Merged LoRA + Phi Model
This repository contains a fully merged LoRA adapter with the Phi base model.
Usage
from transformers import AutoModelForCausalLM, AutoTokenizer
model = AutoModelForCausalLM.from_pretrained("XabUG7/itsasargi", device_map="auto")
tokenizer = AutoTokenizer.from_pretrained("XabUG7/itsasargi")
inputs = tokenizer("Hello!", return_tensors="pt").to(model.device)
outputs = model.generate(**inputs, max_new_tokens=50)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))
