Update app.py

app.py

@@ -1,26 +1,74 @@
 # app.py
 import gradio as gr
-from transformers import pipeline
+from transformers import pipeline, AutoTokenizer, AutoConfig, AutoModelForCausalLM
 import torch
+import os
 import re
+import json
+import time
+from datetime import datetime
 
 # ====== Load Model ======
 device = 0 if torch.cuda.is_available() else -1
+model_name = "rahul7star/Qwen0.5-3B-Gita"
+
+log_lines = []
+
+def log(msg):
+    """Append timestamped message to log."""
+    line = f"[{datetime.now().strftime('%H:%M:%S')}] {msg}"
+    print(line)
+    log_lines.append(line)
+
+log("🔍 Initializing model load sequence...")
+log(f"Using model: {model_name}")
+log(f"Detected device: {'GPU' if device == 0 else 'CPU'}")
+
+# Inspect model folder (once downloaded from HF cache)
+hf_cache = os.path.expanduser("~/.cache/huggingface/hub")
+log(f"Model will be loaded from local cache directory: {hf_cache}")
+
+try:
+    config = AutoConfig.from_pretrained(model_name)
+    log("✅ Loaded configuration file:")
+    log(json.dumps(config.to_dict(), indent=2)[:800] + " ...")
+except Exception as e:
+    log(f"⚠️ Could not read model config: {e}")
+
+try:
+    tokenizer = AutoTokenizer.from_pretrained(model_name)
+    log("✅ Tokenizer loaded successfully.")
+    log(f"Tokenizer vocab size: {tokenizer.vocab_size}")
+    log(f"Tokenizer files found in: {tokenizer.pretrained_vocab_files_map}")
+except Exception as e:
+    log(f"⚠️ Could not load tokenizer: {e}")
+
+# Load model pipeline
+start_load = time.time()
 pipe = pipeline(
     "text-generation",
-    model="rahul7star/Qwen0.5-3B-Gita",
+    model=model_name,
     device=device,
 )
+log(f"✅ Model pipeline loaded in {time.time() - start_load:.2f} seconds.")
 
 # ====== Chat Function ======
 def chat_with_model(message, history):
-    # Build conversation context as plain text
+    log_lines.clear()
+    log("💭 Starting chat generation process...")
+    log(f"User message: {message}")
+
+    # 1️⃣ Build conversation context
     context = "The following is a conversation between a user and an AI assistant inspired by the Bhagavad Gita.\n"
     for user, bot in history:
         context += f"User: {user}\nAssistant: {bot}\n"
     context += f"User: {message}\nAssistant:"
+    log("📄 Built conversation context:")
+    log(context)
 
-    # Generate response
+    # 2️⃣ Encode and run model
+    log("🧠 Encoding input and generating response...")
+    start_time = time.time()
     output = pipe(
         context,
         max_new_tokens=200,
@@ -30,31 +78,48 @@ def chat_with_model(message, history):
         repetition_penalty=1.1,
         truncation=True,
     )[0]["generated_text"]
+    log(f"⏱️ Inference took {time.time() - start_time:.2f} seconds")
 
-    # Extract assistant reply only (remove repeated context)
+    # 3️⃣ Extract clean assistant reply
     reply = output[len(context):].strip()
-
-    # Clean junk or repeated tokens
     reply = re.sub(r"(ContentLoaded|<\/?[^>]+>|[\r\n]{2,})", " ", reply)
     reply = re.sub(r"\s{2,}", " ", reply).strip()
-
-    # Cut off weird repetitions
     reply = reply.split("User:")[0].split("Assistant:")[0].strip()
 
+    log("🪄 Raw model output processed successfully.")
+    log(f"Model reply (cleaned): {reply}")
+
+    # 4️⃣ Log tokenizer + model folders
+    try:
+        model_dir = pipe.model.name_or_path
+        log(f"📂 Model files are read from: {model_dir}")
+        if os.path.exists(model_dir):
+            for root, dirs, files in os.walk(model_dir):
+                for file in files[:5]:  # show first 5 files only
+                    log(f"  - {os.path.join(root, file)}")
+                break
+    except Exception as e:
+        log(f"⚠️ Could not list model folder files: {e}")
+
+    # 5️⃣ Finalize
     history.append((message, reply))
-    return "", history
+    return "", history, "\n".join(log_lines)
 
 
 # ====== Gradio Interface ======
 with gr.Blocks(theme=gr.themes.Soft(primary_hue="blue")) as demo:
-    gr.Markdown("## 💬 Qwen0.5-3B-Gita — Conversational Assistant")
+    gr.Markdown("## 💬 Qwen0.5-3B-Gita — Conversational Assistant with Debug Log")
 
-    chatbot = gr.Chatbot(height=500)
-    msg = gr.Textbox(placeholder="Ask about the Gita, life, or philosophy...", label="Your Message")
-    clear = gr.Button("Clear")
+    with gr.Row():
+        with gr.Column(scale=2):
+            chatbot = gr.Chatbot(height=500)
+            msg = gr.Textbox(placeholder="Ask about the Gita, life, or philosophy...", label="Your Message")
+            clear = gr.Button("Clear")
+        with gr.Column(scale=1):
+            log_box = gr.Textbox(label="Detailed Model Log", lines=25, interactive=False)
 
-    msg.submit(chat_with_model, [msg, chatbot], [msg, chatbot])
-    clear.click(lambda: None, None, chatbot, queue=False)
+    msg.submit(chat_with_model, [msg, chatbot], [msg, chatbot, log_box])
+    clear.click(lambda: (None, None, ""), None, [chatbot, log_box], queue=False)
 
 # ====== Launch ======
 if __name__ == "__main__":