[C/PyTorch/Jax] Add support for more bias shapes #677
Merged
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: Alp Dener <[email protected]>
Signed-off-by: Alp Dener <[email protected]>
Signed-off-by: Charlene Yang <[email protected]>
cyanguwa
changed the title
cyanguwa
changed the title
Signed-off-by: Charlene Yang <[email protected]>
…ze' into fused_attn/add_dbias_shapes_c_pytorch Signed-off-by: Alp Dener <[email protected]>
cyanguwa
changed the title
cyanguwa
changed the title
Signed-off-by: Charlene Yang <[email protected]>
|
/te-ci |
Signed-off-by: Alp Dener <[email protected]>
Signed-off-by: Alp Dener <[email protected]>
Signed-off-by: Alp Dener <[email protected]>
|
/te-ci jax |
Signed-off-by: Alp Dener <[email protected]>
Signed-off-by: Alp Dener <[email protected]>
|
/te-ci jax |
|
Pytorch pipeline with CUDA 12.3: 13014488. Fixing A100 errors now. |
Signed-off-by: Charlene Yang <[email protected]>
|
Pipeline 13043757 for PyTorch and Paddle has passed. |
Signed-off-by: Alp Dener <[email protected]>
|
/te-ci jax |
…h JAX Signed-off-by: Alp Dener <[email protected]>
…and h_q == h_kv conditions Signed-off-by: Alp Dener <[email protected]>
|
With the latest commit, all FP16 tests in TE/JAX CI are now passing with The BF16 failures look like this:
It's not clear to me yet if these BF16 failures are due to a bug in the pure-JAX/Flax reference function or the TE/JAX fused attn custom op. |
zlsh80826
reviewed
Feb 27, 2024
…27 for Bfloat16 and -2**15 for Float16 Signed-off-by: Alp Dener <[email protected]>
Signed-off-by: Alp Dener <[email protected]>
|
/te-ci jax |
Signed-off-by: Alp Dener <[email protected]>
|
/te-ci jax |
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.
This PR
[b,1,s,s],[b,h,s,s],[1,1,s,s]bias shapes, when dBias is not required. This is applicable to inference (bias.requires_grad = False), when bias is a workaround for an arbitrary mask (True/False -> 0/-inf), or when ALiBi slopes tensor is in[b,h]shape.F16_arbitrary_seqlenbackend of the cuDNN fused attention only.