GPU out of memory in Jupyter

Question

My OS specs are 8GB dedicated GPU and 24GB shared GPU, I am trying to fine tune a model and I am facing CUDA out of memory errors . I am not able to use Shared memory. Is there any way to use it when my dedicated memory was full.

Code Snippet for reference

training_args = TrainingArguments(
        output_dir=local_output_dir,
        per_device_train_batch_size=1,
        per_device_eval_batch_size=1,
        fp16=False,
        bf16=False,
        learning_rate=1e-1,
        num_train_epochs=2,
        deepspeed=None,
        gradient_checkpointing=True,
        logging_dir=f"{local_output_dir}/runs",
        logging_strategy="steps",
        logging_steps=10,
        evaluation_strategy="steps",
        eval_steps=10,
        save_strategy="steps",
        save_steps=20,
        save_total_limit=10,
        load_best_model_at_end=False,
        report_to="tensorboard",
        disable_tqdm=True,
        remove_unused_columns=False,
        local_rank=2,
        warmup_steps=0,
    )

trainer = Trainer(
    model=model,
    tokenizer=tokenizer,
    args=training_args,
    train_dataset=split_dataset["train"],
    eval_dataset=split_dataset["test"],
)