Skip to content

Commit

Permalink
Merge pull request #166 from linto-ai/features/transformers
Browse files Browse the repository at this point in the history 
Enable transformers as a backend
  • Loading branch information
@Jeronymous
Jeronymous authored Feb 25, 2024
2 parents 713626e + 58909dc commit 79cc85e
Show file tree
Hide file tree
Showing 4 changed files with 554 additions and 28 deletions.
2 changes: 1 addition & 1 deletion setup.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -50,7 +50,7 @@
},
include_package_data=True,
extras_require={
'dev': ['matplotlib', 'transformers'],
'dev': ['matplotlib==3.7.4', 'transformers'],
'vad_silero': ['onnxruntime', 'torchaudio'],
'vad_auditok': ['auditok'],
'test': ['jsonschema'],
Expand Down
104 changes: 102 additions & 2 deletions tests/json_schema.json
View file
Original file line number Diff line number Diff line change
Expand Up @@ -39,8 +39,108 @@
},
"language": {"type": "string"},
"language_probs": {
"type": "array",
"items": {"type": "number", "minimum":0, "maximum":1}
"type": "object",
"properties": {
"en": {"type": "number", "minimum":0, "maximum":1},
"zh": {"type": "number", "minimum":0, "maximum":1},
"de": {"type": "number", "minimum":0, "maximum":1},
"es": {"type": "number", "minimum":0, "maximum":1},
"ru": {"type": "number", "minimum":0, "maximum":1},
"ko": {"type": "number", "minimum":0, "maximum":1},
"fr": {"type": "number", "minimum":0, "maximum":1},
"ja": {"type": "number", "minimum":0, "maximum":1},
"pt": {"type": "number", "minimum":0, "maximum":1},
"tr": {"type": "number", "minimum":0, "maximum":1},
"pl": {"type": "number", "minimum":0, "maximum":1},
"ca": {"type": "number", "minimum":0, "maximum":1},
"nl": {"type": "number", "minimum":0, "maximum":1},
"ar": {"type": "number", "minimum":0, "maximum":1},
"sv": {"type": "number", "minimum":0, "maximum":1},
"it": {"type": "number", "minimum":0, "maximum":1},
"id": {"type": "number", "minimum":0, "maximum":1},
"hi": {"type": "number", "minimum":0, "maximum":1},
"fi": {"type": "number", "minimum":0, "maximum":1},
"vi": {"type": "number", "minimum":0, "maximum":1},
"he": {"type": "number", "minimum":0, "maximum":1},
"uk": {"type": "number", "minimum":0, "maximum":1},
"el": {"type": "number", "minimum":0, "maximum":1},
"ms": {"type": "number", "minimum":0, "maximum":1},
"cs": {"type": "number", "minimum":0, "maximum":1},
"ro": {"type": "number", "minimum":0, "maximum":1},
"da": {"type": "number", "minimum":0, "maximum":1},
"hu": {"type": "number", "minimum":0, "maximum":1},
"ta": {"type": "number", "minimum":0, "maximum":1},
"no": {"type": "number", "minimum":0, "maximum":1},
"th": {"type": "number", "minimum":0, "maximum":1},
"ur": {"type": "number", "minimum":0, "maximum":1},
"hr": {"type": "number", "minimum":0, "maximum":1},
"bg": {"type": "number", "minimum":0, "maximum":1},
"lt": {"type": "number", "minimum":0, "maximum":1},
"la": {"type": "number", "minimum":0, "maximum":1},
"mi": {"type": "number", "minimum":0, "maximum":1},
"ml": {"type": "number", "minimum":0, "maximum":1},
"cy": {"type": "number", "minimum":0, "maximum":1},
"sk": {"type": "number", "minimum":0, "maximum":1},
"te": {"type": "number", "minimum":0, "maximum":1},
"fa": {"type": "number", "minimum":0, "maximum":1},
"lv": {"type": "number", "minimum":0, "maximum":1},
"bn": {"type": "number", "minimum":0, "maximum":1},
"sr": {"type": "number", "minimum":0, "maximum":1},
"az": {"type": "number", "minimum":0, "maximum":1},
"sl": {"type": "number", "minimum":0, "maximum":1},
"kn": {"type": "number", "minimum":0, "maximum":1},
"et": {"type": "number", "minimum":0, "maximum":1},
"mk": {"type": "number", "minimum":0, "maximum":1},
"br": {"type": "number", "minimum":0, "maximum":1},
"eu": {"type": "number", "minimum":0, "maximum":1},
"is": {"type": "number", "minimum":0, "maximum":1},
"hy": {"type": "number", "minimum":0, "maximum":1},
"ne": {"type": "number", "minimum":0, "maximum":1},
"mn": {"type": "number", "minimum":0, "maximum":1},
"bs": {"type": "number", "minimum":0, "maximum":1},
"kk": {"type": "number", "minimum":0, "maximum":1},
"sq": {"type": "number", "minimum":0, "maximum":1},
"sw": {"type": "number", "minimum":0, "maximum":1},
"gl": {"type": "number", "minimum":0, "maximum":1},
"mr": {"type": "number", "minimum":0, "maximum":1},
"pa": {"type": "number", "minimum":0, "maximum":1},
"si": {"type": "number", "minimum":0, "maximum":1},
"km": {"type": "number", "minimum":0, "maximum":1},
"sn": {"type": "number", "minimum":0, "maximum":1},
"yo": {"type": "number", "minimum":0, "maximum":1},
"so": {"type": "number", "minimum":0, "maximum":1},
"af": {"type": "number", "minimum":0, "maximum":1},
"oc": {"type": "number", "minimum":0, "maximum":1},
"ka": {"type": "number", "minimum":0, "maximum":1},
"be": {"type": "number", "minimum":0, "maximum":1},
"tg": {"type": "number", "minimum":0, "maximum":1},
"sd": {"type": "number", "minimum":0, "maximum":1},
"gu": {"type": "number", "minimum":0, "maximum":1},
"am": {"type": "number", "minimum":0, "maximum":1},
"yi": {"type": "number", "minimum":0, "maximum":1},
"lo": {"type": "number", "minimum":0, "maximum":1},
"uz": {"type": "number", "minimum":0, "maximum":1},
"fo": {"type": "number", "minimum":0, "maximum":1},
"ht": {"type": "number", "minimum":0, "maximum":1},
"ps": {"type": "number", "minimum":0, "maximum":1},
"tk": {"type": "number", "minimum":0, "maximum":1},
"nn": {"type": "number", "minimum":0, "maximum":1},
"mt": {"type": "number", "minimum":0, "maximum":1},
"sa": {"type": "number", "minimum":0, "maximum":1},
"lb": {"type": "number", "minimum":0, "maximum":1},
"my": {"type": "number", "minimum":0, "maximum":1},
"bo": {"type": "number", "minimum":0, "maximum":1},
"tl": {"type": "number", "minimum":0, "maximum":1},
"mg": {"type": "number", "minimum":0, "maximum":1},
"as": {"type": "number", "minimum":0, "maximum":1},
"tt": {"type": "number", "minimum":0, "maximum":1},
"haw": {"type": "number", "minimum":0, "maximum":1},
"ln": {"type": "number", "minimum":0, "maximum":1},
"ha": {"type": "number", "minimum":0, "maximum":1},
"ba": {"type": "number", "minimum":0, "maximum":1},
"jw": {"type": "number", "minimum":0, "maximum":1},
"su": {"type": "number", "minimum":0, "maximum":1}
}
}
}
}
4 changes: 3 additions & 1 deletion tests/test_transcribe.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -661,15 +661,17 @@ def test_hugging_face_model(self):
)

import tempfile
from transformers import WhisperForConditionalGeneration
from transformers import WhisperForConditionalGeneration, WhisperProcessor, GenerationConfig
tempfolder = os.path.join(tempfile.gettempdir(), "tmp_whisper-tiny-french-cased")

for safe_serialization in False, True,:
for max_shard_size in "100MB", "10GB", :
shutil.rmtree(tempfolder, ignore_errors=True)
model = WhisperForConditionalGeneration.from_pretrained("qanastek/whisper-tiny-french-cased")
processor = WhisperProcessor.from_pretrained("qanastek/whisper-tiny-french-cased")
try:
model.save_pretrained(tempfolder, safe_serialization=safe_serialization, max_shard_size=max_shard_size)
processor.save_pretrained(tempfolder)
self._test_cli_(
["--model", tempfolder, "--verbose", "True"],
"verbose", files=["bonjour.wav"], extensions=None,
Expand Down
Loading

0 comments on commit 79cc85e

Please sign in to comment.