Closed aaronrmm closed 5 years ago
I guess they are the same problems since the training data will also be loaded at the inference stage for the accompaniment model. You need to have >5G RAM for loading the entire training data. One solution is to load only part of the training data if you don't have enough RAM. You can modify the following function to achieve this (by setting the length of the first axis to a smaller number).
def load_data_from_npz(filename):
"""Load and return the training data from a npz file (sparse format)."""
with np.load(filename) as f:
data = np.zeros(f['shape'], np.bool_)
data[[x for x in f['nonzero']]] = True
return data
I'm getting a memory bus error on trying to load in train_x_lpd_5_phr.npz, even when attempting to load a pretrained model.