Skip to content

migrate metric learning example to keras 3 #2206

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 2 commits into
base: master
Choose a base branch
from
Open

migrate metric learning example to keras 3 #2206

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
ba408a9
migrate metric learning example to keras 3
Shi-pra-19 Oct 28, 2025
87c5a8f
Fix duplicate assignment in no_match_queries
Shi-pra-19 Oct 28, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
9 changes: 5 additions & 4 deletions examples/vision/metric_learning_tf_similarity.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -54,7 +54,7 @@
import numpy as np

import tensorflow as tf
from tensorflow import keras
import keras

import tensorflow_similarity as tfsim

Expand Down Expand Up @@ -216,13 +216,14 @@
val_steps = 50

# init similarity loss
loss = tfsim.losses.MultiSimilarityLoss()
loss = tfsim.losses.MultiSimilarityLoss(reduction='sum_over_batch_size')

# compiling and training
model.compile(
optimizer=keras.optimizers.Adam(learning_rate),
loss=loss,
steps_per_execution=10,
run_eagerly=True,
Copy link
Contributor

@gemini-code-assist gemini-code-assist bot Oct 28, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

high

Enabling run_eagerly=True can significantly degrade training performance as it disables graph optimizations. While it can be useful for debugging, it's generally recommended to make the model and training loop compatible with graph execution (i.e., by setting run_eagerly=False or removing the argument). Was this added to work around a specific incompatibility with Keras 3's graph mode? If so, it would be better to address the underlying issue to ensure the example runs efficiently.

)
history = model.fit(
train_ds, epochs=epochs, validation_data=val_ds, validation_steps=val_steps
Expand Down Expand Up @@ -321,7 +322,7 @@
for idx in np.argsort(y_display):
tfsim.visualization.viz_neigbors_imgs(
x_display[idx],
y_display[idx],
y_display[idx].numpy(),
nns[idx],
class_mapping=class_mapping,
fig_size=(16, 2),
Expand Down Expand Up @@ -394,7 +395,7 @@
"""

idx_no_match = np.where(np.array(matches) == 10)
no_match_queries = x_confusion[idx_no_match]
no_match_queries = keras.ops.take(x_confusion, keras.ops.cast(idx_no_match[0], dtype="int32"), axis=0)
if len(no_match_queries):
plt.imshow(no_match_queries[0])
else:
Expand Down