Spaces:
Runtime error
Runtime error
MorenoLQ
commited on
Commit
·
7fce27b
1
Parent(s):
856bef6
Updated for file upload and missing inputs
Browse files- app.py +66 -45
- demo_example_1.mp3 +0 -0
- gradio_queue.db +0 -0
app.py
CHANGED
|
@@ -25,61 +25,78 @@ DICT_MODELS = {
|
|
| 25 |
MODELS = sorted(DICT_MODELS.keys())
|
| 26 |
CACHED_MODELS_BY_ID = {}
|
| 27 |
|
| 28 |
-
def
|
| 29 |
-
|
| 30 |
-
|
| 31 |
-
|
| 32 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 33 |
|
|
|
|
|
|
|
| 34 |
model = DICT_MODELS.get(model_name)
|
| 35 |
-
|
| 36 |
-
|
| 37 |
-
|
| 38 |
-
"error_message": f"Model size {model_size} not found for {language} language :("
|
| 39 |
-
})
|
| 40 |
-
elif decoding_type == "Guided by Language Model" and not model["has_lm"]:
|
| 41 |
history.append({
|
| 42 |
-
"
|
|
|
|
|
|
|
|
|
|
| 43 |
})
|
| 44 |
else:
|
| 45 |
|
| 46 |
-
|
| 47 |
-
|
| 48 |
-
if model_instance is None:
|
| 49 |
-
model_instance = AutoModelForCTC.from_pretrained(model["model_id"])
|
| 50 |
-
CACHED_MODELS_BY_ID[model["model_id"]] = model_instance
|
| 51 |
|
| 52 |
-
|
| 53 |
-
|
| 54 |
-
|
| 55 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 56 |
else:
|
| 57 |
-
processor = Wav2Vec2Processor.from_pretrained(model["model_id"])
|
| 58 |
-
asr = pipeline("automatic-speech-recognition", model=model_instance, tokenizer=processor.tokenizer,
|
| 59 |
-
feature_extractor=processor.feature_extractor, decoder=None)
|
| 60 |
|
| 61 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 62 |
|
| 63 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 64 |
|
| 65 |
-
|
| 66 |
-
"model_id": model["model_id"],
|
| 67 |
-
"decoding_type": decoding_type,
|
| 68 |
-
"transcription": transcription,
|
| 69 |
-
"error_message": None
|
| 70 |
-
})
|
| 71 |
|
| 72 |
-
|
| 73 |
-
|
| 74 |
-
|
| 75 |
-
|
| 76 |
-
|
| 77 |
-
|
| 78 |
-
|
| 79 |
-
|
| 80 |
-
|
| 81 |
-
|
| 82 |
-
html_output += "</div>"
|
| 83 |
|
| 84 |
return html_output, history
|
| 85 |
|
|
@@ -87,7 +104,8 @@ def run(input_file, model_name, decoding_type, history):
|
|
| 87 |
gr.Interface(
|
| 88 |
run,
|
| 89 |
inputs=[
|
| 90 |
-
gr.inputs.Audio(source="
|
|
|
|
| 91 |
gr.inputs.Radio(label="Model", choices=MODELS),
|
| 92 |
gr.inputs.Radio(label="Decoding type", choices=["Standard", "Guided by Language Model"]),
|
| 93 |
"state"
|
|
@@ -106,5 +124,8 @@ gr.Interface(
|
|
| 106 |
""",
|
| 107 |
allow_screenshot=False,
|
| 108 |
allow_flagging="never",
|
| 109 |
-
theme="huggingface"
|
|
|
|
|
|
|
|
|
|
| 110 |
).launch(enable_queue=True)
|
|
|
|
| 25 |
MODELS = sorted(DICT_MODELS.keys())
|
| 26 |
CACHED_MODELS_BY_ID = {}
|
| 27 |
|
| 28 |
+
def build_html(history):
|
| 29 |
+
html_output = "<div class='result'>"
|
| 30 |
+
for item in history:
|
| 31 |
+
if item["error_message"] is not None:
|
| 32 |
+
html_output += f"<div class='result_item result_item_error'>{item['error_message']}</div>"
|
| 33 |
+
else:
|
| 34 |
+
url_suffix = " + Guided by Language Model" if item["decoding_type"] == "Guided by Language Model" else ""
|
| 35 |
+
html_output += "<div class='result_item result_item_success'>"
|
| 36 |
+
html_output += f'<strong><a target="_blank" href="https://huggingface.co/{item["model_id"]}">{item["model_id"]}{url_suffix}</a></strong><br/><br/>'
|
| 37 |
+
html_output += f'{item["transcription"]}<br/>'
|
| 38 |
+
html_output += "</div>"
|
| 39 |
+
html_output += "</div>"
|
| 40 |
+
return html_output
|
| 41 |
|
| 42 |
+
def run(uploaded_file, input_file, model_name, decoding_type, history):
|
| 43 |
+
|
| 44 |
model = DICT_MODELS.get(model_name)
|
| 45 |
+
history = history or []
|
| 46 |
+
|
| 47 |
+
if uploaded_file is None and input_file is None:
|
|
|
|
|
|
|
|
|
|
| 48 |
history.append({
|
| 49 |
+
"model_id": model["model_id"],
|
| 50 |
+
"decoding_type": decoding_type,
|
| 51 |
+
"transcription": "",
|
| 52 |
+
"error_message": "No input provided."
|
| 53 |
})
|
| 54 |
else:
|
| 55 |
|
| 56 |
+
if input_file is None:
|
| 57 |
+
input_file = uploaded_file
|
|
|
|
|
|
|
|
|
|
| 58 |
|
| 59 |
+
logger.info(f"Running ASR {model_name}-{decoding_type} for {input_file}")
|
| 60 |
+
|
| 61 |
+
history = history or []
|
| 62 |
+
|
| 63 |
+
if model is None:
|
| 64 |
+
history.append({
|
| 65 |
+
"error_message": f"Model size {model_size} not found for {language} language :("
|
| 66 |
+
})
|
| 67 |
+
elif decoding_type == "Guided by Language Model" and not model["has_lm"]:
|
| 68 |
+
history.append({
|
| 69 |
+
"error_message": f"LM not available for {language} language :("
|
| 70 |
+
})
|
| 71 |
else:
|
|
|
|
|
|
|
|
|
|
| 72 |
|
| 73 |
+
# model_instance = AutoModelForCTC.from_pretrained(model["model_id"])
|
| 74 |
+
model_instance = CACHED_MODELS_BY_ID.get(model["model_id"], None)
|
| 75 |
+
if model_instance is None:
|
| 76 |
+
model_instance = AutoModelForCTC.from_pretrained(model["model_id"])
|
| 77 |
+
CACHED_MODELS_BY_ID[model["model_id"]] = model_instance
|
| 78 |
|
| 79 |
+
if decoding_type == "Guided by Language Model":
|
| 80 |
+
processor = Wav2Vec2ProcessorWithLM.from_pretrained(model["model_id"])
|
| 81 |
+
asr = pipeline("automatic-speech-recognition", model=model_instance, tokenizer=processor.tokenizer,
|
| 82 |
+
feature_extractor=processor.feature_extractor, decoder=processor.decoder)
|
| 83 |
+
else:
|
| 84 |
+
processor = Wav2Vec2Processor.from_pretrained(model["model_id"])
|
| 85 |
+
asr = pipeline("automatic-speech-recognition", model=model_instance, tokenizer=processor.tokenizer,
|
| 86 |
+
feature_extractor=processor.feature_extractor, decoder=None)
|
| 87 |
|
| 88 |
+
transcription = asr(input_file, chunk_length_s=5, stride_length_s=1)["text"]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 89 |
|
| 90 |
+
logger.info(f"Transcription for {input_file}: {transcription}")
|
| 91 |
+
|
| 92 |
+
history.append({
|
| 93 |
+
"model_id": model["model_id"],
|
| 94 |
+
"decoding_type": decoding_type,
|
| 95 |
+
"transcription": transcription,
|
| 96 |
+
"error_message": None
|
| 97 |
+
})
|
| 98 |
+
|
| 99 |
+
html_output = build_html(history)
|
|
|
|
| 100 |
|
| 101 |
return html_output, history
|
| 102 |
|
|
|
|
| 104 |
gr.Interface(
|
| 105 |
run,
|
| 106 |
inputs=[
|
| 107 |
+
gr.inputs.Audio(source="upload", type='filepath', optional=True),
|
| 108 |
+
gr.inputs.Audio(source="microphone", type="filepath", label="Record something...", optional=True),
|
| 109 |
gr.inputs.Radio(label="Model", choices=MODELS),
|
| 110 |
gr.inputs.Radio(label="Decoding type", choices=["Standard", "Guided by Language Model"]),
|
| 111 |
"state"
|
|
|
|
| 124 |
""",
|
| 125 |
allow_screenshot=False,
|
| 126 |
allow_flagging="never",
|
| 127 |
+
theme="huggingface",
|
| 128 |
+
examples = [
|
| 129 |
+
['demo_example_1.mp3', 'demo_example_1.mp3', 'robust-300m', 'Guided by Language Model']
|
| 130 |
+
]
|
| 131 |
).launch(enable_queue=True)
|
demo_example_1.mp3
ADDED
|
Binary file (121 kB). View file
|
|
|
gradio_queue.db
CHANGED
|
Binary files a/gradio_queue.db and b/gradio_queue.db differ
|
|
|