevinamiri commented 8 months ago

Does attaching --use_deepspeed helps?

or 

api.TextToSpeech(use_deepspeed=True)
MagicChakram commented 8 months ago

Does attaching --use_deepspeed helps?

or

api.TextToSpeech(use_deepspeed=True)

according to docs it is more about speedup of process (which itself on windonws didn't work for me for varios reasons - can't compile lib etc) that is stull executed on a single device. What I'm looking for is to utilise more than one gpu at once in system. Or have it selected. For example defult param would be use_gpu = 0 (uses all available cards in system) or use_gpu = 1 (uses first card of N cards in system)

xiaoyangnihao commented 8 months ago

how you solve this? I want to train ar module with multi-gpus too.

MagicChakram commented 8 months ago

how you solve this? I want to train ar module with multi-gpus too.

unfortunately I have no solution to run this on specified gpus under windows or force it to use both installed dedicated cards in the system..