Update app.py

app.py

@@ -1,128 +1,161 @@
 import gradio as gr
-from unsloth import FastLanguageModel
+from transformers import AutoTokenizer, AutoModelForCausalLM
 import torch
 import time
 
 # ----------------------------
-# 🚀 Load Model
+# 🚀 Load Model (CPU-friendly)
 # ----------------------------
+@gr.cache
 def load_model():
-    max_seq_length = 1024
-    model, tokenizer = FastLanguageModel.from_pretrained(
-        model_name="umarfarzan/my-finetuned-model2-lora",
-        max_seq_length=max_seq_length,
-        dtype=None,
-        load_in_4bit=True,
+    model_name = "umarfarzan/my-finetuned-model2-lora"
+    tokenizer = AutoTokenizer.from_pretrained(model_name)
+    model = AutoModelForCausalLM.from_pretrained(
+        model_name,
+        torch_dtype=torch.float32  # CPU-friendly
     )
-    FastLanguageModel.for_inference(model)
+    model.to("cpu")
     return model, tokenizer
 
-print("🔄 Loading model ...")
+print("Loading model...")
 model, tokenizer = load_model()
 print("✅ Model loaded successfully!")
 
-
 # ----------------------------
 # 💡 Generate Training Program
 # ----------------------------
-def generate_program(instruction, max_tokens=4000, temperature=0.7, top_p=0.9):
-    """Generate a detailed training program from a natural-language prompt"""
-    prompt_text = f"""Below is an instruction that describes a task, paired with an input that provides further context.
-Write a response that appropriately completes the request.
+def generate_training_program(
+    instruction,
+    max_tokens=500,
+    temperature=0.7,
+    top_p=0.9
+):
+    """Generate a training program based on user instruction"""
+    
+    prompt_text = f"""Below is an instruction that describes a task. Write a response that appropriately completes the request.
 
 ### Instruction:
 {instruction}
 
-### Input:
-
-
 ### Response:
 """
+    inputs = tokenizer(prompt_text, return_tensors="pt")
 
-    inputs = tokenizer([prompt_text], return_tensors="pt").to("cuda")
-
-    start = time.time()
+    start_time = time.time()
     outputs = model.generate(
         **inputs,
         max_new_tokens=max_tokens,
         temperature=temperature,
         top_p=top_p,
         do_sample=True,
-        use_cache=True,
+        use_cache=True
     )
-    end = time.time()
-
-    result = tokenizer.batch_decode(outputs, skip_special_tokens=True)[0]
-    if "### Response:" in result:
-        result = result.split("### Response:")[-1].strip()
+    generation_time = time.time() - start_time
 
-    return result, f"⏱️ Generated in {end - start:.2f} sec"
+    generated_text = tokenizer.decode(outputs[0], skip_special_tokens=True)
+    
+    # Extract response after "### Response:"
+    if "### Response:" in generated_text:
+        response = generated_text.split("### Response:")[-1].strip()
+    else:
+        response = generated_text
 
+    return response, f"⏱️ Generated in {generation_time:.2f} seconds"
 
 # ----------------------------
 # 🎨 Gradio Interface
 # ----------------------------
-examples = [
-    ["Design a detailed 1-week training program titled 'The Leader's Blueprint for Strategic Problem-Solving' for mid-level to senior-level managers, team leads, and high-potential employees."],
-    ["Create a 3-day innovation workshop based on TRIZ principles for engineers and R&D specialists."],
-    ["Develop a 5-day leadership bootcamp focused on emotional intelligence and decision-making."],
-    ["Design a 2-day workshop on cross-cultural communication for international teams."],
-    ["Create a 4-week mentorship program for junior data scientists to learn project management and collaboration skills."],
-]
-
 custom_css = """
+.gradio-container { font-family: 'Inter', sans-serif; }
 .main-header {
     text-align: center;
-    background: linear-gradient(135deg, #6a11cb 0%, #2575fc 100%);
+    background: linear-gradient(135deg, #667eea 0%, #764ba2 100%);
     color: white;
     padding: 2rem;
-    border-radius: 12px;
+    border-radius: 10px;
     margin-bottom: 2rem;
 }
-.gradio-container {
-    font-family: 'Inter', sans-serif;
-}
 """
 
+examples = [
+    ["Design a detailed 1-week training program titled 'The Leader's Blueprint for Strategic Problem-Solving' for mid-level to senior-level managers, team leads, and high-potential employees."],
+    ["Create a 3-day workshop on effective communication skills for remote teams."],
+    ["Develop a 5-day leadership bootcamp for new managers focusing on team management and conflict resolution."],
+    ["Design a half-day training session on data-driven decision making for executives."],
+    ["Create a 2-week onboarding program for new software engineers including technical and cultural training."]
+]
+
 with gr.Blocks(css=custom_css, theme=gr.themes.Soft()) as demo:
+
+    # Header
     gr.HTML("""
         <div class="main-header">
             <h1>🎯 AI Training Program Generator</h1>
-            <p>Generate professional, structured training programs instantly with your fine-tuned Qwen2 model</p>
+            <p style="font-size: 1.1rem; margin-top: 0.5rem;">
+                Generate comprehensive, professional training programs instantly using AI
+            </p>
         </div>
     """)
 
     with gr.Row():
         with gr.Column(scale=1):
-            instruction = gr.Textbox(
-                label="🧠 Describe the training program you want",
-                placeholder="Example: Design a 3-day workshop on innovation for engineers...",
-                lines=5,
+            instruction_input = gr.Textbox(
+                label="📝 Training Program Description",
+                placeholder="Example: Design a 1-week training program on strategic problem-solving for managers...",
+                lines=5
             )
 
             with gr.Accordion("⚙️ Advanced Settings", open=False):
-                max_tokens = gr.Slider(500, 8000, value=4000, step=100, label="Max Output Length")
-                temperature = gr.Slider(0.1, 1.5, value=0.7, step=0.1, label="Creativity (Temperature)")
-                top_p = gr.Slider(0.5, 1.0, value=0.9, step=0.05, label="Diversity (Top-p)")
-
-            submit = gr.Button("🚀 Generate", variant="primary")
+                max_tokens_slider = gr.Slider(
+                    minimum=100,
+                    maximum=5500,
+                    value=500,
+                    step=50,
+                    label="Max Output Length",
+                    info="Longer programs take more time on CPU"
+                )
+                temperature_slider = gr.Slider(
+                    minimum=0.1,
+                    maximum=1.5,
+                    value=0.7,
+                    step=0.1,
+                    label="Creativity (Temperature)"
+                )
+                top_p_slider = gr.Slider(
+                    minimum=0.5,
+                    maximum=1.0,
+                    value=0.9,
+                    step=0.05,
+                    label="Diversity (Top-p)"
+                )
+
+            generate_btn = gr.Button("🚀 Generate Training Program", variant="primary", size="lg")
 
         with gr.Column(scale=1):
-            output = gr.Textbox(
+            output_text = gr.Textbox(
                 label="📋 Generated Training Program",
                 lines=25,
-                show_copy_button=True,
+                show_copy_button=True
+            )
+            generation_info = gr.Textbox(
+                label="ℹ️ Generation Info",
+                interactive=False,
+                show_label=False
             )
-            info = gr.Textbox(label="ℹ️ Info", interactive=False, show_label=False)
 
-    gr.Examples(examples, inputs=[instruction], label="💡 Example Prompts")
+    gr.Examples(
+        examples=examples,
+        inputs=[instruction_input],
+        label="💡 Quick Start Examples"
+    )
 
-    submit.click(
-        generate_program,
-        inputs=[instruction, max_tokens, temperature, top_p],
-        outputs=[output, info],
+    generate_btn.click(
+        fn=generate_training_program,
+        inputs=[instruction_input, max_tokens_slider, temperature_slider, top_p_slider],
+        outputs=[output_text, generation_info]
     )
 
-demo.queue(max_size=10)
+# Launch
 if __name__ == "__main__":
+    demo.queue(max_size=5)
     demo.launch(server_name="0.0.0.0", server_port=7860)