don't fail if optim name doesn't exist

src/neptune_fastai/impl/__init__.py

@@ -154,7 +154,9 @@ def _batch_size(self) -> int:
 
  @property
  def _optimizer_name(self) -> Optional[str]:
- return self.opt_func.__name__
+ optim_name = getattr(self.opt_func, "__name__", "NA")
+ warnings.warn("NeptuneCallback: Couldn't retrieve the optimizer name, so it will not be logged.")
+ return optim_name
 
  @property
  def _device(self) -> str:

tests/neptune_fastai/test_e2e.py

@@ -13,6 +13,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 #
+from functools import partial
 from itertools import islice
 from pathlib import Path
 
@@ -24,6 +25,7 @@
  untar_data,
 )
 from fastai.callback.all import SaveModelCallback
+from fastai.optimizer import Adam
 from fastai.tabular.all import (
  Categorify,
  FillMissing,
@@ -71,12 +73,15 @@ def test_vision_classification_with_handler(self):
  device=torch.device("cpu"),
  )
 
+ opt_func = partial(Adam, lr=3e-3, wd=0.01)
+
  learn = cnn_learner(
  dls,
  squeezenet1_0,
  metrics=error_rate,
  cbs=[NeptuneCallback(run, "experiment")],
  pretrained=False,
+ opt_func=opt_func,
  )
 
  learn.fit(1)
@@ -91,6 +96,7 @@ def test_vision_classification_with_handler(self):
  exp_config = run["experiment/config"].fetch()
  assert exp_config["batch_size"] == 64
  assert exp_config["criterion"] == "CrossEntropyLoss()"
+ assert exp_config["optimizer"]["name"] == "NA"
  assert exp_config["input_shape"] == {"x": "[3, 224, 224]", "y": 1}
 
  # and