File size: 1,124 Bytes
e27b9eb e7020af 3c4bdb3 e27b9eb e7020af e27b9eb e7020af e27b9eb 3c4bdb3 e27b9eb e7020af e27b9eb e7020af e27b9eb e7020af e27b9eb e7020af |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 |
import torch
from unsloth import FastLanguageModel
# Check if CUDA is available
device = "cuda" if torch.cuda.is_available() else "cpu"
# Load the base model
base_model_name = "unsloth/Llama-3.2-3B-Instruct"
base_model, tokenizer = FastLanguageModel.from_pretrained(
model_name=base_model_name,
max_seq_length=2048,
dtype=None, # Auto-detect data type
load_in_4bit=False, # Disable 4-bit quantization for CPU
)
base_model.to(device)
# Apply LoRA adapters
from peft import PeftModel
lora_model_name = "oskaralf/lora_model" # Replace with your LoRA model path
model = PeftModel.from_pretrained(base_model, lora_model_name)
model.to(device)
# Prepare for inference
FastLanguageModel.for_inference(model)
# Gradio interface
import gradio as gr
def chatbot(input_text):
inputs = tokenizer(input_text, return_tensors="pt").to(device)
outputs = model.generate(input_ids=inputs["input_ids"], max_new_tokens=64)
response = tokenizer.decode(outputs[0], skip_special_tokens=True)
return response
iface = gr.Interface(fn=chatbot, inputs="text", outputs="text", title="Chatbot")
iface.launch()
|