deepspeed --num_gpus=2 your_program.py <normal cl args> --do_eval --deepspeed ds_config.json | |
Non-Trainer DeepSpeed integration | |
DeepSpeed also works with Transformers without the [Trainer] class. |
deepspeed --num_gpus=2 your_program.py <normal cl args> --do_eval --deepspeed ds_config.json | |
Non-Trainer DeepSpeed integration | |
DeepSpeed also works with Transformers without the [Trainer] class. |