import json, spaces, torch, gradio as gr
from transformers import AutoTokenizer, AutoModelForCausalLM
M="Qwen/Qwen3-14B"
tok=AutoTokenizer.from_pretrained(M)
model=AutoModelForCausalLM.from_pretrained(M, torch_dtype=torch.bfloat16).to("cuda")
@spaces.GPU(duration=40)
def chat(messages_json, max_tokens=300):
    msgs=json.loads(messages_json)
    ids=tok.apply_chat_template(msgs, add_generation_prompt=True, enable_thinking=False, return_tensors="pt").to("cuda")
    out=model.generate(ids, max_new_tokens=int(max_tokens), do_sample=True, temperature=0.8, top_p=0.9)
    return tok.decode(out[0][ids.shape[1]:], skip_special_tokens=True).strip()
gr.Interface(chat,[gr.Textbox(),gr.Number(value=300)],gr.Textbox(),api_name="chat").queue().launch()