Worked on the facebookresearch/fairseq2 repository to enhance the end-to-end fine-tuning documentation, focusing on distributed training workflows. Updated the RST documentation to clarify the use of the --gpus-per-node flag within the srun command, providing explicit guidance for configuring multi-GPU environments. This documentation improvement aimed to align with best practices in distributed training and reduce user friction during setup. The work emphasized clear communication and onboarding for developers, ensuring that CLI arguments were accurately described. Leveraged skills in technical writing and documentation, using RST to deliver precise, actionable instructions that help prevent misconfigurations in complex training scenarios.