Give file_pretrained_model an example

[rcannood]

Describe your changes

novel_train and simple_mlp_train cannot pass viash test on main, and haven't been able to for a while:

+/tmp/viash_test_novel_train_.../test_run_and_check_output/test_executable ---cpus=2 ---memory=14gb
[error] --output is a required argument. Use --help to get more information on the parameters.
AssertionError: Script exited with an error. Return code: 1

get_argument_sets() only gives an argument a value when the spec has a default or an example:

if arg.get("default") is not None:
    default_or_example = arg["default"]
elif arg.get("example") is not None:
    default_or_example = arg["example"]
...
if default_or_example is not None and arg["type"] == "file":

file_pretrained_model.yaml was the only file spec with neither, so --output never made it onto the command line:

  --input_train_mod1  type=file  dir=input   example=[...train_mod1.h5ad]
  --input_train_mod2  type=file  dir=input   example=[...train_mod2.h5ad]
  --input_test_mod1   type=file  dir=input   example=[...test_mod1.h5ad]
  --output            type=file  dir=output  example=None   <-- here

With an example added:

  novel_train          [run] --output passed: True
  simple_mlp_train     [run] --output passed: True
  novel_predict        [with_model] --output passed: True
  simple_mlp_predict   [with_model] --output passed: True

For an output argument only the extension is read off the example, so a directory-shaped path gives the right thing (the pretrained model is a directory). The predict components already override --input_model through info.test_setup, so they're unaffected either way.

Worth knowing: this is why #46 is red, and #33 was red for the same reason before it was merged. CI only tests components a PR touches, so it only shows up when someone edits novel_train or simple_mlp_train -- and then it looks like that PR broke something.

Part of a series of PRs coming out of a pre-run review of the benchmark.

Checklist before requesting a review

• I have performed a self-review of my code

• Check the correct box. Does this PR contain:

  • Breaking changes
  • New functionality
  • Major changes
  • Minor changes
  • Bug fixes

• Proposed changes are described in the CHANGELOG.md

• CI Tests succeed and look good!

[rcannood]

these examples should be set in the predict methods under info.test_setup.with_model.input_model because they are different for each method.