import os
os.environ["CUDA_DEVICE_ORDER"]="PCI_BUS_ID"
os.environ["CUDA_VISIBLE_DEVICES"] = "1"
from transformers import AutoTokenizer, AutoModelForCausalLM

name = "google/gemma-3-4b-it"  # or your Gemma 3 variant
tokenizer = AutoTokenizer.from_pretrained(name)
model = AutoModelForCausalLM.from_pretrained(name)

print(tokenizer.eos_token, tokenizer.eos_token_id)
print(model.config.eos_token, model.config.eos_token_id)