search

tenstorrent / tt-metal

:metal: TT-NN operator library, and TT-Metalium low level kernel programming model.
Apache License 2.0
396 stars 48 forks source link

SSD CPU Implementation #897

Closed boris-drazic closed 1 year ago

boris-drazic commented 1 year ago

Create a test for SSD (Single Shot MultiBox Detector) CPU implementation. This test will run fully in PyTorch on CPU and will not use TT accelerator device.

This test should be organized in three parts:

  1. input processing
  2. running SSD
  3. output processing

The test should take images as inputs and provide images with detected objects as output.

Primary purpose of this test is to set up all that is needed to start implementing using TT ops and accelerator device.

ruthreshx commented 1 year ago

@boris-drazic , The CPU implementation of the SSD model is complete. A PR has been created for the same here The corresponding commit is here

Kindly review.

ruthreshx commented 1 year ago

Hi @boris-drazic , Please find the new commit here

Kindly review, Thanks.

ruthreshx commented 1 year ago

Hi @boris-drazic , I have run the pipelines (regression and post_commit) to ensure the commit passing in CI. Both pipelines were passing on this new commit. Kindly review, Thanks

ruthreshx commented 1 year ago

@boris-drazic @tt-rkim, The CPU implementation of the SSD model is completed by following the VGG model standards. Please find the commit here. Kindly review.

tt-rkim commented 1 year ago

Hi @ruthreshx - I appreciate the tagging, but please from now on refer to @farbabi @boris-drazic @ashayestehmanesh for models review. This is now their domain

Furthermore, tagging us to review on the issue rather than the PR is redundant. Please restrict tagging to just the PR review assignment, as opposed to a comment in the issue or a comment in the PR.