Create app.py

app.py

@@ -0,0 +1,90 @@
+import torch
+from transformers import AutoModelForCausalLM, AutoTokenizer, Trainer, TrainingArguments
+import gradio as gr
+from datasets import load_dataset
+
+# Load the GPT-J model and tokenizer from Hugging Face
+model_name = "EleutherAI/gpt-j-6B"  # GPT-J model
+tokenizer = AutoTokenizer.from_pretrained(model_name)
+model = AutoModelForCausalLM.from_pretrained(model_name)
+
+# Define a function for generating responses
+def generate_response(prompt):
+    # Tokenize the input text
+    inputs = tokenizer(prompt, return_tensors="pt")
+    
+    # Generate response using the model
+    outputs = model.generate(
+        inputs.input_ids, 
+        max_length=200,      # Adjust response length
+        do_sample=True,      # Sample from the model
+        temperature=0.7,     # Adjust creativity (lower = more conservative)
+        top_p=0.9,           # Nucleus sampling for diverse answers
+        pad_token_id=tokenizer.eos_token_id  # EOS token to avoid error
+    )
+    
+    # Decode and return the response
+    return tokenizer.decode(outputs[0], skip_special_tokens=True)
+
+# Create a Gradio interface for the chatbot
+def chatbot_interface(user_input):
+    response = generate_response(user_input)
+    return response
+
+# Fine-tuning function
+def fine_tune_model(dataset_name, output_dir='./fine-tuned-model'):
+    # Load dataset (assuming it's in Hugging Face format or a similar format)
+    dataset = load_dataset(dataset_name)
+
+    # Define training arguments
+    training_args = TrainingArguments(
+        output_dir=output_dir,        # Output directory for the fine-tuned model
+        per_device_train_batch_size=2, # Adjust batch size according to GPU capacity
+        num_train_epochs=1,           # Adjust the number of epochs based on the dataset
+        save_steps=500,               # Save checkpoint every 500 steps
+        save_total_limit=2,           # Keep only last two checkpoints
+    )
+
+    # Define the Trainer object
+    trainer = Trainer(
+        model=model,
+        args=training_args,
+        train_dataset=dataset['train'],
+        eval_dataset=dataset['test'],
+    )
+
+    # Fine-tune the model
+    trainer.train()
+
+    return f"Model fine-tuned and saved to {output_dir}"
+
+# Gradio interface for fine-tuning the model
+def fine_tune_interface(dataset_name):
+    result = fine_tune_model(dataset_name)
+    return result
+
+# Gradio UI for both chatbot and fine-tuning functionalities
+with gr.Blocks() as demo:
+    gr.Markdown("# Cable Industry Customer Service Chatbot")
+
+    # Chatbot section
+    with gr.Row():
+        with gr.Column():
+            chatbot_input = gr.Textbox(label="Ask the Cable Industry Chatbot")
+            chatbot_output = gr.Textbox(label="Chatbot Response")
+            chatbot_button = gr.Button("Generate Response")
+        
+        chatbot_button.click(fn=chatbot_interface, inputs=chatbot_input, outputs=chatbot_output)
+
+    # Fine-tuning section
+    with gr.Row():
+        with gr.Column():
+            gr.Markdown("### Fine-tune the Chatbot")
+            dataset_input = gr.Textbox(label="Enter Hugging Face Dataset Name (e.g., 'cable_data')")
+            fine_tune_button = gr.Button("Fine-tune Model")
+            fine_tune_output = gr.Textbox(label="Fine-tune Status")
+        
+        fine_tune_button.click(fn=fine_tune_interface, inputs=dataset_input, outputs=fine_tune_output)
+
+# Launch the Gradio interface
+demo.launch()