using model.log outside model

[sordonia]

Hi all :)

I am trying to log results of a computation happening outside the model class itself, after trainer.fit(model, datamodule) has taken place.

# average over top-5 checkpoints
test_results = []
for ckpt in glob.glob(hparams.save_dir + '/*.ckpt'):
  test_results.append(trainer.test(ckpt_path=ckpt))

test_results = pandas.DataFrame(test_results)
for item, value in test_results.mean(0).to_dict().items():
  model.log(f'{item}_mean', value, on_step=True)

The problem is that these results are not getting logged. Am I doing something wrong?

Thanks a lot for the support!