Get weights from a parallel model and save as a single GPU model in Keras

Here’s the table of contents:

1. Load into a multi-gpu model and save the weights for a single-gpu model

Load into a multi-gpu model and save the weights for a single-gpu model

    import numpy as np
    weights = np.load("./parallel_model.npy", allow_pickle=True)
    parallel_model.set_weights(weights)
    print (parallel_model.layers[-2])
    single_gpu_model = parallel_model.layers[-2]
    np.save("./single_model", single_gpu_model.get_weights())

See!