Ray is a unified framework for scaling AI and Python applications. Ray consists of a core distributed runtime and a set of AI Libraries for accelerating ML workloads.
The value for use_critic is incorrectly assigned to lr_schedule in a3c.py. This silently passes for use_critic=False with the value for self.use_critic not being set correctly and raises an error for use_critic=True.
E lr_schedule, outside_value=lr_schedule[-1][-1], framework=None
E TypeError: 'bool' object is not subscriptable
Versions / Dependencies
Present in master branch
Reproduction script
import ray
from ray.rllib.algorithms.a3c import A3CConfig
ray.init()
a3c = (
A3CConfig()
.environment('CartPole-v1')
.training(use_critic=True)
.build()
)
What happened + What you expected to happen
The value for
use_critic
is incorrectly assigned tolr_schedule
ina3c.py
. This silently passes foruse_critic=False
with the value forself.use_critic
not being set correctly and raises an error foruse_critic=True
.Versions / Dependencies
Present in
master
branchReproduction script
Issue Severity
High: It blocks me from completing my task.