Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Update demoBERT input dimensions to match Triton requirement #1051

Merged
merged 5 commits into from
May 11, 2021

Conversation

mengdong
Copy link
Contributor

@mengdong mengdong commented Feb 8, 2021

Hi @rajeevsrao, this is a similar change to #890, make builder generate an engine that works with Triton. I have tried 1 opt profile and multiple opt profiles and it seems to work well. Rebased based on the new master. I messed up last one due to a sign-off error

@rajeevsrao
Copy link
Collaborator

@mengdong this seems to be breaking the inference script. Can you please confirm?
https://github.com/NVIDIA/TensorRT/tree/master/demo/BERT#quick-start-guide

@mengdong
Copy link
Contributor Author

Yes, seems like we need to change the inference code for fixed seqlen model.

@rajeevsrao
Copy link
Collaborator

@mengdong please rebase on NVIDIA:master.

@rajeevsrao rajeevsrao force-pushed the triton-bert branch 2 times, most recently from 27b0961 to 27d4680 Compare May 11, 2021 05:41
@mengdong mengdong force-pushed the triton-bert branch 2 times, most recently from cfe52ba to ce910da Compare May 11, 2021 06:17
mengdong and others added 5 commits May 10, 2021 23:31
Signed-off-by: Dong Meng <dongm@nvidia.com>
Signed-off-by: Rajeev Rao <rajeevrao@nvidia.com>
Signed-off-by: Rajeev Rao <rajeevrao@nvidia.com>
delete checkpoints

Signed-off-by: Rajeev Rao <rajeevrao@nvidia.com>
Signed-off-by: Rajeev Rao <rajeevrao@nvidia.com>
Signed-off-by: Rajeev Rao <rajeevrao@nvidia.com>
@rajeevsrao rajeevsrao changed the title fix bert input dimension to match Triton requirement Update demoBERT input dimensions to match Triton requirement May 11, 2021
@rajeevsrao rajeevsrao merged commit ab20a8a into NVIDIA:master May 11, 2021
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants