DeepSpeed
adb0b972 - align alibi&mask addition with HF new changes

Commit

2 years ago

align alibi&mask addition with HF new changes

References

#2083 - Add Inference support for running the BigScience-BLOOM Architecture

Author

Reza Yazdani

Parents

Files4

csrc/transformer/inference
- csrc
  - pt_binding.cpp
  - softmax.cu
- includes
  - custom_cuda_layers.h
deepspeed/ops/transformer/inference
- transformer_inference.py

Loading