From 4d303d1e5f9d6c26ad406f13783f2733a1c23465 Mon Sep 17 00:00:00 2001
From: dcarron <daniel.carron@idiap.ch>
Date: Tue, 14 May 2024 11:06:07 +0200
Subject: [PATCH] [mednet] Fixes after rebase
---
.../classification/engine/saliency/generator.py | 6 +++---
.../libs/classification/engine/saliency/viewer.py | 2 +-
.../libs/classification/scripts/experiment.py | 2 --
src/mednet/libs/classification/scripts/train.py | 10 ++--------
src/mednet/libs/common/models/loss_weights.py | 14 +++++++-------
src/mednet/libs/common/models/normalizer.py | 2 +-
6 files changed, 14 insertions(+), 22 deletions(-)
diff --git a/src/mednet/libs/classification/engine/saliency/generator.py b/src/mednet/libs/classification/engine/saliency/generator.py
index 28a11674..2b9badf1 100644
--- a/src/mednet/libs/classification/engine/saliency/generator.py
+++ b/src/mednet/libs/classification/engine/saliency/generator.py
@@ -212,14 +212,14 @@ def run(
for sample in tqdm.tqdm(v, desc="samples", leave=False, disable=None):
name = sample[1]["name"][0]
- label = sample[1]["label"].item()
+ target = sample[1]["target"].item()
image = sample[0].to(
device=device,
non_blocking=torch.cuda.is_available(),
)
- # in binary classification systems, negative labels may be skipped
- if positive_only and (model.num_classes == 1) and (label == 0):
+ # in binary classification systems, negative targets may be skipped
+ if positive_only and (model.num_classes == 1) and (target == 0):
continue
# chooses target outputs to generate saliency maps for
diff --git a/src/mednet/libs/classification/engine/saliency/viewer.py b/src/mednet/libs/classification/engine/saliency/viewer.py
index 588ba70c..063ce737 100644
--- a/src/mednet/libs/classification/engine/saliency/viewer.py
+++ b/src/mednet/libs/classification/engine/saliency/viewer.py
@@ -233,7 +233,7 @@ def run(
# WARNING: following code assumes a batch size of 1. Will break if
# not the case.
name = str(sample[1]["name"][0])
- label = int(sample[1]["label"].item())
+ label = int(sample[1]["target"].item())
data = sample[0][0]
if label != target_label:
diff --git a/src/mednet/libs/classification/scripts/experiment.py b/src/mednet/libs/classification/scripts/experiment.py
index 3118b14b..814f3b76 100644
--- a/src/mednet/libs/classification/scripts/experiment.py
+++ b/src/mednet/libs/classification/scripts/experiment.py
@@ -48,7 +48,6 @@ def experiment(
seed,
parallel,
monitoring_interval,
- balance_classes,
augmentations,
**_,
): # numpydoc ignore=PR01
@@ -89,7 +88,6 @@ def experiment(
seed=seed,
parallel=parallel,
monitoring_interval=monitoring_interval,
- balance_classes=balance_classes,
augmentations=augmentations,
)
train_stop_timestamp = datetime.now()
diff --git a/src/mednet/libs/classification/scripts/train.py b/src/mednet/libs/classification/scripts/train.py
index 6e1e28e8..85a570d2 100644
--- a/src/mednet/libs/classification/scripts/train.py
+++ b/src/mednet/libs/classification/scripts/train.py
@@ -86,14 +86,8 @@ def train(
# of class examples available in the training set. Also affects the
# validation loss if a validation set is available on the DataModule.
if balance_classes:
- logger.info("Applying DataModule train sampler balancing...")
- datamodule.balance_sampler_by_class = True
- # logger.info("Applying train/valid loss balancing...")
- # model.balance_losses_by_class(datamodule)
- else:
- logger.info(
- "Skipping sample class/dataset ownership balancing on user request",
- )
+ logger.info("Applying train/valid loss balancing...")
+ model.balance_losses(datamodule)
checkpoint_file = get_checkpoint_file(output_folder)
load_checkpoint(checkpoint_file, datamodule, model)
diff --git a/src/mednet/libs/common/models/loss_weights.py b/src/mednet/libs/common/models/loss_weights.py
index b3c2edec..596c9e7b 100644
--- a/src/mednet/libs/common/models/loss_weights.py
+++ b/src/mednet/libs/common/models/loss_weights.py
@@ -129,13 +129,13 @@ def get_positive_weights(
number of negative and positive samples (scalar). The weight can be used
to adjust minimisation criteria to in cases there is a huge data imbalance.
- It returns a vector with weights (inverse counts) for each label.
+ It returns a vector with weights (inverse counts) for each target.
Parameters
----------
dataloader
A DataLoader from which to compute the positive weights. Entries must
- be a dictionary which must contain a ``label`` key.
+ be a dictionary which must contain a ``target`` key.
Returns
-------
@@ -147,9 +147,9 @@ def get_positive_weights(
targets = defaultdict(list)
for batch in dataloader:
- for class_idx, class_targets in enumerate(batch[1]["label"]):
+ for class_idx, class_targets in enumerate(batch[1]["target"]):
# Targets are either a single tensor (binary case) or a list of tensors (multilabel)
- if isinstance(batch[1]["label"], list):
+ if isinstance(batch[1]["target"], list):
targets[class_idx].extend(tensor_to_list(class_targets))
else:
targets[0].extend(tensor_to_list(class_targets))
@@ -161,17 +161,17 @@ def get_positive_weights(
targets_tensor = torch.tensor(targets_list)
if targets_tensor.shape[0] == 1:
- logger.info("Computing positive weights assuming binary labels.")
+ logger.info("Computing positive weights assuming binary targets.")
positive_weights = compute_binary_weights(targets_tensor)
else:
if is_multicalss_exclusive(targets_tensor):
logger.info(
- "Computing positive weights assuming multiclass, exclusive labels."
+ "Computing positive weights assuming multiclass, exclusive targets."
)
positive_weights = compute_multiclass_weights(targets_tensor)
else:
logger.info(
- "Computing positive weights assuming multiclass, non-exclusive labels."
+ "Computing positive weights assuming multiclass, non-exclusive targets."
)
positive_weights = compute_non_exclusive_multiclass_weights(targets_tensor)
diff --git a/src/mednet/libs/common/models/normalizer.py b/src/mednet/libs/common/models/normalizer.py
index 63a5f10c..603de9c1 100644
--- a/src/mednet/libs/common/models/normalizer.py
+++ b/src/mednet/libs/common/models/normalizer.py
@@ -38,7 +38,7 @@ def make_z_normalizer(
# Ensure targets are ints
try:
- target = batch[1]["label"][0].item()
+ target = batch[1]["target"][0].item()
if not isinstance(target, int):
logger.info(
"Targets are not Integer type, skipping z-normalization."
--
GitLab