qwen model host

app.py

@@ -1,69 +1,91 @@
-import gradio as gr
-from huggingface_hub import InferenceClient
+#!/usr/bin/env python3
+"""
+Load and run Qwen3-0.6B (or 1.7B/4B) locally from Hugging Face.
+Install deps first:
+    pip install transformers torch accelerate
+    # For faster inference (optional):
+    pip install bitsandbytes   # 4-bit/8-bit quantization
+"""
 
+from transformers import AutoModelForCausalLM, AutoTokenizer
+import torch
 
-def respond(
-    message,
-    history: list[dict[str, str]],
-    system_message,
-    max_tokens,
-    temperature,
-    top_p,
-    hf_token: gr.OAuthToken,
-):
-    """
-    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(token=hf_token.token, model="openai/gpt-oss-20b")
+# ── Config ────────────────────────────────────────────────────────────────────
+MODEL_ID = "Qwen/Qwen3-4B"   # swap to "Qwen/Qwen3-1.7B" or "Qwen/Qwen3-0.6B"
+DEVICE   = "cuda" if torch.cuda.is_available() else "cpu"
 
-    messages = [{"role": "system", "content": system_message}]
+# Set to True to load in 4-bit (needs bitsandbytes + GPU)
+USE_4BIT = False
+# ──────────────────────────────────────────────────────────────────────────────
 
-    messages.extend(history)
 
-    messages.append({"role": "user", "content": message})
+def load_model():
+    print(f"Loading tokenizer for {MODEL_ID} ...")
+    tokenizer = AutoTokenizer.from_pretrained(MODEL_ID)
 
-    response = ""
+    print(f"Loading model on {DEVICE} (4-bit={USE_4BIT}) ...")
 
-    for message in client.chat_completion(
-        messages,
-        max_tokens=max_tokens,
-        stream=True,
-        temperature=temperature,
-        top_p=top_p,
-    ):
-        choices = message.choices
-        token = ""
-        if len(choices) and choices[0].delta.content:
-            token = choices[0].delta.content
+    kwargs = dict(
+        device_map="auto",          # spreads across available GPU(s)/CPU
+        torch_dtype=torch.float16 if DEVICE == "cuda" else torch.float32,
+    )
 
-        response += token
-        yield response
+    if USE_4BIT:
+        from transformers import BitsAndBytesConfig
+        kwargs["quantization_config"] = BitsAndBytesConfig(
+            load_in_4bit=True,
+            bnb_4bit_compute_dtype=torch.float16,
+        )
 
+    model = AutoModelForCausalLM.from_pretrained(MODEL_ID, **kwargs)
+    model.eval()
+    print("Model ready.\n")
+    return tokenizer, model
 
-"""
-For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
-"""
-chatbot = gr.ChatInterface(
-    respond,
-    additional_inputs=[
-        gr.Textbox(value="You are a friendly Chatbot, Hailey Ho.", 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)",
-        ),
-    ],
-)
-
-with gr.Blocks() as demo:
-    with gr.Sidebar():
-        gr.LoginButton()
-    chatbot.render()
+
+def generate(tokenizer, model, prompt: str, max_new_tokens: int = 256) -> str:
+    # Qwen3 uses a chat template; fall back to raw prompt if needed
+    messages = [{"role": "user", "content": prompt}]
+    try:
+        text = tokenizer.apply_chat_template(
+            messages,
+            tokenize=False,
+            add_generation_prompt=True,
+            enable_thinking=False,   # set True for chain-of-thought (Qwen3)
+        )
+    except Exception:
+        text = prompt
+
+    inputs = tokenizer(text, return_tensors="pt").to(model.device)
+
+    with torch.no_grad():
+        output_ids = model.generate(
+            **inputs,
+            max_new_tokens=max_new_tokens,
+            do_sample=True,
+            temperature=0.7,
+            top_p=0.9,
+            pad_token_id=tokenizer.eos_token_id,
+        )
+
+    # Decode only the newly generated tokens
+    new_tokens = output_ids[0][inputs["input_ids"].shape[-1]:]
+    return tokenizer.decode(new_tokens, skip_special_tokens=True)
+
+
+def main():
+    tokenizer, model = load_model()
+
+    print("=" * 50)
+    print("Qwen3 local inference  |  type 'quit' to exit")
+    print("=" * 50)
+
+    while True:
+        prompt = "Give me 10 facts about portugal"
+
+        response = generate(tokenizer, model, prompt)
+        print(f"\nQwen3: {response}")
 
 
 if __name__ == "__main__":
-    demo.launch()
+    main()