Optim-wip: Add the pre-trained InceptionV1 Places365 model #935
Conversation
|
@ProGamerGov, is it possible to also include the test fixes ? CircleCI is now failing. |
|
@NarineK I fixed the black linting errors. |
captum/attr/_core/noise_tunnel.py
Outdated
| @@ -205,7 +205,7 @@ def compute_expected_attribution_and_sq(attribution): | |||
|
|
|||
| attribution = attribution.view(attribution_shape) | |||
| expected_attribution = attribution.mean(dim=1, keepdim=False) | |||
| expected_attribution_sq = torch.mean(attribution ** 2, dim=1, keepdim=False) | |||
| expected_attribution_sq = torch.mean(attribution**2, dim=1, keepdim=False) | |||
There was a problem hiding this comment.
@ProGamerGov, this PR has also formatted the files in the non optim related code. Perhaps you can bring this back. Also, maybe you have a different version of code formatter that's why it changed the formatting ?
There was a problem hiding this comment.
@NarineK The black tests fail currently because the optim-wip branch's version of black was never pinned to a specific version. I've reverted the changes as they are not required in the master branch's pinned version of black.
|
@ProGamerGov, let's remove the changes that aren't related to optim and copy the fix related to flask. Is this flask error related to flask version ? |
ProGamerGov
force-pushed
the
optim-wip-places365-new
branch
from
249f2d9 to
418027c
Compare
|
@NarineK It looks like the Conda failure is due to the version of flask, but I'm not sure what the fix would be. It's also unrelated to the optim module. |
See: #655 for more details on this PR