fbgemm-xpu: add CI workflow - #86
Conversation
|
There are couple attempts to break current CI workflow into dedicated torchcodec and fbgemm parts (#86 and #88). Unfortunately we will have some issues with this approach centered around the fact that we have one more package to take care about - Overall, I suggest a little bit different strategy. I think we should not break current monolithic CI workflow. I believe that it's better to have single pass to build all the components - this part is cheap enough and 90% the same across both projects. What we need to do is to optimize test time. For that I suggest to add skips of the test jobs depending on the modified paths. Here is respective PR, please, take a look: @flezaalv , can you, please, modify the PR building on top of #93? |
I will do |
Signed-off-by: Felipe Leza Alvarez <felipe.leza.alvarez@intel.com>
… triggers Signed-off-by: Felipe Leza Alvarez <felipe.leza.alvarez@intel.com>
Signed-off-by: Felipe Leza Alvarez <felipe.leza.alvarez@intel.com>
Signed-off-by: Felipe Leza Alvarez <felipe.leza.alvarez@intel.com>
Signed-off-by: Felipe Leza Alvarez <felipe.leza.alvarez@intel.com>
Signed-off-by: Felipe Leza Alvarez <felipe.leza.alvarez@intel.com>
Signed-off-by: Felipe Leza Alvarez <felipe.leza.alvarez@intel.com>
Signed-off-by: Felipe Leza Alvarez <felipe.leza.alvarez@intel.com>
4956d82 to
606c4a8
Compare
Signed-off-by: Felipe Leza Alvarez <felipe.leza.alvarez@intel.com>
dvrogozh
left a comment
There was a problem hiding this comment.
Can you, please, move packages/fbgemm-xpu/test/patches/.gitkeep to packages/fbgemm-xpu/patches/.gitkeep? This will correlate with the placement of the patches for torchcodec.
Signed-off-by: Felipe Leza Alvarez <felipe.leza.alvarez@intel.com>
Done. |
Signed-off-by: Felipe Leza Alvarez <felipe.leza.alvarez@intel.com>
Signed-off-by: Felipe Leza Alvarez <felipe.leza.alvarez@intel.com>
Co-authored-by: Alberto <109159606+aagalleg@users.noreply.github.com>
cc: @aagalleg, @manuelhsantana, @luis-real