Skip to content
Closed
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
8 changes: 3 additions & 5 deletions .azure-pipelines/azure-pipelines-linux.yml

Some generated files are not rendered by default. Learn more about how customized files appear on GitHub.

25 changes: 23 additions & 2 deletions .gitignore

Some generated files are not rendered by default. Learn more about how customized files appear on GitHub.

16 changes: 9 additions & 7 deletions .scripts/build_steps.sh

Some generated files are not rendered by default. Learn more about how customized files appear on GitHub.

4 changes: 2 additions & 2 deletions .scripts/logging_utils.sh

Some generated files are not rendered by default. Learn more about how customized files appear on GitHub.

9 changes: 9 additions & 0 deletions .scripts/run_docker_build.sh

Some generated files are not rendered by default. Learn more about how customized files appear on GitHub.

2 changes: 1 addition & 1 deletion README.md

Some generated files are not rendered by default. Learn more about how customized files appear on GitHub.

5 changes: 3 additions & 2 deletions build-locally.py

Some generated files are not rendered by default. Learn more about how customized files appear on GitHub.

3 changes: 2 additions & 1 deletion recipe/meta.yaml
Original file line number Diff line number Diff line change
Expand Up @@ -23,7 +23,8 @@ requirements:
- pillow
- pretrainedmodels ==0.7.4
- python >=3.6
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Suggested change
- python >=3.6
- python >=3.7

Upstream says that this should be 3.7.
https://github.com/qubvel/segmentation_models.pytorch/blob/v0.3.3/setup.py#L20

though it may be entirely true that 3.6 is compatible.

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

- timm ==0.9.2
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Smp 0.3.3 has a hard pin on the version of timm, see error at https://dev.azure.com/conda-forge/feedstock-builds/_build/results?buildId=888834&view=logs&j=656edd35-690f-5c53-9ba3-09c10d0bea97&t=986b1512-c876-5f92-0d81-ba851554a0a3&l=1048:

segmentation-models-pytorch 0.3.3 has requirement timm==0.9.2, but you have timm 0.9.16

Will need to update upstream at https://github.com/qubvel/segmentation_models.pytorch/blob/6db76a1106426ac5b55f39fba68168f3bccae7f8/requirements.txt#L4, and get them to make a new release before it can be changed here.

Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Maintainer of https://github.com/conda-forge/pytorch-cpu-feedstock/ here.
Thank you for pinging me @jefromson .

Sometimes it is appropriate to patch the source code.

We do this a bit with conda-forge because often we are more at the bleeding edge of releases compared to what people are working with upstream.

It may be that we should:

  1. Set the pin to >=0.9.2,<0.10a.0a instead of ==0.9.2 with conda build's infrastructure, this translates to any "version 0.9".
  2. Patch the file you pointed to with a conda-forge patch. Preferably, we leave a comment pointing to a tracking issue upstream, or even more preferred, a PR showing that the proposed patch was merged and is slated for the next release.

This is sometimes quite risky without tests. I can help try to run the small test suite in the recipe if you want.

Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

#23 <-- leaving this here as an example of what I have in mind. but I haven't run any analysis code, so it might be fully broken.

I did leave a reference to the open issue asking for a new release with expanded timm compatibility.

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Keeping an eye on qubvel-org/segmentation_models.pytorch#885 for smp to support timm 1.x

- six
- timm
- torchvision >=0.5.0
- tqdm

Expand Down