how to get what type of activation is used?
Question:
I have found that model.layers[index].output prints the info I need. But, I couldn’t get what activation function was used by looking at this output:
Tensor("dense_11_1/clip_by_value:0", shape=(?, 256), dtype=float32)
Usually, it is like Tensor("block5_conv3_1/Relu:0", shape=(?, ?, ?, 512), dtype=float32) and I can see Relu was used in that layer.
How to determine the activation function of the above-look output?
Thanks.
Answers:
It’s easier than you think, layers with activations will expose them with the activation attribute, so you can do:
model.layers[index].activation
And should give you information about the activation used by that layer.
I have found that model.layers[index].output prints the info I need. But, I couldn’t get what activation function was used by looking at this output:
Tensor("dense_11_1/clip_by_value:0", shape=(?, 256), dtype=float32)
Usually, it is like Tensor("block5_conv3_1/Relu:0", shape=(?, ?, ?, 512), dtype=float32) and I can see Relu was used in that layer.
How to determine the activation function of the above-look output?
Thanks.
It’s easier than you think, layers with activations will expose them with the activation attribute, so you can do:
model.layers[index].activation
And should give you information about the activation used by that layer.