Update app.py

app.py

@@ -5,7 +5,6 @@ from transformers import AutoTokenizer, AutoModelForCausalLM, GenerationConfig,B
 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")
@@ -19,17 +18,24 @@ 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):
+def generate_text(*args):
+    if args[0] == "":
+        return "Please input text"
+    args[1] = 0.6 if args[1] == 0 else args[1]
+    args[3] = 5 if args[3] == 0 else args[3]
+    args[4] = 0.5 if args[4] == 0 else args[4]
+    args[5] = 50 if args[5] == 0 else args[5]
+
+    generation_config = GenerationConfig(
+    penalty_alpha=args[1],
+    do_sample=args[2],
+    top_k=args[3],
+    temperature=args[4],
+    repetition_penalty=args[5],
+    max_new_tokens=args[6],
+    pad_token_id=tokenizer.eos_token_id
+    )
     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)
@@ -45,8 +51,7 @@ def recognize_speech(audio_data):
     recognizer = sr.Recognizer()
     try:
         text = recognizer.recognize_google(audio_data)
-        question_text = text
-        return f"Recognized Speech: {text}"
+        return text
 
     except sr.UnknownValueError:
         return "Speech Recognition could not understand audio."
@@ -65,13 +70,22 @@ def recognize_speech(audio_data):
 
 with gr.Blocks() as demo:
     with gr.Row():
-        gr.Label("Speech Recognition")
+        inp_text = gr.Textbox(label="Input Text")
         inp = gr.Audio(type="numpy")
         out_text_predict = gr.Textbox(label="Recognized Speech")
-        button = gr.Button("Recognize Speech")
+        button = gr.Button("Recognize Speech" , size="sm")
         button.click(recognize_speech, inp, out_text_predict)
+    with gr.Row():
+        with gr.Row():
+            penalty_alpha_slider = gr.Slider(minimum=0, maximum=1, step=0.1, label="penalty alpha")
+            do_sample_checkbox = gr.Checkbox(label="do sample")
+            top_k_slider = gr.Slider(minimum=0, maximum=10, step=1, label="top k")
+        with gr.Row():
+            temperature_slider = gr.Slider(minimum=0, maximum=1, step=0.1, label="temperature")
+            repetition_penalty_slider = gr.Slider(minimum=0, maximum=2, step=0.1, label="repetition penalty")
+            max_new_tokens_slider = gr.Slider(minimum=0, maximum=100, step=1, label="max new tokens")
     with gr.Row():
         out_answer = gr.Textbox(label="Answer")
-        button_answer = gr.Button("Generate Answer")
-        button_answer.click(generate_text, out_text_predict, out_answer)
+        button_answer = gr.Button("Answer")
+        button_answer.click(generate_text, [out_text_predict, penalty_alpha_slider, do_sample_checkbox, top_k_slider, temperature_slider, repetition_penalty_slider, max_new_tokens_slider], out_answer)
 demo.launch()