Merge pull request #75 from Ph0rk0z/patch-2

Empty cuda cache between inferences.
This commit is contained in:
Puyuan Peng
2024-04-06 10:55:24 -05:00
committed by GitHub

View File

@@ -98,7 +98,9 @@ def inference_one_sample(model, model_args, phn2num, text_tokenizer, audio_token
gen_sample = audio_tokenizer.decode(
[(gen_frames, None)]
)
#Empty cuda cache between runs
if torch.cuda.is_available():
torch.cuda.empty_cache()
# return
return concat_sample, gen_sample