DeepSpeed
a5adb90d - Enabling CUDA-graph for the bert-type models (#1952)

Commit

3 years ago

Enabling CUDA-graph for the bert-type models (#1952) Co-authored-by: Olatunji Ruwase <olruwase@microsoft.com> Co-authored-by: Jeff Rasley <jerasley@microsoft.com>

References

#1952 - Enabling CUDA-graph for the bert-type models

Author

RezaYazdaniAminabadi

Parents

5053217e

Files2

deepspeed
- __init__.py
- inference
  - engine.py

DeepSpeed a5adb90d - Enabling CUDA-graph for the bert-type models (#1952)

DeepSpeed
a5adb90d - Enabling CUDA-graph for the bert-type models (#1952)