feat: add chat UI + JSON API

app.py

@@ -1,64 +1,55 @@
+import torch
+from transformers import AutoTokenizer, AutoModelForCausalLM
 import gradio as gr
-from huggingface_hub import InferenceClient
-
-"""
-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("HuggingFaceH4/zephyr-7b-beta")
-
-
-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})
-
-    response = ""
-
-    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
-
-        response += token
-        yield response
-
+import os
+
+MODEL_ID = "TinyLlama/TinyLlama-1.1B-Chat-v1.0"
+
+# 1) load model & tokenizer
+tokenizer = AutoTokenizer.from_pretrained(MODEL_ID, trust_remote_code=True)
+model = AutoModelForCausalLM.from_pretrained(
+    MODEL_ID,
+    load_in_4bit=True,            # comment out to use full precision
+    torch_dtype=torch.float16,
+    device_map="auto",
+    trust_remote_code=True,
+)
 
-"""
-For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
-"""
+# 2) define inference function
+def generate(messages):
+    """
+    messages: List of alternating [user, assistant, user, ...]
+    returns: [user, assistant, user, assistant, ...] with model's reply appended
+    """
+    # rebuild a single prompt string
+    prompt = ""
+    for i in range(0, len(messages), 2):
+        prompt += f"User: {messages[i]}\n"
+        if i+1 < len(messages):
+            prompt += f"Assistant: {messages[i+1]}\n"
+    prompt += "Assistant:"
+
+    inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
+    outputs = model.generate(
+        **inputs,
+        max_new_tokens=128,
+        do_sample=True,
+        temperature=0.7,
+    )
+    text = tokenizer.decode(outputs[0], skip_special_tokens=True)
+    # strip everything before the last "Assistant:"
+    reply = text.split("Assistant:")[-1].strip()
+
+    messages.append(reply)
+    return messages
+
+# 3) build Gradio ChatInterface
 demo = gr.ChatInterface(
-    respond,
-    additional_inputs=[
-        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)",
-        ),
-    ],
+    fn=generate,
+    title="TinyLlama-1.1B Chat API",
+    description="Chat with TinyLlama-1.1B and call via /api/predict",
 )
 
-
+# 4) launch
 if __name__ == "__main__":
     demo.launch()

requirements.txt

@@ -1 +1,5 @@
-huggingface_hub==0.25.2
+transformers
+torch
+accelerate
+bitsandbytes       # for 4-bit quant
+gradio             # Gradio UI + auto-API