DeepSpeed
Fix the tensor-slicing with multi-GPU inference and kernel-injection
#1724
Merged
Merged