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

Support FP16 for user buffer #690

Merged
merged 6 commits into from
Mar 7, 2024
Merged

Conversation

lhb8125
Copy link
Contributor

@lhb8125 lhb8125 commented Feb 29, 2024

No description provided.

Signed-off-by: Hongbin Liu <[email protected]>
@timmoon10 timmoon10 self-requested a review March 4, 2024 22:42
@timmoon10
Copy link
Collaborator

/te-ci pytorch

Copy link
Collaborator

@timmoon10 timmoon10 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM. Have you tested that userbuffers works as expected with FP16 in your use-case?

@ksivaman
Copy link
Member

ksivaman commented Mar 6, 2024

/te-ci pytorch

Signed-off-by: Kirthi Shankar Sivamani <[email protected]>
@ksivaman
Copy link
Member

ksivaman commented Mar 7, 2024

/te-ci pytorch

Copy link
Member

@ksivaman ksivaman left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Looks good

@ksivaman ksivaman merged commit 8255f87 into NVIDIA:main Mar 7, 2024
20 checks passed
@lhb8125
Copy link
Contributor Author

lhb8125 commented Mar 8, 2024

LGTM. Have you tested that userbuffers works as expected with FP16 in your use-case?

@timmoon10 A China customer, Baidu, verified the fp16+user buffer by hard-coding the dtype to fp16.

rachitgarg91 pushed a commit to rachitgarg91/TransformerEngine that referenced this pull request Mar 12, 2024
* add_dtype_for_userbuf

Signed-off-by: Hongbin Liu <[email protected]>

* Update transformer_engine/pytorch/module/base.py

Signed-off-by: Tim Moon <[email protected]>

* Fix syntax

Signed-off-by: Kirthi Shankar Sivamani <[email protected]>

* Fix lint

Signed-off-by: Kirthi Shankar Sivamani <[email protected]>

---------

Signed-off-by: Hongbin Liu <[email protected]>
Signed-off-by: Tim Moon <[email protected]>
Signed-off-by: Kirthi Shankar Sivamani <[email protected]>
Co-authored-by: Hongbin Liu <[email protected]>
Co-authored-by: Tim Moon <[email protected]>
Co-authored-by: Kirthi Shankar Sivamani <[email protected]>
Signed-off-by: rachitg <[email protected]>
Kite0011 pushed a commit to Kite0011/TransformerEngine that referenced this pull request Mar 19, 2024
* add_dtype_for_userbuf

Signed-off-by: Hongbin Liu <[email protected]>

* Update transformer_engine/pytorch/module/base.py

Signed-off-by: Tim Moon <[email protected]>

* Fix syntax

Signed-off-by: Kirthi Shankar Sivamani <[email protected]>

* Fix lint

Signed-off-by: Kirthi Shankar Sivamani <[email protected]>

---------

Signed-off-by: Hongbin Liu <[email protected]>
Signed-off-by: Tim Moon <[email protected]>
Signed-off-by: Kirthi Shankar Sivamani <[email protected]>
Co-authored-by: Hongbin Liu <[email protected]>
Co-authored-by: Tim Moon <[email protected]>
Co-authored-by: Kirthi Shankar Sivamani <[email protected]>
Signed-off-by: kitefang <[email protected]>
Kite0011 pushed a commit to Kite0011/TransformerEngine that referenced this pull request Mar 20, 2024
* add_dtype_for_userbuf

Signed-off-by: Hongbin Liu <[email protected]>

* Update transformer_engine/pytorch/module/base.py

Signed-off-by: Tim Moon <[email protected]>

* Fix syntax

Signed-off-by: Kirthi Shankar Sivamani <[email protected]>

* Fix lint

Signed-off-by: Kirthi Shankar Sivamani <[email protected]>

---------

Signed-off-by: Hongbin Liu <[email protected]>
Signed-off-by: Tim Moon <[email protected]>
Signed-off-by: Kirthi Shankar Sivamani <[email protected]>
Co-authored-by: Hongbin Liu <[email protected]>
Co-authored-by: Tim Moon <[email protected]>
Co-authored-by: Kirthi Shankar Sivamani <[email protected]>
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