pixelartist101/MyChatBot

1import gradio as gr2import torch3from transformers import AutoModelForCausalLM, AutoTokenizer4from peft import PeftModel5 6# Your repository details7base_model_name = "Qwen/Qwen2.5-7B" # Use the exact base you trained on8adapter_name = "pixelartist101/MyChatBot"9 10# Load the model and tokenizer11tokenizer = AutoTokenizer.from_pretrained(base_model_name)12base_model = AutoModelForCausalLM.from_pretrained(13    base_model_name, 14    torch_dtype=torch.float16, 15    device_map="auto"16)17 18# Merge your adapter weights19model = PeftModel.from_pretrained(base_model, adapter_name)20 21def chat(message, history):22    inputs = tokenizer(message, return_tensors="pt").to("cuda" if torch.cuda.is_available() else "cpu")23    outputs = model.generate(**inputs, max_new_tokens=100)24    response = tokenizer.decode(outputs[0], skip_special_tokens=True)25    return response26 27gr.ChatInterface(chat).launch()