make TransformerLayer accept a bshd or sbhd tensor format
#557
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Signed-off-by: Sudhakar Singh <sudhakars@nvidia.com>
cyanguwa
approved these changes
Dec 8, 2023
cyanguwa
reviewed
Dec 8, 2023
ptrendx
reviewed
Dec 13, 2023
Signed-off-by: Sudhakar Singh <sudhakars@nvidia.com>
…to plumb_tensor_format_thru_transformer_layer
cyanguwa
reviewed
Jan 5, 2024
cyanguwa
reviewed
Jan 5, 2024
cyanguwa
reviewed
Jan 5, 2024
cyanguwa
requested changes
Jan 5, 2024
ptrendx
reviewed
Jan 9, 2024
Signed-off-by: Sudhakar Singh <sudhakars@nvidia.com>
…to plumb_tensor_format_thru_transformer_layer
|
@sudhakarsingh27 Please resolve the merge conflicts @cyanguwa Could you rereview? |
…to plumb_tensor_format_thru_transformer_layer Signed-off-by: Sudhakar Singh <sudhakars@nvidia.com>
Signed-off-by: Sudhakar Singh <sudhakars@nvidia.com>
|
/te-ci/pytorch |
|
/te-ci pytorch |
1 similar comment
|
/te-ci pytorch |
cyanguwa
reviewed
Jan 18, 2024
cyanguwa
reviewed
Jan 18, 2024
Signed-off-by: Sudhakar Singh <sudhakars@nvidia.com>
|
@sudhakarsingh27 Please fix CI before merging -- looks like there's a few failed jobs. |
sudhakarsingh27
added a commit
to sudhakarsingh27/TransformerEngine
that referenced
this pull request
Jan 19, 2024
Signed-off-by: Sudhakar Singh <sudhakars@nvidia.com>
ksivaman
pushed a commit
that referenced
this pull request
Jan 20, 2024
Wong4j
pushed a commit
to Wong4j/TransformerEngine
that referenced
this pull request
Jan 22, 2024
…#557) * make TransformerLayer accept a `bshd` or `sbhd` tensor format Signed-off-by: Sudhakar Singh <sudhakars@nvidia.com> * Fixes from feedback Signed-off-by: Sudhakar Singh <sudhakars@nvidia.com> * more feedback fixes Signed-off-by: Sudhakar Singh <sudhakars@nvidia.com> * remove incorrect info from docstring Signed-off-by: Sudhakar Singh <sudhakars@nvidia.com> * fix from feedback Signed-off-by: Sudhakar Singh <sudhakars@nvidia.com> --------- Signed-off-by: Sudhakar Singh <sudhakars@nvidia.com>
ptrendx
pushed a commit
that referenced
this pull request
Jan 22, 2024
* make TransformerLayer accept a `bshd` or `sbhd` tensor format Signed-off-by: Sudhakar Singh <sudhakars@nvidia.com> * Fixes from feedback Signed-off-by: Sudhakar Singh <sudhakars@nvidia.com> * more feedback fixes Signed-off-by: Sudhakar Singh <sudhakars@nvidia.com> * remove incorrect info from docstring Signed-off-by: Sudhakar Singh <sudhakars@nvidia.com> * fix from feedback Signed-off-by: Sudhakar Singh <sudhakars@nvidia.com> --------- Signed-off-by: Sudhakar Singh <sudhakars@nvidia.com>
ptrendx
pushed a commit
that referenced
this pull request
Jan 22, 2024
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Changes:
DotProductAttentioncan interpret the input tensor insbhdorbshdformat butMultiHeadAttentionandTransformerLayeraren't aware of this. This PR plumbs this information throughTransformerLayer->MultiHeadAttention->DotProductAttention.RotaryPositionEmbeddingalso needs to be aware of this format to correctly apply the rope cache to theqandkinputs.Uses:
LlamaDecoderLayerwithTransformerLayer, it helps to have this control since HF isbshdby default and TE issbhd.