add bhashini for hindi

app.py

@@ -3,24 +3,59 @@
 import gradio as gr
 import openai
 import os
+import torch
 
 
 
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+print("Device type:", device)
+
+from transformers import pipeline
+from transformers import (
+    AutoTokenizer,
+    WhisperProcessor,
+    WhisperForConditionalGeneration,
+)
+
 api_key = os.getenv('OPEN_API_KEY')
 openai.api_key = api_key
 
 
 global_history = [{"role": "assistant", "content": "Hi, I am a chatbot. I can converse in English. I can answer your questions about farming in India. Ask me anything!"}]
 
-
 from pydub import AudioSegment
 
+def get_asr_model_processor(model_id):
+    processor = WhisperProcessor.from_pretrained(model_id,model_max_length=225)
+    model = WhisperForConditionalGeneration.from_pretrained(model_id).to(device)
+    # model.forced_decoder_ids =None
+    model.config.max_new_token = 200
+    return {
+        "model": model,
+        "processor": processor
+    }
+model_proc_dict  = get_asr_model_processor("vasista22/whisper-hindi-large-v2")
+asr_pipe = pipeline("automatic-speech-recognition", model=model_proc_dict["model"], tokenizer=model_proc_dict["processor"].tokenizer, feature_extractor=model_proc_dict["processor"].feature_extractor,device=device)
+
+
+
+
 def get_asr_output(audio_path,lang):
     audio = AudioSegment.from_file(audio_path)
-    audio.export("temp.mp3", format="mp3")
-    file = open("temp.mp3","rb")
+    audio.export("temp.wav", format="wav")
+    file = open("temp.wav","rb")
+    
+    print(lang)
+    if lang == "hi":
+        op_text = asr_pipe("temp.wav") 
+        print('ai4bharat',op_text)    
+    
     transcription = openai.Audio.transcribe("whisper-1", file, language=lang)
-    return transcription.text
+    print('whisper',transcription)    
+    op_text = transcription.text
+    print(transcription)
+    
+    return op_text
 
 def add_text(history, audio_path,lang):
     global global_history
@@ -79,4 +114,4 @@ with gr.Blocks(title="Krishi GPT Demo") as demo:
     lang.change(clear_history, [lang], chatbot, queue=False)
 
 
-demo.launch(share=False)
+demo.launch(share=True)

requirements.txt

@@ -1,3 +1,3 @@
 openai==0.27.4
 gradio==3.21.0
-pydub==0.25.1
+pydub==0.25.1