Spaces:
Sleeping
Sleeping
Commit
•
6a2808e
1
Parent(s):
fea335b
Changed the way the ASR model is being called. (whisper)
Browse files
app.py
CHANGED
@@ -37,29 +37,25 @@ Transcript: I need to buy a birthday
|
|
37 |
Prediction: Present, Gift, Cake, Card
|
38 |
Transcript: """
|
39 |
|
40 |
-
# whisper model specification
|
41 |
-
|
42 |
|
43 |
# openai.api_key = os.environ["Openai_APIkey"]
|
44 |
|
45 |
-
def debug_inference(audio, prompt, model, temperature, state=""):
|
46 |
-
# load audio data
|
47 |
-
audio = whisper.load_audio(audio)
|
48 |
-
# ensure sample is in correct format for inference
|
49 |
-
audio = whisper.pad_or_trim(audio)
|
50 |
|
51 |
-
# generate a log-mel spetrogram of the audio data
|
52 |
-
mel = whisper.log_mel_spectrogram(audio)
|
53 |
-
|
54 |
-
_, probs = model.detect_language(mel)
|
55 |
|
56 |
-
|
57 |
-
|
58 |
-
|
59 |
-
|
60 |
-
|
61 |
|
62 |
-
|
|
|
|
|
|
|
|
|
|
|
63 |
|
64 |
response = openai.Completion.create(
|
65 |
model=model,
|
@@ -82,7 +78,7 @@ def debug_inference(audio, prompt, model, temperature, state=""):
|
|
82 |
infers = list(map(lambda x: x.replace("\n", ""), temp))
|
83 |
#infered = list(map(lambda x: x.split(','), infers))
|
84 |
|
85 |
-
return
|
86 |
|
87 |
# get audio from microphone
|
88 |
gr.Interface(
|
|
|
37 |
Prediction: Present, Gift, Cake, Card
|
38 |
Transcript: """
|
39 |
|
40 |
+
# whisper model specification
|
41 |
+
asr_model = whisper.load_model("tiny")
|
42 |
|
43 |
# openai.api_key = os.environ["Openai_APIkey"]
|
44 |
|
|
|
|
|
|
|
|
|
|
|
45 |
|
|
|
|
|
|
|
|
|
46 |
|
47 |
+
# Transcribe function
|
48 |
+
def transcribe(audio_file):
|
49 |
+
print("Transcribing")
|
50 |
+
transcription = asr_model.transcribe(audio_file)["text"]
|
51 |
+
return transcription
|
52 |
|
53 |
+
def debug_inference(audio, prompt, model, temperature, state=""):
|
54 |
+
# Transcribe with Whisper
|
55 |
+
print("The audio is:", audio)
|
56 |
+
transcript = transcribe(audio)
|
57 |
+
|
58 |
+
text = prompt + transcript + "\nPrediction: "
|
59 |
|
60 |
response = openai.Completion.create(
|
61 |
model=model,
|
|
|
78 |
infers = list(map(lambda x: x.replace("\n", ""), temp))
|
79 |
#infered = list(map(lambda x: x.split(','), infers))
|
80 |
|
81 |
+
return transcript, state, infers, text
|
82 |
|
83 |
# get audio from microphone
|
84 |
gr.Interface(
|