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

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

jeffra merged 4 commits into master from bert-cuda-graph

enabling cuda-graph for the bert-type models

9a7e056f

add enable_cuda_graph at init_inference

fc0bcd86

RezaYazdaniAminabadi requested a review from

jeffra 3 years ago

RezaYazdaniAminabadi requested a review from

samyam 3 years ago

RezaYazdaniAminabadi requested a review from

tjruwase 3 years ago

RezaYazdaniAminabadi requested a review from

ShadenSmith 3 years ago

RezaYazdaniAminabadi requested a review from

conglongli 3 years ago

RezaYazdaniAminabadi requested a review from

awan-10 3 years ago

RezaYazdaniAminabadi requested a review from

cli99 3 years ago

RezaYazdaniAminabadi requested a review from

eltonzheng 3 years ago

RezaYazdaniAminabadi requested a review from

minjiaz 3 years ago

Merge branch 'master' into bert-cuda-graph

4332342a

explictly opt-in to cuda-graph mode

5780b53a

jeffra approved these changes on 2022-05-17

jeffra enabled auto-merge (squash) 3 years ago

jeffra merged a5adb90d into master 3 years ago

jeffra deleted the bert-cuda-graph branch 3 years ago

Reviewers

jeffra

samyam

tjruwase

ShadenSmith

conglongli

awan-10

cli99

eltonzheng

minjiaz

Assignees

No one assigned

Labels

None yet

Milestone

No milestone

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

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

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

Merged