Description
New Operator
PartitionedCall
Describe the operator
This operator is needed for any ConvNeXt Keras implementation conversion to ONNX
Do you know this operator be constructed using existing ONNX operators?
Not sure.
Is this operator used by any model currently? Which one?
ConvNeXt Keras implementation all sizes (Large, XL, etc)
Are you willing to contribute it? (Y/N)
N
Notes
When converting the model to ONNX, I get the following errors:
Tensorflow op [sequential/convnext_xlarge/convnext_xlarge_stage_0_block_0_depthwise_conv/PartitionedCall: PartitionedCall] is not supported
Tensorflow op [sequential/convnext_xlarge/convnext_xlarge_stage_0_block_1_depthwise_conv/PartitionedCall: PartitionedCall] is not supported
Tensorflow op [sequential/convnext_xlarge/convnext_xlarge_stage_0_block_2_depthwise_conv/PartitionedCall: PartitionedCall] is not supported
Tensorflow op [sequential/convnext_xlarge/convnext_xlarge_stage_1_block_0_depthwise_conv/PartitionedCall: PartitionedCall] is not supported
Tensorflow op [sequential/convnext_xlarge/convnext_xlarge_stage_1_block_1_depthwise_conv/PartitionedCall: PartitionedCall] is not supported
Tensorflow op [sequential/convnext_xlarge/convnext_xlarge_stage_1_block_2_depthwise_conv/PartitionedCall: PartitionedCall] is not supported
Tensorflow op [sequential/convnext_xlarge/convnext_xlarge_stage_2_block_0_depthwise_conv/PartitionedCall: PartitionedCall] is not supported
Tensorflow op [sequential/convnext_xlarge/convnext_xlarge_stage_2_block_1_depthwise_conv/PartitionedCall: PartitionedCall] is not supported
Tensorflow op [sequential/convnext_xlarge/convnext_xlarge_stage_2_block_2_depthwise_conv/PartitionedCall: PartitionedCall] is not supported
Tensorflow op [sequential/convnext_xlarge/convnext_xlarge_stage_2_block_3_depthwise_conv/PartitionedCall: PartitionedCall] is not supported
Tensorflow op [sequential/convnext_xlarge/convnext_xlarge_stage_2_block_4_depthwise_conv/PartitionedCall: PartitionedCall] is not supported
Tensorflow op [sequential/convnext_xlarge/convnext_xlarge_stage_2_block_5_depthwise_conv/PartitionedCall: PartitionedCall] is not supported
Tensorflow op [sequential/convnext_xlarge/convnext_xlarge_stage_2_block_6_depthwise_conv/PartitionedCall: PartitionedCall] is not supported
Tensorflow op [sequential/convnext_xlarge/convnext_xlarge_stage_2_block_7_depthwise_conv/PartitionedCall: PartitionedCall] is not supported
Tensorflow op [sequential/convnext_xlarge/convnext_xlarge_stage_2_block_8_depthwise_conv/PartitionedCall: PartitionedCall] is not supported
Tensorflow op [sequential/convnext_xlarge/convnext_xlarge_stage_2_block_9_depthwise_conv/PartitionedCall: PartitionedCall] is not supported
Tensorflow op [sequential/convnext_xlarge/convnext_xlarge_stage_2_block_10_depthwise_conv/PartitionedCall: PartitionedCall] is not supported
Tensorflow op [sequential/convnext_xlarge/convnext_xlarge_stage_2_block_11_depthwise_conv/PartitionedCall: PartitionedCall] is not supported
Tensorflow op [sequential/convnext_xlarge/convnext_xlarge_stage_2_block_12_depthwise_conv/PartitionedCall: PartitionedCall] is not supported
Tensorflow op [sequential/convnext_xlarge/convnext_xlarge_stage_2_block_13_depthwise_conv/PartitionedCall: PartitionedCall] is not supported
Tensorflow op [sequential/convnext_xlarge/convnext_xlarge_stage_2_block_14_depthwise_conv/PartitionedCall: PartitionedCall] is not supported
Tensorflow op [sequential/convnext_xlarge/convnext_xlarge_stage_2_block_15_depthwise_conv/PartitionedCall: PartitionedCall] is not supported
Tensorflow op [sequential/convnext_xlarge/convnext_xlarge_stage_2_block_16_depthwise_conv/PartitionedCall: PartitionedCall] is not supported
Tensorflow op [sequential/convnext_xlarge/convnext_xlarge_stage_2_block_17_depthwise_conv/PartitionedCall: PartitionedCall] is not supported
Tensorflow op [sequential/convnext_xlarge/convnext_xlarge_stage_2_block_18_depthwise_conv/PartitionedCall: PartitionedCall] is not supported
Tensorflow op [sequential/convnext_xlarge/convnext_xlarge_stage_2_block_19_depthwise_conv/PartitionedCall: PartitionedCall] is not supported
Tensorflow op [sequential/convnext_xlarge/convnext_xlarge_stage_2_block_20_depthwise_conv/PartitionedCall: PartitionedCall] is not supported
Tensorflow op [sequential/convnext_xlarge/convnext_xlarge_stage_2_block_21_depthwise_conv/PartitionedCall: PartitionedCall] is not supported
Tensorflow op [sequential/convnext_xlarge/convnext_xlarge_stage_2_block_22_depthwise_conv/PartitionedCall: PartitionedCall] is not supported
Tensorflow op [sequential/convnext_xlarge/convnext_xlarge_stage_2_block_23_depthwise_conv/PartitionedCall: PartitionedCall] is not supported
Tensorflow op [sequential/convnext_xlarge/convnext_xlarge_stage_2_block_24_depthwise_conv/PartitionedCall: PartitionedCall] is not supported
Tensorflow op [sequential/convnext_xlarge/convnext_xlarge_stage_2_block_25_depthwise_conv/PartitionedCall: PartitionedCall] is not supported
Tensorflow op [sequential/convnext_xlarge/convnext_xlarge_stage_2_block_26_depthwise_conv/PartitionedCall: PartitionedCall] is not supported
Tensorflow op [sequential/convnext_xlarge/convnext_xlarge_stage_3_block_0_depthwise_conv/PartitionedCall: PartitionedCall] is not supported
Tensorflow op [sequential/convnext_xlarge/convnext_xlarge_stage_3_block_1_depthwise_conv/PartitionedCall: PartitionedCall] is not supported
Tensorflow op [sequential/convnext_xlarge/convnext_xlarge_stage_3_block_2_depthwise_conv/PartitionedCall: PartitionedCall] is not supported
Unsupported ops: Counter({'PartitionedCall': 36})
This can be easily recreated with the following code:
import tensorflow as tf
import tf2onnx
pretrained_model= tf.keras.applications.ConvNeXtXLarge(include_top=False,
input_shape=(112,112,3),
pooling='avg',classes=3,
weights='imagenet')
onnx_model, _ = tf2onnx.convert.from_keras(pretrained_model)
I receive the same errors when using the command line interface with tf2onnx and using a saved ConvNeXt model.
Also, I have seen similar issues such as in #1864 requesting this op but that they have not been updated. Was a solution or workaround already created for this?
Thanks for any help.