Closed omerferhatt closed 3 years ago
Hello again, the main problem is
model.input.set_shape((1,) + model.input.shape[1:])
In MobileNetV2, constructed model can be compiled without editing. But in V3Small and V3Large, model input shape needs to be set before saving model. After setting batch shape, model compiled successfully.
Hi, It seems you found the solution. Are you still looking for Mobilenet v3 model implementation on TF2?
Hello, actually no. It's working now. But it would be great if docs will be updated for newcomers.
Can someone add Mobilenet v3 model implementation on TF2 because Keras pre-trained model not working.