Skip to content

Fix/dqn example #121

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
KaleabTessera wants to merge 2 commits into
base: main
Choose a base branch
from
Open

Fix/dqn example #121

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
157a52c
fix: Fixed experiment import.
KaleabTessera Mar 14, 2023
2ffc089
fix: Fixed epsilon scheduling due to actor count getting reset every …
KaleabTessera Mar 14, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
14 changes: 8 additions & 6 deletions examples/experiment.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -28,19 +28,21 @@ def run_loop(
params = agent.initial_params(next(rng))
learner_state = agent.initial_learner_state(params)

train_actor_state = agent.initial_actor_state()

print(f"Training agent for {train_episodes} episodes")
for episode in range(train_episodes):

# Prepare agent, environment and accumulator for a new episode.
timestep = environment.reset()
accumulator.push(timestep, None)
actor_state = agent.initial_actor_state()


while not timestep.last():

# Acting.
actor_output, actor_state = agent.actor_step(
params, timestep, actor_state, next(rng), evaluation=False)
actor_output, train_actor_state = agent.actor_step(
params, timestep, train_actor_state, next(rng), evaluation=False)

# Agent-environment interaction.
action = int(actor_output.actions)
Expand All @@ -57,13 +59,13 @@ def run_loop(
# Evaluation.
if not episode % evaluate_every:
returns = 0.
eval_actor_state = agent.initial_actor_state()
for _ in range(eval_episodes):
timestep = environment.reset()
actor_state = agent.initial_actor_state()

while not timestep.last():
actor_output, actor_state = agent.actor_step(
params, timestep, actor_state, next(rng), evaluation=True)
actor_output, eval_actor_state = agent.actor_step(
params, timestep, eval_actor_state, next(rng), evaluation=True)
timestep = environment.step(int(actor_output.actions))
returns += timestep.reward

Expand Down
2 changes: 1 addition & 1 deletion examples/simple_dqn.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -19,14 +19,14 @@
from absl import app
from absl import flags
from bsuite.environments import catch
from examples import experiment
import haiku as hk
from haiku import nets
import jax
import jax.numpy as jnp
import numpy as np
import optax
import rlax
from rlax.examples import experiment

Params = collections.namedtuple("Params", "online target")
ActorState = collections.namedtuple("ActorState", "count")
Expand Down