alihan/speech-to-speech
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Merge pull request #58 from huggingface/fix_stt_compile_mode

Browse Source 
Assigning min new tokens to a compiled whisper graph on a thread brea…
This commit is contained in:
Andrés Marafioti
2024-08-28 12:09:22 +02:00
committed by GitHub
parent fc9f960285 3b685e2629
commit feb12358ad
2 changed files with 2 additions and 7 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
STT/whisper_stt_handler.py
View File

@@ -66,8 +66,9 @@ class WhisperSTTHandler(BaseHandler):
if self.compile_mode not in (None, "default"): if self.compile_mode not in (None, "default"):
# generating more tokens than previously will trigger CUDA graphs capture # generating more tokens than previously will trigger CUDA graphs capture
# one should warmup with a number of generated tokens above max tokens targeted for subsequent generation # one should warmup with a number of generated tokens above max tokens targeted for subsequent generation
# hence, having min_new_tokens < max_new_tokens in the future doesn't make sense
warmup_gen_kwargs = { warmup_gen_kwargs = {
"min_new_tokens": self.gen_kwargs["min_new_tokens"], "min_new_tokens": self.gen_kwargs["max_new_tokens"], # Yes, assign max_new_tokens to min_new_tokens
"max_new_tokens": self.gen_kwargs["max_new_tokens"], "max_new_tokens": self.gen_kwargs["max_new_tokens"],
**self.gen_kwargs, **self.gen_kwargs,
} }

6
arguments_classes/whisper_stt_arguments.py
View File

@@ -33,12 +33,6 @@ class WhisperSTTHandlerArguments:
"help": "The maximum number of new tokens to generate. Default is 128." "help": "The maximum number of new tokens to generate. Default is 128."
}, },
) )
stt_gen_min_new_tokens: int = field(
default=0,
metadata={
"help": "The minimum number of new tokens to generate. Default is 0."
},
)
stt_gen_num_beams: int = field( stt_gen_num_beams: int = field(
default=1, default=1,
metadata={ metadata={