40 lines
1.2 KiB
Python
40 lines
1.2 KiB
Python
|
from glob import glob
|
||
|
|
|
||
|
|
import tensorboard_reducer as tbr
|
||
|
|
|
||
|
|
from policy import policies
|
||
|
|
|
||
|
|
strict_steps=False
|
||
|
|
overwrite = True
|
||
|
|
reduce_ops = ("mean", "min", "max")
|
||
|
|
|
||
|
|
for policy_name in policies:
|
||
|
|
print(f"Calculating {policy_name}")
|
||
|
|
|
||
|
|
# Define glob pattern to find the directories
|
||
|
|
input_event_dirs = sorted(glob(f"tnsrbrd/sb3/{policy_name}_[0-9]"))
|
||
|
|
print(f"Found {input_event_dirs}")
|
||
|
|
|
||
|
|
# Define output dir
|
||
|
|
events_output_dir = f"tnsrbrd/results/{policy_name}"
|
||
|
|
|
||
|
|
# Load TB events
|
||
|
|
events_dict = tbr.load_tb_events(input_event_dirs, strict_steps=False, verbose=True)
|
||
|
|
|
||
|
|
# Get number of steps and other params
|
||
|
|
n_scalars = len(events_dict)
|
||
|
|
n_steps, n_events = list(events_dict.values())[0].shape
|
||
|
|
|
||
|
|
print(f"Loaded {n_events} TensorBoard runs with {n_scalars} scalars and {n_steps} steps each")
|
||
|
|
print(", ".join(events_dict))
|
||
|
|
|
||
|
|
# Reduce events
|
||
|
|
reduced_events = tbr.reduce_events(events_dict, reduce_ops, verbose=True)
|
||
|
|
|
||
|
|
for op in reduce_ops:
|
||
|
|
print(f"Writing '{op}' reduction to '{events_output_dir}-{op}'")
|
||
|
|
|
||
|
|
tbr.write_tb_events(reduced_events, events_output_dir, overwrite)
|
||
|
|
|
||
|
|
print(f"Reduction of {policy_name} done.")
|