https://keras.io/api/layers/base_layer/#set_weights-method

1
2
3
4
5
6
7
8
9
10
f = h5py.File("encoder_weights_0.hdf5", "r")
print(f.filename, ":")
print(f['dense_1'])
print([key for key in f.keys()], "\n")
for key in f.keys():
    print(key,f[key])
    for k in f[key].keys():
        print(k,f[key][k])
        for l in f[key][k].keys():
            print(l, f[key][k][l])
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
<HDF5 group "/dense_1" (1 members)>
['dense_1', 'dense_2', 'dense_3', 'input_1']

dense_1 <HDF5 group "/dense_1" (1 members)>
dense_1 <HDF5 group "/dense_1/dense_1" (2 members)>
bias:0 <HDF5 dataset "bias:0": shape (64,), type "<f4">
kernel:0 <HDF5 dataset "kernel:0": shape (100, 64), type "<f4">
dense_2 <HDF5 group "/dense_2" (1 members)>
dense_2 <HDF5 group "/dense_2/dense_2" (2 members)>
bias:0 <HDF5 dataset "bias:0": shape (16,), type "<f4">
kernel:0 <HDF5 dataset "kernel:0": shape (64, 16), type "<f4">
dense_3 <HDF5 group "/dense_3" (1 members)>
dense_3 <HDF5 group "/dense_3/dense_3" (2 members)>
bias:0 <HDF5 dataset "bias:0": shape (8,), type "<f4">
kernel:0 <HDF5 dataset "kernel:0": shape (16, 8), type "<f4">
input_1 <HDF5 group "/input_1" (0 members)>

a Dense layer returns a list of two values– per-output weights and the bias value.
在