Skip to content
This repository has been archived by the owner on Aug 26, 2022. It is now read-only.

Commit

Permalink
Update USAGE.md
Browse files Browse the repository at this point in the history
  • Loading branch information
hyunwoongko authored Jan 15, 2022
1 parent 504944a commit 4b33288
Showing 1 changed file with 3 additions and 2 deletions.
5 changes: 3 additions & 2 deletions USAGE.md
Original file line number Diff line number Diff line change
Expand Up @@ -8,8 +8,8 @@
- [Saving Checkpoints](#saving-checkpoints)
- [Kernel Fusion](#kernel-fusion)
- [Fused MLP and Softmax Kernels](#fused-mlp-and-softmax-kernels)
- [Selective Kernel Fusion](#selective-kernel-fusion)
- [Fused N-Gram Blocking Kernels](#fused-n-gram-blocking-kernels)
- [Selective Kernel Fusion](#selective-kernel-fusion)
- [DeepSpeed Support](#deepspeed-support)
- [Initialization](#initialization)
- [Training](#training-1)
Expand Down Expand Up @@ -342,7 +342,8 @@ model.generate(..., fused_no_repeat_ngram_blocking=True)

## Selective Kernel Fusion
Since version 1.1.2, you can fuse only partial kernels, not all kernels. Currently, only Attention class and MLP class are supported.
```

```python
from oslo import GPT2MLP, GPT2Attention

# MLP only fusion
Expand Down

0 comments on commit 4b33288

Please sign in to comment.