import gradio as gr from huggingface_hub import InferenceClient import torch from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer import os from threading import Thread !gdown 17_uyVadOd6pCzZ044hFOWUX4o6DkB00t with open("scraped_data.txt") as f : context = f.read() system_prompt = f"""you are twensa hosting chat bot to know more about twensa hosting this is an ad about them : {context} """ model = AutoModelForCausalLM.from_pretrained("KingNish/Qwen2.5-0.5b-Test-ft", torch_dtype=torch.float16) tokenizer = AutoTokenizer.from_pretrained("KingNish/Qwen2.5-0.5b-Test-ft") device = torch.device('cuda') model = model.to(device) def chat(message, history): chat = [{"role":"system","content":system_prompt}] for item in history: chat.append({"role": "user", "content": item[0]}) if item[1] is not None: chat.append({"role": "assistant", "content": item[1]}) chat.append({"role": "user", "content": message}) messages = tokenizer.apply_chat_template(chat, tokenize=False, add_generation_prompt=True) # Tokenize the messages string model_inputs = tokenizer([messages], return_tensors="pt").to(device) streamer = TextIteratorStreamer( tokenizer, timeout=10., skip_prompt=True, skip_special_tokens=True) generate_kwargs = dict( model_inputs, streamer=streamer, max_new_tokens=1024, do_sample=True, top_p=0.95, top_k=1000, temperature=0.75, num_beams=1, ) t = Thread(target=model.generate, kwargs=generate_kwargs) t.start() # Initialize an empty string to store the generated text partial_text = "" for new_text in streamer: partial_text += new_text yield partial_text demo = gr.ChatInterface(fn=chat, chatbot=gr.Chatbot(show_label=True, show_share_button=True, show_copy_button=True,layout="bubble", bubble_full_width=False), theme="dark", examples=[["what is twensa hosting ?"]], title="TWENSA HOSTING CHAT BOT") # Launch the app demo.launch()