Current text: "to run our training distributed across multiple gpus." Correction: "to run our training in distributed mode across multiple GPUs."