fails to convert Keras models to ONNX when the model contains multiple `BatchNormalization` layers followed by a custom layer that performs element-wise multiplication in `channels_first` data format

[Inquisitive-ME]

Describe the bug
tf2onnx fails to convert Keras models to ONNX when the model contains multiple BatchNormalization layers followed by a custom layer that performs element-wise multiplication in channels_first data format.

System information

• OS Platform and Distribution (e.g., Linux Ubuntu 18.04*): Linux Ubuntu 22.04
• TensorFlow Version: 2.20.0
• Python version: 3.12.9
• ONNX version (if applicable, e.g. 1.11*): 1.16.1
• ONNXRuntime version (if applicable, e.g. 1.11*):

To Reproduce
Run the script below it will fail with:
"Process finished with exit code 134 (interrupted by signal 6:SIGABRT)"
tf2onnx_bug_report.py

Additional context
Issue is from trying to use Mobilenetv4Hybrid models in Channels First

[jaytiwarihub]

this issue resolved ?

[jaytiwarihub]

Root cause seems related to implicit broadcasting in NCHW.

When replacing:
x * scale

with:
scale = tf.broadcast_to(scale, tf.shape(x))
x = tf.multiply(x, scale)

conversion succeeds for both NHWC and NCHW.

This suggests tf2onnx has issues handling implicit broadcasting
for element-wise ops in channels_first format.