mirror of
https://github.com/vale981/ray
synced 2025-03-09 12:56:46 -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.
50 lines
1.6 KiB
Python
50 lines
1.6 KiB
Python
import numpy as np
|
|
from ray.rllib.evaluation import SampleBatch
|
|
|
|
from ray.rllib.utils.filter import MeanStdFilter
|
|
|
|
|
|
class _MockWorker:
|
|
def __init__(self, sample_count=10):
|
|
self._weights = np.array([-10, -10, -10, -10])
|
|
self._grad = np.array([1, 1, 1, 1])
|
|
self._sample_count = sample_count
|
|
self.obs_filter = MeanStdFilter(())
|
|
self.rew_filter = MeanStdFilter(())
|
|
self.filters = {
|
|
"obs_filter": self.obs_filter,
|
|
"rew_filter": self.rew_filter
|
|
}
|
|
|
|
def sample(self):
|
|
samples_dict = {"observations": [], "rewards": []}
|
|
for i in range(self._sample_count):
|
|
samples_dict["observations"].append(
|
|
self.obs_filter(np.random.randn()))
|
|
samples_dict["rewards"].append(self.rew_filter(np.random.randn()))
|
|
return SampleBatch(samples_dict)
|
|
|
|
def compute_gradients(self, samples):
|
|
return self._grad * samples.count, {"batch_count": samples.count}
|
|
|
|
def apply_gradients(self, grads):
|
|
self._weights += self._grad
|
|
|
|
def get_weights(self):
|
|
return self._weights
|
|
|
|
def set_weights(self, weights):
|
|
self._weights = weights
|
|
|
|
def get_filters(self, flush_after=False):
|
|
obs_filter = self.obs_filter.copy()
|
|
rew_filter = self.rew_filter.copy()
|
|
if flush_after:
|
|
self.obs_filter.clear_buffer(), self.rew_filter.clear_buffer()
|
|
|
|
return {"obs_filter": obs_filter, "rew_filter": rew_filter}
|
|
|
|
def sync_filters(self, new_filters):
|
|
assert all(k in new_filters for k in self.filters)
|
|
for k in self.filters:
|
|
self.filters[k].sync(new_filters[k])
|