Upload 2 files

app.py

@@ -0,0 +1,43 @@
+import streamlit as st
+import torch
+from transformers import WhisperForConditionalGeneration, WhisperProcessor
+from peft import PeftModel, PeftConfig
+import librosa
+
+# Model sozlamalari
+peft_model_id = "Elyordev/fine_tune_whisper_uzbek"
+language = "Uzbek"
+task = "transcribe"
+
+# PEFT konfiguratsiyasini yuklash
+peft_config = PeftConfig.from_pretrained(peft_model_id)
+
+# CPU uchun model yuklash
+model = WhisperForConditionalGeneration.from_pretrained(
+    peft_config.base_model_name_or_path, 
+    device_map="cpu"
+)
+model = PeftModel.from_pretrained(model, peft_model_id)
+
+# Tokenizer va Processor sozlash
+processor = WhisperProcessor.from_pretrained(peft_config.base_model_name_or_path, language=language, task=task)
+
+# Streamlit interfeysi
+st.title("Uzbek Whisper STT Hugging Face Spaces App")
+st.write("Fine-tuned Whisper model for Uzbek speech recognition. Upload your audio to get the transcription.")
+
+# Audio yuklash
+uploaded_file = st.file_uploader("Ovozli fayl yuklang", type=["wav", "mp3", "m4a"])
+
+def transcribe(audio_file):
+    audio, sr = librosa.load(audio_file, sr=16000)
+    inputs = processor(audio, sampling_rate=16000, return_tensors="pt").input_features
+    predicted_ids = model.generate(inputs, forced_decoder_ids=processor.get_decoder_prompt_ids(language="uz", task="transcribe"))
+    transcription = processor.batch_decode(predicted_ids, skip_special_tokens=True)[0]
+    return transcription
+
+if uploaded_file:
+    st.audio(uploaded_file, format="audio/wav")
+    st.write("**Transkripsiya natijasi:**")
+    transcription = transcribe(uploaded_file)
+    st.success(transcription)

requirements.txt

@@ -0,0 +1,12 @@
+git+https://github.com/huggingface/transformers.git@main
+git+https://github.com/huggingface/peft.git@main
+torch
+gradio
+huggingface_hub
+librosa
+evaluate>=0.4.3
+jiwer
+bitsandbytes
+datasets
+accelerate
+loralib