populate some more legacy checkpoints

[Borda]

What does this PR do?

adding some more legacy checkpoints, followup on #5223

Before submitting

• Was this discussed/approved via a GitHub issue? (not for typos and docs)
• 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? (if necessary)
• Did you write any new necessary tests? (not for typos and docs)
• Did you verify new and existing tests pass locally with your changes?
• Did you update the CHANGELOG? (not for typos, docs, test updates, or internal minor changes/refactorings)

PR review

Anyone in the community is free to review the PR once the tests have passed.
Before you start reviewing make sure you have read Review guidelines. In short, see the following bullet-list:

• Is this pull request ready for review? (if not, please submit in draft mode)
• Check that all items from Before submitting are resolved
• Make sure the title is self-explanatory and the description concisely explains the PR
• Add labels and milestones (and optionally projects) to the PR so it can be classified
• Check that target branch and milestone match!

Did you have fun?

Make sure you had fun coding 🙃

[codecov]

Codecov Report

Merging #5457 (0f5d2ca) into master (499d503) will decrease coverage by 0%.
The diff coverage is n/a.

@@          Coverage Diff           @@
##           master   #5457   +/-   ##
======================================
- Coverage      93%     93%   -0%     
======================================
  Files         135     135           
  Lines       10011   10011           
======================================
- Hits         9324    9317    -7     
- Misses        687     694    +7     

[Borda]

well not sure why the newly added ckpt fails, locally it works fine...

[Borda]

seems that there is a problem with PT, as the same files can be loaded in 1.5 1.6 1.7 but not in 1.4
and it includes also the later PL versions such as 1.1.3

FAILED tests/checkpointing/test_legacy_checkpoints.py::test_resume_legacy_checkpoints[0.8.3]
FAILED tests/checkpointing/test_legacy_checkpoints.py::test_resume_legacy_checkpoints[0.8.4]
FAILED tests/checkpointing/test_legacy_checkpoints.py::test_resume_legacy_checkpoints[0.8.5]
FAILED tests/checkpointing/test_legacy_checkpoints.py::test_resume_legacy_checkpoints[0.9.0]
FAILED tests/checkpointing/test_legacy_checkpoints.py::test_resume_legacy_checkpoints[1.0.8]
FAILED tests/checkpointing/test_legacy_checkpoints.py::test_resume_legacy_checkpoints[1.1.0]
FAILED tests/checkpointing/test_legacy_checkpoints.py::test_resume_legacy_checkpoints[1.1.1]
FAILED tests/checkpointing/test_legacy_checkpoints.py::test_resume_legacy_checkpoints[1.1.2]
FAILED tests/checkpointing/test_legacy_checkpoints.py::test_resume_legacy_checkpoints[1.1.3]

[carmocca]

seems that there is a problem with PT, as the same files can be loaded in 1.5 1.6 1.7 but not in 1.4
and it includes also the later PL versions such as 1.1.3

what's the error?

[Borda]

seems that there is a problem with PT, as the same files can be loaded in 1.5 1.6 1.7 but not in 1.4 and it includes also the later PL versions such as 1.1.3

what's the error?

PT somehow states that the ckpt is an archive...

      if _is_zipfile(f):
          # .zip is used for torch.jit.save and will throw an un-pickling error here
>         raise RuntimeError("{} is a zip archive (did you mean to use torch.jit.load()?)".format(f.name))
E         RuntimeError: /home/runner/work/pytorch-lightning/pytorch-lightning/legacy/checkpoints/1.1.3/epoch=4-step=319.ckpt is a zip archive (did you mean to use torch.jit.load()?)

[Borda]

this seems to be very similart issue:

• torch.load issue on loading file created by torch.save pytorch/pytorch#25214

so the problem is that we cannot save checkpoints with and load with lover version, the breaking change is in 1.7 so for creating legacy checkpoints we have to use the lowest supported version...

[Borda]

ok, seems that there is some problem with v0.8.5 and v0.9.0 so I would leave it for another PR...
eventually, we may consider making some workaround to transform archive to loadable format or use torch.jit