Commit 907ad20

authored

【Inference】Optimize top_p kernel performance (#9132)

* add top_p kernel * fix code style

1 parent c93bada commit 907ad20Copy full SHA for 907ad20

5 files changed

+791

-3

lines changed

csrc
- gpu/sample_kernels
- setup_cuda.py
paddlenlp/experimental/transformers
- generation_utils.py

5 files changed

+791

-3

lines changed

Comments

(0)