diff --git a/research/audioset/yamnet/README.md b/research/audioset/yamnet/README.md
index 4f3caddfd0f..e57b84339d2 100644
--- a/research/audioset/yamnet/README.md
+++ b/research/audioset/yamnet/README.md
@@ -15,11 +15,13 @@ YAMNet depends on the following Python packages:
 * [`numpy`](http://www.numpy.org/)
 * [`resampy`](http://resampy.readthedocs.io/en/latest/)
 * [`tensorflow`](http://www.tensorflow.org/)
+* [`tf_keras`](https://github.com/keras-team/tf-keras)
 * [`pysoundfile`](https://pysoundfile.readthedocs.io/)
 
 These are all easily installable via, e.g., `pip install numpy` (as in the
 example command sequence below). Any reasonably recent version of these
-packages should work.
+packages should work. Note that YAMNet currently relies on Keras 2 and is
+incompatible with Keras 3 (the default as of TF 2.16).
 
 YAMNet also requires downloading the following data file:
 
diff --git a/research/audioset/yamnet/yamnet.py b/research/audioset/yamnet/yamnet.py
index cac7f87d99e..06a246b043b 100644
--- a/research/audioset/yamnet/yamnet.py
+++ b/research/audioset/yamnet/yamnet.py
@@ -19,7 +19,7 @@
 
 import numpy as np
 import tensorflow as tf
-from tensorflow.keras import Model, layers
+from tf_keras import Model, layers
 
 import features as features_lib