[examples] unit test for run_bart_sum #3544
Conversation
…formers_fork into testability-run-bart-sum
| @@ -102,8 +102,8 @@ def optimizer_step(self, epoch, batch_idx, optimizer, optimizer_idx, second_orde | |||
| self.lr_scheduler.step() | |||
|
|
|||
| def get_tqdm_dict(self): | |||
| tqdm_dict = {"loss": "{:.3f}".format(self.trainer.avg_loss), "lr": self.lr_scheduler.get_last_lr()[-1]} | |||
|
|
|||
| avg_loss = getattr(self.trainer, "avg_loss", 0.0) | |||
There was a problem hiding this comment.
self.trainer.avg_loss was raising AttributeError
Codecov Report
@@ Coverage Diff @@
## master #3544 +/- ##
==========================================
+ Coverage 76.84% 77.81% +0.97%
==========================================
Files 100 100
Lines 17064 17064
==========================================
+ Hits 13112 13279 +167
+ Misses 3952 3785 -167
Continue to review full report at Codecov.
|
| from utils import SummarizationDataset | ||
|
|
||
|
|
||
| try: |
There was a problem hiding this comment.
to support being called from from pwd or examples/
sshleifer
changed the title
| @@ -0,0 +1,33 @@ | |||
| # Script for verifying that run_bart_sum can be invoked from its directory | |||
There was a problem hiding this comment.
Runs same logic as run_train.sh on a tiny dataset
|
LGTM from a superficial glance |
| try: | ||
| from .utils import SummarizationDataset | ||
| except ImportError: | ||
| from utils import SummarizationDataset |
There was a problem hiding this comment.
I asked @aaugustin if there's an easy way to not resort to this kind of hack
|
Planning on merging this April 15 at 7pm EST barring objections. |
examples/requirements.txt