try minimal requirements

[Borda]

What does this PR do?

unfreeze requirements, cc: @SeanNaren

Before submitting

• Was this discussed/approved via a Github issue? (no need for typos and docs improvements)
• Did you read the contributor guideline, Pull Request section?
• Did you make sure your PR does only one thing, instead of bundling different changes together?
• Did you make sure to update the documentation with your changes?
• Did you write any new necessary tests? [not needed for typos/docs]
• Did you verify new and existing tests pass locally with your changes?
• If you made a notable change (that affects users), did you update the CHANGELOG?

PR review

• Is this pull request ready for review? (if not, please submit in draft mode)

Anyone in the community is free to review the PR once the tests have passed.
If we didn't discuss your PR in Github issues there's a high chance it will not be merged.

Did you have fun?

Make sure you had fun coding 🙃

[codecov]

Codecov Report

Merging #62 (957ff0a) into master (522cf75) will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff           @@
##           master      #62   +/-   ##
=======================================
  Coverage   87.39%   87.39%           
=======================================
  Files          49       49           
  Lines        1579     1579           
=======================================
  Hits         1380     1380           
  Misses        199      199           

Flag	Coverage Δ
unittests	87.39% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 522cf75...991678c. Read the comment docs.

[carmocca]

I'm not sure we can have torch<1.7.1 because for earlier versions, the loaded weights in the examples will fail due to PyTorch breaking compatibility in their serialization

[Borda]

I'm not sure we can have torch<1.7.1 because for earlier versions, the loaded weights in the examples will fail due to PyTorch breaking compatibility in their serialization

good point but in such case, we shall have it tested somewhere... anyway fix free on one particular version is very bad :/

[carmocca]

We have a test for it

https://github.com/PyTorchLightning/lightning-flash/blob/3c5b535ce26a12fd141acff7e20bc2b18ee56a0a/tests/core/test_model.py#L115-L129

We would need to run it with different PyTorch versions

[Borda]

I'm not sure we can have torch<1.7.1 because for earlier versions, the loaded weights in the examples will fail due to PyTorch breaking compatibility in their serialization

but this breaking change was in 1.6? @SeanNaren

[carmocca]

Yes, the change was in 1.6 but we had some issues before release and fixed it to 1.7.1 to be safe.

Can you add a single job with PyTorch 1.6 in CI? To be sure

[Borda]

Can you add a single job with PyTorch 1.6 in CI? To be sure

now the min is 1.6 so all 6 tests are using it - Linux/Mac/Win and py3.6 and py3.8

[carmocca]

We need to think about how to solve the weights compatibility issue. See failing tests:

https://github.com/PyTorchLightning/lightning-flash/blob/90a5d798d7d494cc66ca3e9b1294679769095f32/tests/core/test_model.py#L115-L129

RuntimeError: Only one file(not dir) is allowed in the zipfile

The weights were generated with torch==1.7.1 and fail to load with torch==1.6.0

[Borda]

The weights were generated with torch==1.7.1 and fail to load with torch==1.6.0

Yes, the weights shall be generated in 1.6 so they would be compatible with all after 1.6...

[carmocca]

Yes, the weights shall be generated in 1.6 so they would be compatible with all after 1.6...

That's what we thought initially but didn't seem to work before release. We'll have to try again

[Borda]

@carmocca it is also what I did for PL checkpoint legacy testing, all are created with PT 1.4 as they are possible to load in any future version up to 1.8

[Borda]

@carmocca how is it going with generating weights with lower pytorch?

added todo

[carmocca]

@carmocca how is it going with generating weights with lower pytorch?

Bit busy with the 1.2 release. @teddykoker can you do it? I believe you did it originally

[Borda]

@carmocca how is it going with generating weights with lower pytorch?

Bit busy with the 1.2 release. @teddykoker can you do it? I believe you did it originally

no rush, I have created an issue for it... #128

[Borda]

ok, let's try reverse strategy :D

[Borda]

The Win-latest is failing also on master - https://github.com/PyTorchLightning/lightning-flash/runs/1936420296