Create app.py

app.py

@@ -0,0 +1,77 @@
+import speech_recognition as sr
+import gradio as gr
+import numpy as np
+from transformers import AutoTokenizer, AutoModelForCausalLM, GenerationConfig,BitsAndBytesConfig
+import torch
+
+Medical_finetunned_model = "truongghieu/deci-finetuned_Prj2"
+question_text = "This is a question"
+answer_text = "This is an answer"
+
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+bnb_config = BitsAndBytesConfig(
+        load_in_4bit=True, bnb_4bit_quant_type="nf4", bnb_4bit_compute_dtype="float16", bnb_4bit_use_double_quant=True
+    )
+tokenizer = AutoTokenizer.from_pretrained(Medical_finetunned_model, trust_remote_code=True)
+if torch.cuda.is_available():
+    model = AutoModelForCausalLM.from_pretrained(Medical_finetunned_model, trust_remote_code=True, quantization_config=bnb_config)
+else:
+    model = AutoModelForCausalLM.from_pretrained("truongghieu/deci-finetuned", trust_remote_code=True)
+
+
+generation_config = GenerationConfig(
+    penalty_alpha=0.6,
+    do_sample=True,
+    top_k=3,
+    temperature=0.5,
+    repetition_penalty=1.2,
+    max_new_tokens=50,
+    pad_token_id=tokenizer.eos_token_id
+)
+
+def generate_text(text):
+    input_text = f'###Human: \"{text}\"'
+    input_ids = tokenizer.encode(input_text, return_tensors="pt").to(device)
+    output_ids = model.generate(input_ids, generation_config=generation_config)
+    output_text = tokenizer.decode(output_ids[0], skip_special_tokens=True)
+    return output_text
+
+
+def recognize_speech(audio_data):
+
+    # return text
+    audio_data = sr.AudioData(np.array(audio_data[1]), sample_rate=audio_data[0] , sample_width=2)
+
+    recognizer = sr.Recognizer()
+    try:
+        text = recognizer.recognize_google(audio_data)
+        question_text = text
+        return f"Recognized Speech: {text}"
+
+    except sr.UnknownValueError:
+        return "Speech Recognition could not understand audio."
+    except sr.RequestError as e:
+        return f"Could not request results from Google Speech Recognition service; {e}"
+
+
+# audio_input = gr.Audio(type="numpy")
+
+# iface = gr.Interface(fn=recognize_speech, inputs=audio_input , outputs="text", title="Speech to Text")
+# # create a place to generate answer
+# answer_text = gr.Textbox(label="Answer")
+# iface_answer = gr.Interface(fn=generate_text, inputs=question_text , outputs="text", title="Answer")
+
+# iface.launch()
+
+with gr.Blocks as demo:
+    with gr.Row():
+        gr.Label("Speech Recognition")
+        inp = gr.Audio(type="numpy")
+        out_text_predict = gr.Textbox(label="Recognized Speech")
+        button = gr.Button(label="Recognize Speech", type="button")
+        button.click(recognize_speech, inp, out_text_predict)
+    with gr.Row():
+        out_answer = gr.Textbox(label="Answer")
+        button_answer = gr.Button(label="Generate Answer", type="button")
+        button_answer.click(generate_text, out_text_predict, out_answer)
+demo.launch()