diff --git a/keras_hub/src/models/efficientnet/efficientnet_backbone.py b/keras_hub/src/models/efficientnet/efficientnet_backbone.py index 05e46967f8..99fbc68b5f 100644 --- a/keras_hub/src/models/efficientnet/efficientnet_backbone.py +++ b/keras_hub/src/models/efficientnet/efficientnet_backbone.py @@ -136,13 +136,17 @@ def __init__( ): num_stacks = len(stackwise_kernel_sizes) if "depth_coefficient" in kwargs: - stackwise_depth_coefficients = [ - kwargs.pop("depth_coefficient") - ] * num_stacks + depth_coefficient = kwargs.pop("depth_coefficient") + if not isinstance(depth_coefficient, (list, tuple)): + stackwise_depth_coefficients = [depth_coefficient] * num_stacks + else: + stackwise_depth_coefficients = depth_coefficient if "width_coefficient" in kwargs: - stackwise_width_coefficients = [ - kwargs.pop("width_coefficient") - ] * num_stacks + width_coefficient = kwargs.pop("width_coefficient") + if not isinstance(width_coefficient, (list, tuple)): + stackwise_width_coefficients = [width_coefficient] * num_stacks + else: + stackwise_width_coefficients = width_coefficient image_input = keras.layers.Input(shape=input_shape) diff --git a/keras_hub/src/models/xlm_roberta/xlm_roberta_presets.py b/keras_hub/src/models/xlm_roberta/xlm_roberta_presets.py index 20b6c7e150..9a4ed4dbd7 100644 --- a/keras_hub/src/models/xlm_roberta/xlm_roberta_presets.py +++ b/keras_hub/src/models/xlm_roberta/xlm_roberta_presets.py @@ -21,6 +21,6 @@ "params": 558837760, "path": "xlm_roberta", }, - "kaggle_handle": "kaggle://keras/xlm_roberta/keras/xlm_roberta_large_multi/2", + "kaggle_handle": "kaggle://keras/xlm_roberta/keras/xlm_roberta_large_multi/3", }, } diff --git a/keras_hub/src/utils/keras_utils.py b/keras_hub/src/utils/keras_utils.py index 6cc4dddc4b..f792bcf3a9 100644 --- a/keras_hub/src/utils/keras_utils.py +++ b/keras_hub/src/utils/keras_utils.py @@ -36,7 +36,9 @@ def print_msg(message, line_break=True): logging.info(message) +# Register twice for backwards compat. @keras.saving.register_keras_serializable(package="keras_hub") +@keras.saving.register_keras_serializable(package="keras_nlp") def gelu_approximate(x): return keras.activations.gelu(x, approximate=True)