reset to basic

app.py

@@ -1,72 +1,69 @@
 import gradio as gr
-import torch
-from transformers import AutoModelForCausalLM, AutoTokenizer
-import uuid
-import os
-from datetime import datetime
-import spaces
+from huggingface_hub import InferenceClient
 
-MODEL_NAME = "HuggingFaceH4/zephyr-7b-beta"
 with open("system_prompt.txt", "r") as f:
     SYSTEM_PROMPT = f.read()
-LOG_DIR = "chat_logs"
-os.makedirs(LOG_DIR, exist_ok=True)
 
-# Globals
-model = None
-tokenizer = None
-session_id = str(uuid.uuid4())
+MODEL_NAME = "HuggingFaceH4/zephyr-7b-beta"
+
+"""
+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
+"""
+client = InferenceClient(MODEL_NAME)
+
+
+def respond(
+    message,
+    history: list[tuple[str, str]],
+    system_message,
+    max_tokens,
+    temperature,
+    top_p,
+):
+    messages = [{"role": "system", "content": system_message}]
+
+    for val in history:
+        if val[0]:
+            messages.append({"role": "user", "content": val[0]})
+        if val[1]:
+            messages.append({"role": "assistant", "content": val[1]})
+
+    messages.append({"role": "user", "content": message})
 
-def log_chat(session_id, user_msg, bot_msg):
-    timestamp = datetime.now().strftime("%Y-%m-%d %H:%M:%S")
-    with open(os.path.join(LOG_DIR, f"{session_id}.txt"), "a") as f:
-        f.write(f"[{timestamp}] User: {user_msg}\n")
-        f.write(f"[{timestamp}] Bot: {bot_msg}\n\n")
+    response = ""
 
-# This function will be run by ZeroGPU
-@spaces.GPU
-def load_model():
-    tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
-    model = AutoModelForCausalLM.from_pretrained(
-        MODEL_NAME,
-        torch_dtype=torch.float16 if torch.cuda.is_available() else torch.float32,
-        device_map="auto"
-    )
-    return tokenizer, model
+    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
 
-def format_chat_prompt(history, new_input):
-    messages = [{"role": "system", "content": SYSTEM_PROMPT}]
-    for user_msg, bot_msg in history:
-        messages.append({"role": "user", "content": user_msg})
-        messages.append({"role": "assistant", "content": bot_msg})
-    messages.append({"role": "user", "content": new_input})
-    return tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
+        response += token
+        yield response
 
-@torch.no_grad()
-def respond(message, history):
-    global tokenizer, model
 
-    # Lazy-load model only when needed
-    if tokenizer is None or model is None:
-        tokenizer, model = load_model()
+"""
+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=SYSTEM_PROMPT, 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)",
+        ),
+    ],
+)
 
-    prompt = format_chat_prompt(history, message)
-    inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
-    output = model.generate(
-        **inputs,
-        max_new_tokens=512,
-        do_sample=True,
-        temperature=0.7,
-        top_p=0.95,
-        pad_token_id=tokenizer.eos_token_id
-    )
-    decoded = tokenizer.decode(output[0], skip_special_tokens=True)
-    response = decoded.split(message)[-1].strip().split("\n")[0].strip()
-    log_chat(session_id, message, response)
-    return response
 
-gr.ChatInterface(
-    fn=respond,
-    title="BoundrAI",
-    theme="soft"
-).launch()
+if __name__ == "__main__":
+    demo.launch()

requirements.txt

@@ -1,6 +1 @@
-huggingface_hub==0.25.2
-gradio
-transformers
-torch
-spaces
-accelerate>=0.26.0
+huggingface_hub==0.25.2