mirror of
https://github.com/vale981/ray
synced 2025-03-09 04:46:38 -04:00
60d4d5e1aa
* Remove all __future__ imports from RLlib. * Remove (object) again from tf_run_builder.py::TFRunBuilder. * Fix 2xLINT warnings. * Fix broken appo_policy import (must be appo_tf_policy) * Remove future imports from all other ray files (not just RLlib). * Remove future imports from all other ray files (not just RLlib). * Remove future import blocks that contain `unicode_literals` as well. Revert appo_tf_policy.py to appo_policy.py (belongs to another PR). * Add two empty lines before Schedule class. * Put back __future__ imports into determine_tests_to_run.py. Fails otherwise on a py2/print related error.
32 lines
902 B
Python
32 lines
902 B
Python
import numpy as np
|
|
|
|
|
|
def flatten(weights, start=0, stop=2):
|
|
"""This methods reshapes all values in a dictionary.
|
|
|
|
The indices from start to stop will be flattened into a single index.
|
|
|
|
Args:
|
|
weights: A dictionary mapping keys to numpy arrays.
|
|
start: The starting index.
|
|
stop: The ending index.
|
|
"""
|
|
for key, val in weights.items():
|
|
new_shape = val.shape[0:start] + (-1, ) + val.shape[stop:]
|
|
weights[key] = val.reshape(new_shape)
|
|
return weights
|
|
|
|
|
|
def concatenate(weights_list):
|
|
keys = weights_list[0].keys()
|
|
result = {}
|
|
for key in keys:
|
|
result[key] = np.concatenate([l[key] for l in weights_list])
|
|
return result
|
|
|
|
|
|
def shuffle(trajectory):
|
|
permutation = np.random.permutation(trajectory["actions"].shape[0])
|
|
for key, val in trajectory.items():
|
|
trajectory[key] = val[permutation]
|
|
return trajectory
|