classifer.train_loop
I'm testing out DenseNet with the default recipe file (from ketos.neural_networks.densenet import default_densenet_recipe) and upon running classifier.train_loop(n_epochs=30, verbose=True), I get a very length error message (full version attached)error.txt ending in:
ValueError: Negative dimension size caused by subtracting 2 from 1 for '{{node dense_net_arch_1/sequential_5/transition_block_4/average_pooling2d_4/AvgPool}} = AvgPoolT=DT_FLOAT, data_format="NHWC", ksize=[1, 2, 2, 1], padding="VALID", strides=[1, 2, 2, 1]' with input shapes: [128,1,1500,128].
If I switch out the DenseNet default recipe file for the default ResNet recipe I don't get any error message and it starts running. Is there something I've overlooked? Thanks very much, Shaula :)