Megatron-DeepSpeed
f01fa4a5 - make the tests adaptable to the number of available gpus

Commit

3 years ago

make the tests adaptable to the number of available gpus

References

#239 - Reshape deepspeed checkpoint

#289 - No-ZeRO reshaping

#292 - a branch combining layer-norm-auto-sync and ds_ckpt_reshape

Author

stas00

stas00

Parents

Loading