Upload app.py with huggingface_hub

app.py

@@ -1,66 +1,82 @@
 
 import gradio as gr
 import time
-import requests
-import json
+from transformers import AutoTokenizer, AutoModelForCausalLM
+import torch
 
-# Demo responses for HR testing
-DEMO_RESPONSES = {
-    "Hi there!": "Hi there! 👋 I'm the RML-AI demo, and I'd love to answer any questions you have about AI, technology, science, or anything else!",
-    "What can you do?": "Great question! I'm a demo of RML-AI, which is pretty amazing. I can chat about AI, technology, science, business, health - you name it! The full system has access to 100GB of knowledge and can answer complex questions with sources. What interests you?",
-    "Tell me about AI": "Oh, AI! That's a fascinating topic. AI and machine learning are changing the world, and RML-AI is actually a revolutionary new approach to AI. The full system would give you detailed, accurate answers with sources from its massive knowledge base. What specifically about AI interests you?",
-    "How are you?": "I'm doing great, thanks for asking! 😊 I'm excited to be chatting with you and showing off what RML-AI can do. What's on your mind today?",
-    "What is RML?": "RML is what makes this system special! It's Resonant Memory Learning - a completely new way of doing AI that's 100x more efficient and 10x faster than traditional systems. It's exactly the kind of technology that's changing the future of AI!"
-}
+MODEL_ID = "akshaynayaks9845/rml-ai-phi1_5-rml-100k"
 
-SAMPLES = list(DEMO_RESPONSES.keys())
+# Global model and tokenizer
+_model = None
+_tokenizer = None
+
+def load_model():
+    global _model, _tokenizer
+    if _model is None:
+        try:
+            print("Loading RML model...")
+            _tokenizer = AutoTokenizer.from_pretrained(MODEL_ID, trust_remote_code=True)
+            if _tokenizer.pad_token is None:
+                _tokenizer.pad_token = _tokenizer.eos_token
+            
+            _model = AutoModelForCausalLM.from_pretrained(
+                MODEL_ID, 
+                trust_remote_code=True,
+                torch_dtype=torch.float16 if torch.cuda.is_available() else torch.float32,
+                device_map="auto" if torch.cuda.is_available() else None,
+                low_cpu_mem_usage=True
+            )
+            print("Model loaded successfully!")
+        except Exception as e:
+            print(f"Error loading model: {e}")
+            return False
+    return True
 
 def generate_response(prompt, max_new_tokens=128, temperature=0.2):
     start = time.time()
     
-    # Check if we have a demo response
-    if prompt in DEMO_RESPONSES:
-        reply = DEMO_RESPONSES[prompt]
-    else:
-        # Generate conversational response based on question content
-        prompt_lower = prompt.lower()
-        
-        # Greetings and casual conversation
-        if any(word in prompt_lower for word in ['hi', 'hello', 'hey', 'good morning', 'good afternoon', 'good evening']):
-            greetings = ["Hi there! 👋", "Hello! Nice to meet you!", "Hey! How can I help you today?", "Hi! I'm excited to chat with you!", "Hello! What would you like to know?"]
-            import random
-            reply = random.choice(greetings) + " I'm the RML-AI demo, and I'd love to answer any questions you have about AI, technology, science, or anything else!"
-        
-        # How are you / personal questions
-        elif any(word in prompt_lower for word in ['how are you', 'how do you do', 'are you ok', 'feeling']):
-            reply = "I'm doing great, thanks for asking! 😊 I'm excited to be chatting with you and showing off what RML-AI can do. What's on your mind today?"
-        
-        # What can you do / capabilities
-        elif any(word in prompt_lower for word in ['what can you do', 'help', 'capabilities', 'features']):
-            reply = "Great question! I'm a demo of RML-AI, which is pretty amazing. I can chat about AI, technology, science, business, health - you name it! The full system has access to 100GB of knowledge and can answer complex questions with sources. What interests you?"
+    if not load_model():
+        return "Error: Could not load the RML model. Please try again."
+    
+    try:
+        # Prepare input
+        inputs = _tokenizer(prompt, return_tensors="pt", truncation=True, max_length=512)
         
-        # AI/ML topics
-        elif any(word in prompt_lower for word in ['ai', 'artificial intelligence', 'machine learning', 'ml', 'neural', 'deep learning']):
-            reply = "Oh, " + prompt + "! That's a fascinating topic. AI and machine learning are changing the world, and RML-AI is actually a revolutionary new approach to AI. The full system would give you detailed, accurate answers with sources from its massive knowledge base. What specifically about AI interests you?"
+        # Generate response
+        with torch.no_grad():
+            outputs = _model.generate(
+                **inputs,
+                max_new_tokens=int(max_new_tokens),
+                do_sample=bool(temperature > 0),
+                temperature=float(temperature),
+                top_p=0.9,
+                repetition_penalty=1.1,
+                pad_token_id=_tokenizer.eos_token_id
+            )
         
-        # Technology topics
-        elif any(word in prompt_lower for word in ['technology', 'tech', 'computer', 'software', 'programming', 'code']):
-            reply = "Technology is incredible, isn't it? " + prompt + " is such an interesting area. The RML-AI system has tons of knowledge about the latest tech developments and can explain complex concepts in simple terms. What aspect of technology would you like to explore?"
+        # Decode response
+        generated_text = _tokenizer.decode(outputs[0], skip_special_tokens=True)
         
-        # Science topics
-        elif any(word in prompt_lower for word in ['science', 'research', 'study', 'physics', 'chemistry', 'biology']):
-            reply = "Science is amazing! " + prompt + " is a great topic to dive into. The RML-AI system has access to scientific literature and research papers, so it can give you evidence-based answers with proper citations. What scientific topic catches your interest?"
+        # Extract only the new part (after the input prompt)
+        if generated_text.startswith(prompt):
+            response = generated_text[len(prompt):].strip()
+        else:
+            response = generated_text.strip()
         
-        # RML-specific questions
-        elif any(word in prompt_lower for word in ['rml', 'resonant', 'memory', 'learning']):
-            reply = "RML is what makes this system special! It's Resonant Memory Learning - a completely new way of doing AI that's 100x more efficient and 10x faster than traditional systems. " + prompt + " is exactly the kind of question RML-AI excels at answering with its revolutionary frequency-based approach."
+        elapsed = int((time.time() - start) * 1000)
+        return response + f"\n\n(⏱️ {elapsed} ms)"
         
-        # General questions
-        else:
-            reply = "That's an interesting question about '" + prompt + "'! I'd love to help you with that. The full RML-AI system would search through its 100GB knowledge base to give you a comprehensive, well-sourced answer. What would you like to know more about?"
-    
-    elapsed = int((time.time() - start) * 1000)
-    return reply + "\n\n(⏱️ " + str(elapsed) + " ms)\n\n💡 This is a demo. The full RML-AI provides detailed, source-attributed responses!"
+    except Exception as e:
+        return f"Error generating response: {str(e)}"
+
+# Sample questions for the demo
+SAMPLES = [
+    "What is artificial intelligence?",
+    "Explain machine learning in simple terms",
+    "What is quantum computing?",
+    "How does RML work?",
+    "Tell me about neural networks"
+]
 
 with gr.Blocks(title="RML-AI Demo") as demo:
     gr.Markdown('''