Simple way to use the model

<code>
model_name = "sanjeev498/en-to-romanian"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForSeq2SeqLM.from_pretrained(model_name)
model = model.half()
model.to("cuda")

sentences = ["This is a test"]

inputs = tokenizer(sentences, padding=True, return_tensors="pt").to("cuda")
translated_tokens = model.generate(
**inputs
)
translations = tokenizer.batch_decode(translated_tokens, skip_special_tokens=True)
</code>