32 lines
705 B
Python
32 lines
705 B
Python
|
import os
|
||
|
|
import glob
|
||
|
|
|
||
|
|
os.environ["TF_CPP_MIN_LOG_LEVEL"] = "3"
|
||
|
|
|
||
|
|
from tensorboard.backend.event_processing import event_accumulator # type: ignore
|
||
|
|
import numpy as np
|
||
|
|
|
||
|
|
|
||
|
|
def get_data(path: str = "log_cnn"):
|
||
|
|
acc = event_accumulator.EventAccumulator(path)
|
||
|
|
acc.Reload()
|
||
|
|
|
||
|
|
which_scalar = "Test Number Correct"
|
||
|
|
te = acc.Scalars(which_scalar)
|
||
|
|
|
||
|
|
np_temp = np.zeros((len(te), 2))
|
||
|
|
|
||
|
|
for id in range(0, len(te)):
|
||
|
|
np_temp[id, 0] = te[id].step
|
||
|
|
np_temp[id, 1] = te[id].value
|
||
|
|
|
||
|
|
print(np_temp[:, 1] / 100)
|
||
|
|
np_temp = np.nan_to_num(np_temp)
|
||
|
|
return np_temp
|
||
|
|
|
||
|
|
|
||
|
|
for path in glob.glob("log_*"):
|
||
|
|
print(path)
|
||
|
|
data = get_data(path)
|
||
|
|
np.save("data_" + path + ".npy", data)
|