fine tune

app.py

@@ -1,31 +1,62 @@
-import streamlit as st
-import whisper
-from tempfile import NamedTemporaryFile
-import ffmpeg
+# import streamlit as st
+# import whisper
+# from tempfile import NamedTemporaryFile
+# import ffmpeg
 
 
-st.title("MinuteBot App")
+# st.title("MinuteBot App")
 
-# upload audio file with streamlit
-audio_file = st.file_uploader("Unggah Meeting Audio", type=["mp3", "wav", "m4a"])
+# # upload audio file with streamlit
+# audio_file = st.file_uploader("Unggah Meeting Audio", type=["mp3", "wav", "m4a"])
 
-# model = whisper.load_model("base") # loading the base model
-st.text("MinuteBot Model telah dimuat:")
+# # model = whisper.load_model("base") # loading the base model
+# st.text("MinuteBot Model telah dimuat:")
 
-def load_whisper_model():
+# def load_whisper_model():
 
- return model
+# return model
 
  
+# if st.sidebar.button("Transkripsikan Audio"):
+# if audio_file is not None:
+# with NamedTemporaryFile() as temp:
+# temp.write(audio_file.getvalue())
+# temp.seek(0)
+# model = whisper.load_model("large")
+# result = model.transcribe(temp.name)
+# st.write(result["text"])
+ 
+# st.sidebar.header("Putar Berkas Audio")
+# st.sidebar.audio(audio_file)
+
+import streamlit as st
+from tempfile import NamedTemporaryFile
+import ffmpeg
+from transformers import Wav2Vec2ForCTC, Wav2Vec2Tokenizer
+import librosa
+
+st.title("TemplarX-Medium-Indonesian Transcription App")
+st.text("Model Whisper (TemplarX-medium-Indonesian) telah dimuat:")
+
+def load_whisper_model():
+ model_name = "jonnatakusuma/TemplarX-medium-Indonesian"
+ tokenizer = Wav2Vec2Tokenizer.from_pretrained(model_name)
+ model = Wav2Vec2ForCTC.from_pretrained(model_name)
+ return tokenizer, model
+
+audio_file = st.file_uploader("Unggah Meeting Audio", type=["mp3", "wav", "m4a"])
+
 if st.sidebar.button("Transkripsikan Audio"):
  if audio_file is not None:
  with NamedTemporaryFile() as temp:
- temp.write(audio_file.getvalue())
+ temp.write(audio_file.read())
  temp.seek(0)
- model = whisper.load_model("large")
- result = model.transcribe(temp.name)
- st.write(result["text"])
- 
-st.sidebar.header("Putar Berkas Audio")
-st.sidebar.audio(audio_file)
+ tokenizer, model = load_whisper_model()
+ # Read the audio file and transcribe using the fine-tuned model
+ audio_path = temp.name
+ audio_input, _ = librosa.load(audio_path, sr=16000)
+ transcription = model.stt(text)
+ st.write(transcription)
 
+st.sidebar.header("Putar Berkas Audio")
+st.sidebar.audio(audio_file, format='audio/wav')