Verl Configuration on GPU: [https://github.com/rllm-org/rllm/blob/e24d8c82e0fb5f1f8f9729d73f6722f35d0e980a/examples/swe/train_deepswe_32b.sh#L19](<https://github.com/rllm-org/rllm/blob/e24d8c82e0fb5f1f8f9729d73f6722f35d0e980a/examples/swe/train_deepswe_32b.sh#L19>)