Rename nikapp.py to app.py

app.py

@@ -0,0 +1,203 @@
+import gradio as gr
+import subprocess
+from huggingface_hub import hf_hub_download
+
+# 1. Install llama-cpp-python in runtime
+subprocess.run("pip install -q 'llama_cpp_python==0.3.15'", shell=True, check=False)
+
+from llama_cpp import Llama
+
+# 2. Load GGUF model
+MODEL_REPO = "Jeppcode/ScalableLab2"
+GGUF_FILENAME = "model-q4_k_m.gguf" 
+
+print(f"Downloading GGUF model {MODEL_REPO}/{GGUF_FILENAME} ...")
+model_path = hf_hub_download(
+    repo_id=MODEL_REPO,
+    filename=GGUF_FILENAME,
+)
+
+print("Initializing llama.cpp LLM ...")
+llm = Llama(
+    model_path=model_path,
+    n_ctx=2048,      
+    n_threads=2,     
+    n_batch=64,      
+    use_mmap=True,
+    use_mlock=False,
+)
+
+# 3. System Prompts
+STYLE_SYSTEM_PROMPTS = {
+    "Default": "You are a helpful, polite assistant.",
+    "Short answer": (
+        "You are a helpful assistant. Answer as concisely as possible, usually in 1–3 sentences."
+    ),
+    "Detailed explanation": (
+        "You are a helpful teaching assistant. Give clear, structured and detailed explanations, "
+        "often with bullet points or numbered steps when useful."
+    ),
+    "Step-by-step reasoning": (
+        "You are a careful problem solver. Think step by step and explain your reasoning clearly "
+        "before giving the final answer."
+    ),
+}
+
+def _extract_text_from_content(content):
+    if isinstance(content, list):
+        texts = []
+        for block in content:
+            if isinstance(block, dict) and block.get("type") == "text":
+                texts.append(block.get("text", ""))
+            else:
+                texts.append(str(block))
+        return "\n".join(t for t in texts if t)
+    else:
+        return str(content)
+
+def build_prompt(message, history, style):
+    system_prompt = STYLE_SYSTEM_PROMPTS.get(style, STYLE_SYSTEM_PROMPTS["Default"])
+    prompt_parts = []
+    prompt_parts.append(f"System: {system_prompt}\n")
+    prompt_parts.append("Conversation:\n")
+    
+    for msg in history or []:
+        role = msg.get("role")
+        content = _extract_text_from_content(msg.get("content", ""))
+        if not content:
+            continue
+        if role == "user":
+            prompt_parts.append(f"User: {content}\n")
+        elif role == "assistant":
+            prompt_parts.append(f"Assistant: {content}\n")
+        elif role == "system":
+            prompt_parts.append(f"System (previous): {content}\n")
+            
+    prompt_parts.append(f"User: {message}\n")
+    prompt_parts.append("Assistant:")
+    return "".join(prompt_parts)
+
+def chat_fn(message, history, max_new_tokens, style):
+    # Removed sliders are now hardcoded defaults here
+    temperature = 0.7
+    top_p = 0.9
+    repetition_penalty = 1.1
+
+    prompt = build_prompt(message, history, style)
+    
+    # Simple logic for temperature
+    if temperature <= 0.0:
+        temperature = 0.0
+        top_p = 1.0
+
+    output = llm(
+        prompt,
+        max_tokens=int(max_new_tokens),
+        temperature=temperature,
+        top_p=top_p,
+        repeat_penalty=repetition_penalty,
+        stop=["User:", "Assistant:", "System:", "Conversation:"],
+    )
+    reply = output["choices"][0]["text"].strip()
+    return reply
+
+# --- Christmas Theme Configuration ---
+
+# 1. THEME: Red (Santa) and Green (Tree)
+christmas_theme = gr.themes.Soft(
+    primary_hue="red",
+    secondary_hue="green",
+    neutral_hue="slate",
+).set(
+    body_background_fill="transparent",
+    block_background_fill="rgba(255, 250, 240, 0.9)", # Creamy white snow color
+    border_color_primary="#D4AF37", # Gold Border
+    button_primary_background_fill="#C62828", # Santa Red
+    button_primary_text_color="white",
+)
+
+# 2. CSS: Background image + Festive Styling
+custom_css = """
+/* Background: A cozy Christmas scene */
+.gradio-container {
+    background: url('https://images.unsplash.com/photo-1544976735-a10c71a39644?q=80&w=2560&auto=format&fit=crop') no-repeat center center fixed; 
+    background-size: cover;
+}
+
+/* Make main container transparent */
+.gradio-container > .main {
+    background: transparent !important;
+}
+
+/* Chatbot Window - Glassy Snow Look */
+.bubble-wrap {
+    background: rgba(255, 255, 255, 0.85) !important;
+    border: 2px solid #D4AF37 !important; /* Gold Border */
+    border-radius: 15px !important;
+}
+
+/* User Message - Christmas Red */
+.user-message {
+    background-color: #D32F2F !important; 
+    color: white !important;
+    border: 1px solid #B71C1C !important;
+}
+
+/* Bot Message - Christmas Green */
+.bot-message {
+    background-color: #2E7D32 !important; 
+    color: white !important;
+    border: 1px solid #1B5E20 !important;
+}
+
+/* Accordion/Settings - Snowy background with Gold border */
+.group, .form {
+    background: rgba(255, 255, 255, 0.9) !important;
+    border: 2px solid #D4AF37 !important;
+    border-radius: 10px;
+    padding: 10px;
+}
+
+/* Labels */
+label, span {
+    color: #3E2723 !important; /* Dark chocolate text for readability */
+    font-weight: bold;
+}
+
+footer {visibility: hidden}
+"""
+
+# 4. Inputs (Only Max Tokens + Style)
+max_new_tokens_slider = gr.Slider(
+    minimum=16, maximum=256, value=64, step=8, label="Max new tokens (Length)",
+)
+
+style_radio = gr.Radio(
+    choices=[
+        "Default",
+        "Short answer",
+        "Detailed explanation",
+        "Step-by-step reasoning",
+    ],
+    value="Detailed explanation",
+    label="Answer style",
+)
+
+# Instantiate ChatInterface
+demo = gr.ChatInterface(
+    fn=chat_fn,
+    title="🎄 Holiday Chat Lab 2 🎅",
+    description="Chat with the fine-tuned model. Grab a hot chocolate and enjoy the holidays.",
+    additional_inputs=[
+        max_new_tokens_slider,
+        style_radio,
+    ],
+    additional_inputs_accordion="Holiday Settings",
+)
+
+# Apply Theme and CSS manually (Safe for older Gradio versions)
+demo.theme = christmas_theme
+demo.css = custom_css
+
+if __name__ == "__main__":
+    demo.launch()

nikapp.py

@@ -1,166 +0,0 @@
-import gradio as gr
-import torch
-from transformers import AutoModelForCausalLM, AutoTokenizer
-
-MODEL_ID = "Jeppcode/ScalableLab2"
-SUBFOLDER = "merged-model-fp16"
-
-print(f"Loading model {MODEL_ID}/{SUBFOLDER} ...")
-
-# Tokenizer
-tokenizer = AutoTokenizer.from_pretrained(
-    MODEL_ID, subfolder=SUBFOLDER,
-)
-
-# Model – fp16 and optimized for CPU
-model = AutoModelForCausalLM.from_pretrained(
-    MODEL_ID, 
-    subfolder=SUBFOLDER, 
-    dtype=torch.float16, 
-    low_cpu_mem_usage=True, 
-    device_map="cpu",
-)
-model.eval()
-
-# Hardcoded system prompt
-SYSTEM_PROMPT = "You are a helpful, polite assistant. Give clear and structured explanations."
-
-def build_prompt(message, history):
-    messages = []
-    messages.append({"role": "system", "content": SYSTEM_PROMPT})
-    
-    for msg in history:
-        role = msg.get("role")
-        content = msg.get("content", "")
-        if isinstance(content, list):
-            texts = []
-            for block in content:
-                if isinstance(block, dict) and block.get("type") == "text":
-                    texts.append(block.get("text", ""))
-                else:
-                    texts.append(str(block))
-            text = "\n".join(t for t in texts if t)
-        else:
-            text = str(content)
-            
-        if text and role in ("user", "assistant", "system"):
-            messages.append({"role": role, "content": text})
-            
-    messages.append({"role": "user", "content": message})
-    
-    prompt = tokenizer.apply_chat_template(
-        messages, tokenize=False, add_generation_prompt=True,
-    )
-    return prompt
-
-def chat_fn(message, history, max_new_tokens):
-    prompt = build_prompt(message, history)
-    inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
-    
-    # Default values 
-    temperature = 0.7 
-    top_p = 0.9
-    repetition_penalty = 1.1 
-
-    gen_kwargs = {
-        **inputs,
-        "max_new_tokens": int(max_new_tokens),
-        "pad_token_id": tokenizer.eos_token_id,
-        "eos_token_id": tokenizer.eos_token_id,
-        "repetition_penalty": float(repetition_penalty),
-        "do_sample": True,
-        "temperature": float(temperature),
-        "top_p": float(top_p),
-    }
-
-    with torch.no_grad():
-        outputs = model.generate(**gen_kwargs)
-        generated = tokenizer.decode(
-            outputs[0][inputs["input_ids"].shape[1]:], skip_special_tokens=True,
-        ).strip()
-
-    return generated
-
-# --- Visual Customization ---
-
-# 1. THEME: Set the base colors to forest greens
-nature_theme = gr.themes.Soft(
-    primary_hue="green",
-    secondary_hue="emerald",
-    neutral_hue="stone",
-).set(
-    # Make the default backgrounds transparent or dark to blend with image
-    body_background_fill="transparent",
-    block_background_fill="rgba(20, 30, 20, 0.8)", # Dark semi-transparent green
-    border_color_primary="#4CAF50", # Bright Green border
-    input_background_fill="rgba(0, 0, 0, 0.5)",
-    button_primary_background_fill="#2E7D32", 
-    text_color_subdued="#A5D6A7", # Light green text for labels
-)
-
-# 2. CSS: Force the background image and specific "Green Border" look
-custom_css = """
-/* The Main Forest Background */
-.gradio-container {
-    background: url('https://images.unsplash.com/photo-1441974231531-c6227db76b6e?q=80&w=2560&auto=format&fit=crop') no-repeat center center fixed; 
-    background-size: cover;
-}
-
-/* Make the main App container transparent so background shows */
-.gradio-container > .main {
-    background: transparent !important;
-}
-
-/* Chatbot Window Styling - The "Glassy" Look */
-.bubble-wrap {
-    background: rgba(0, 0, 0, 0.6) !important;
-    border: 1px solid #4CAF50 !important;
-    border-radius: 10px !important;
-}
-
-/* User and Bot message bubbles */
-.user-message {
-    background-color: #2E7D32 !important; /* Forest Green for user */
-    border: 1px solid #66BB6A !important;
-}
-.bot-message {
-    background-color: rgba(40, 40, 40, 0.9) !important; /* Dark Grey for bot */
-    border: 1px solid #4CAF50 !important;
-}
-
-/* Input Area and Settings - Green Borders */
-.group, .form {
-    background: rgba(10, 20, 10, 0.85) !important; /* Dark semi-transparent */
-    border: 2px solid #4CAF50 !important; /* The prominent green border */
-    border-radius: 8px;
-    padding: 10px;
-}
-
-/* Text colors */
-label, span, p {
-    color: #E8F5E9 !important; /* Very light green/white text */
-}
-
-/* Hide Footer */
-footer {visibility: hidden}
-"""
-
-max_new_tokens_slider = gr.Slider(
-    minimum=16, maximum=512, value=128, step=8, label="Response Length (Tokens)",
-)
-
-# Instantiate ChatInterface
-demo = gr.ChatInterface(
-    fn=chat_fn,
-    title="🌿 NatureChat Lab 2",
-    description="Chat with the fine-tuned Llama model. Relax and enjoy the view.",
-    additional_inputs=[max_new_tokens_slider],
-    additional_inputs_accordion="Settings"
-)
-
-# Apply the Theme and CSS manually (Compatible with older Gradio versions)
-demo.theme = nature_theme
-demo.css = custom_css
-
-if __name__ == "__main__":
-    demo.launch()