new model

app.py

@@ -1,42 +1,42 @@
-import torch
-
-import torch
-torch.cuda.is_available = lambda: False  # Force torch to disable CUDA
-
-from unsloth import FastLanguageModel
-
-
-# Force CPU mode
-device = "cpu"
-
-# Load the base model in CPU mode
-base_model_name = "unsloth/Llama-3.2-3B-Instruct"
-base_model, tokenizer = FastLanguageModel.from_pretrained(
-    model_name=base_model_name,
-    max_seq_length=2048,
-    dtype="float32",  # Use float32 for CPU
-    load_in_4bit=False  # Disable 4-bit quantization for CPU
-)
-base_model.to(device)
-
-# Apply LoRA adapters in CPU mode
-from peft import PeftModel
-
-lora_model_name = "oskaralf/lora_model"  # Replace with your LoRA model path
-model = PeftModel.from_pretrained(base_model, lora_model_name)
-model.to(device)
-
-# Prepare for inference in CPU mode
-FastLanguageModel.for_inference(model)
-
-# Gradio interface
 import gradio as gr
-
-def chatbot(input_text):
-    inputs = tokenizer(input_text, return_tensors="pt").to(device)
-    outputs = model.generate(input_ids=inputs["input_ids"], max_new_tokens=64)
+from transformers import AutoModelForCausalLM, AutoTokenizer
+
+MODEL_NAME = "oskaralf/model_merged" 
+model = AutoModelForCausalLM.from_pretrained(MODEL_NAME, torch_dtype="auto", device_map="auto")
+tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
+
+def generate_response(prompt, max_length=128, temperature=0.7, top_p=0.9):
+    inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
+    outputs = model.generate(
+        **inputs,
+        max_length=max_length,
+        temperature=temperature,
+        top_p=top_p,
+        pad_token_id=tokenizer.eos_token_id
+    )
     response = tokenizer.decode(outputs[0], skip_special_tokens=True)
     return response
 
-iface = gr.Interface(fn=chatbot, inputs="text", outputs="text", title="Chatbot")
-iface.launch()
+def interactive_app():
+    with gr.Blocks() as app:
+        gr.Markdown("# Coding Task Generator")
+        gr.Markdown("Generate coding tasks by entering a prompt below.")
+        
+        prompt = gr.Textbox(label="Enter your prompt:", placeholder="e.g., Create a Python task involving recursion.")
+        max_length = gr.Slider(label="Max Length", minimum=16, maximum=512, value=128, step=16)
+        temperature = gr.Slider(label="Temperature", minimum=0.1, maximum=1.0, value=0.7, step=0.1)
+        top_p = gr.Slider(label="Top-p Sampling", minimum=0.1, maximum=1.0, value=0.9, step=0.1)
+        generate_button = gr.Button("Generate Task")
+        
+        output = gr.Textbox(label="Generated Task", lines=10)
+        
+        generate_button.click(
+            generate_response, 
+            inputs=[prompt, max_length, temperature, top_p],
+            outputs=output
+        )
+    
+    return app
+
+if __name__ == "__main__":
+    interactive_app().launch()