Estimator should still use v1 optimizer.

PiperOrigin-RevId: 301308906

Estimator should still use v1 optimizer.
PiperOrigin-RevId: 301308906
28f6182f · Hongkun Yu · A. Unique TensorFlower · 69837027 · 28f6182f
Commit 28f6182f authored Mar 16, 2020 by Hongkun Yu Committed by A. Unique TensorFlower Mar 16, 2020
Show whitespace changes
Inline Side-by-side

Showing with 1 addition and 1 deletion

official/recommendation/neumf_model.py official/recommendation/neumf_model.py +1 -1

No files found.
--- a/official/recommendation/neumf_model.py
+++ b/official/recommendation/neumf_model.py
@@ -109,7 +109,7 @@ def neumf_model_fn(features, labels, mode, params):
    mlperf_helper.ncf_print(key=mlperf_helper.TAGS.OPT_HP_ADAM_EPSILON,
                            value=params["epsilon"])

-    optimizer = tf.keras.optimizers.Adam(
+    optimizer = tf.compat.v1.train.AdamOptimizer(
        learning_rate=params["learning_rate"],
        beta1=params["beta1"],
        beta2=params["beta2"],