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

Rearrange inference OPS and stop using builder.load #5490

Merged
merged 64 commits into from
Oct 9, 2024

Conversation

oelayan7
Copy link
Contributor

@oelayan7 oelayan7 commented May 1, 2024

This PR mainly handles all places where InferenceBuilder is used to
access any op or a specific implementation for an op.
Instead an op is defined, and its proper implementation is picked inside
and the usage will be transparent to the user.
What was done in the PR:

  1. Added missing ops (added a py file with fallback mechanism)
  2. Added missing fallback implementations for existing ops
  3. removed all usages for builder.load and replaced them with ops instead.
  4. added workspace op and inferenceContext which contains all workspace related functions and inferenceContext is the python fallback of inferenceContext in CUDA
  5. a small change to softmax_context signature to fit the fallback signature.

@loadams
Copy link
Contributor

loadams commented May 22, 2024

Hi @oelayan7 - thanks for the contribution, could you take a look at the failing tests?

@lekurile
Copy link
Contributor

Kicked off a manual run of the nv-ds-chat GH workflow since this PR modifies the Hybrid Engine:
https://github.com/microsoft/DeepSpeed/actions/runs/9199337454

@loadams, @jomayeri, FYI.

This PR mainly handles all places where InferenceBuilder is used to
access any op or a specific implementation for an op.
Instead an op is defined, and its proper implementation is picked inside
and the usage will be transparent to the user.
What was done in the PR:
1) Added missing ops (added a py file with fallback mechanism)
2) Added missing fallback implementations for existing ops
3) removed all usages for builder.load and replaced them with ops instead.
4) a small change to softmax_context signature to fit the fallback signature.
@oelayan7
Copy link
Contributor Author

@loadams @lekurile it is fixed now, can you please trigger the unit tests?

@lekurile
Copy link
Contributor

@loadams @lekurile it is fixed now, can you please trigger the unit tests?

Hi @oelayan7, kicked off the unit tests.

@oelayan7
Copy link
Contributor Author

@loadams Can you please trigger the CI (hopefully the last time :) )

@loadams
Copy link
Contributor

loadams commented Jun 26, 2024

@loadams Can you please trigger the CI (hopefully the last time :) )

@oelayan7 - CI failures appear to be real for a few things if you could take a look

@loadams
Copy link
Contributor

loadams commented Sep 25, 2024

@oelayan7 - I've got to push a fix for nv-accelerate tests, will get this merged shorlty.

@loadams loadams added this pull request to the merge queue Oct 8, 2024
@loadams loadams removed this pull request from the merge queue due to a manual request Oct 8, 2024
@loadams loadams enabled auto-merge October 8, 2024 18:17
@loadams loadams added this pull request to the merge queue Oct 9, 2024
Merged via the queue into microsoft:master with commit 645639b Oct 9, 2024
16 checks passed
@oelayan7 oelayan7 deleted the rearrange_ops branch October 9, 2024 07:41
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.

5 participants