Update app.py

app.py

@@ -1,80 +1,97 @@
-import os
-import torch
 import gradio as gr
-from transformers import AutoModelForCausalLM, AutoTokenizer
+from huggingface_hub import InferenceClient
+
 
 """
-🧮 Root_Math full model chat app
-Supports private/public repo and GPU/CPU auto-detection.
+For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
 """
 
-# ✅ Load Hugging Face API token
-api_token = os.getenv("HUGGINGFACEHUB_API_TOKEN")
-if not api_token:
-    raise ValueError("❌ ERROR: Hugging Face API token is not set. Please set it as an environment variable.")
-
-# ✅ Correct model repo name (case-sensitive)
-model_name = "Hrushi02/Root_Math"  # double-check on HF website
-
-# ✅ Device and dtype
-device = "cuda" if torch.cuda.is_available() else "cpu"
-dtype = torch.float16 if device == "cuda" else torch.float32
-print(f"⚡ Loading model on {device.upper()} with dtype={dtype}")
-
-# ✅ Load the model (directly, full fine-tuned)
-model = AutoModelForCausalLM.from_pretrained(
-    model_name,
-    torch_dtype=dtype,
-    device_map="auto",
-    token=api_token  # required for private repo
-)
+import os
+client = InferenceClient(model="HuggingFaceH4/zephyr-7b-beta", token=os.getenv("HUGGINGFACEHUB_API_TOKEN"))
 
-# ✅ Load tokenizer
-tokenizer = AutoTokenizer.from_pretrained(model_name, token=api_token)
 
+def respond(
+    message,
+    history: list[tuple[str, str]],
+    system_message,
+    max_tokens,
+    temperature,
+    top_p,
+):
+    messages = [{"role": "system", "content": system_message}]
 
-# ✅ Response function
-def respond(message, history, system_message, max_tokens, temperature, top_p):
-    """Generate a response using Root_Math model."""
-    full_prompt = system_message + "\n\n"
-    for user_msg, bot_msg in history:
-        if user_msg:
-            full_prompt += f"User: {user_msg}\n"
-        if bot_msg:
-            full_prompt += f"Assistant: {bot_msg}\n"
-    full_prompt += f"User: {message}\nAssistant:"
+    for val in history:
+        if val[0]:
+            messages.append({"role": "user", "content": val[0]})
+        if val[1]:
+            messages.append({"role": "assistant", "content": val[1]})
 
-    inputs = tokenizer(full_prompt, return_tensors="pt").to(model.device)
+    messages.append({"role": "user", "content": message})
 
-    with torch.no_grad():
-        outputs = model.generate(
-            **inputs,
-            max_new_tokens=int(max_tokens),
-            temperature=float(temperature),
-            top_p=float(top_p),
-            do_sample=True
-        )
+    response = ""
 
-    response = tokenizer.decode(outputs[0], skip_special_tokens=True)
-    if "Assistant:" in response:
-        response = response.split("Assistant:")[-1].strip()
+    for message in client.chat_completion(
+        messages,
+        max_tokens=max_tokens,
+        stream=True,
+        temperature=temperature,
+        top_p=top_p,
+    ):
+        token = message.choices[0].delta.content
 
-    yield response
+        response += token
+        yield response
 
 
-# ✅ Gradio ChatInterface
+"""
+For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
+"""
 demo = gr.ChatInterface(
     respond,
     additional_inputs=[
-        gr.Textbox(value="You are a helpful math assistant.", label="System message"),
-        gr.Slider(minimum=1, maximum=1024, value=256, step=1, label="Max new tokens"),
+        gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
+        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
         gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
-        gr.Slider(minimum=0.1, maximum=1.0, value=0.95, step=0.05, label="Top-p (nucleus sampling)"),
+        gr.Slider(
+            minimum=0.1,
+            maximum=1.0,
+            value=0.95,
+            step=0.05,
+            label="Top-p (nucleus sampling)",
+        ),
     ],
-    title="🧮 Root Math Assistant",
-    description="Fine-tuned by Hrushi02 for mathematical reasoning."
 )
 
-# ✅ Launch app
+
 if __name__ == "__main__":
     demo.launch()
+
+import os
+from transformers import AutoModelForCausalLM, AutoTokenizer
+from peft import PeftModel
+import torch
+
+# Load Hugging Face API token securely
+api_token = os.getenv("HUGGINGFACEHUB_API_TOKEN")
+
+if not api_token:
+    raise ValueError("❌ ERROR: Hugging Face API token is not set. Please set it as an environment variable.")
+
+# Define model names
+base_model_name = "unsloth/qwen2.5-math-7b-bnb-4bit"
+peft_model_name = "Hrushi02/Root_Math"
+
+# Load base model with authentication
+base_model = AutoModelForCausalLM.from_pretrained(
+    base_model_name,
+    torch_dtype=torch.float16,
+    device_map="auto",
+    use_auth_token=api_token  # ✅ Correct
+)
+
+
+# Load fine-tuned model
+model = PeftModel.from_pretrained(base_model, peft_model_name, token=api_token)
+
+# Load tokenizer
+tokenizer = AutoTokenizer.from_pretrained(base_model_name, token=api_token)