47 lines
1.1 KiB
Python
47 lines
1.1 KiB
Python
import os
|
|
os.environ["TF_CPP_MIN_LOG_LEVEL"] = "3"
|
|
|
|
which_scalar = "Test Error"
|
|
|
|
from tensorboard.backend.event_processing import event_accumulator
|
|
import numpy as np
|
|
import glob
|
|
|
|
log_paths: str = "Log*"
|
|
log_paths_list = glob.glob(log_paths)
|
|
assert len(log_paths_list) > 0
|
|
|
|
for path in log_paths_list:
|
|
print(path)
|
|
temp = path.split("_")
|
|
if len(temp) == 2:
|
|
parameter:str | None = temp[-1]
|
|
else:
|
|
parameter = None
|
|
|
|
# ----------------------
|
|
temp = glob.glob(path)
|
|
assert len(temp) == 1
|
|
|
|
acc = event_accumulator.EventAccumulator(path)
|
|
acc.Reload()
|
|
|
|
# Check if the requested scalar exists
|
|
available_scalar = acc.Tags()["scalars"]
|
|
# available_histograms = acc.Tags()["histograms"]
|
|
available_scalar.index(which_scalar)
|
|
|
|
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][1]
|
|
np_temp[id, 1] = te[id][2]
|
|
print(np_temp)
|
|
|
|
if parameter is not None:
|
|
np.save(f"result_{parameter}.npy", np_temp)
|
|
else:
|
|
np.save(f"result.npy", np_temp)
|
|
|