Skip to content

Fix DQN w RNN tutorial #3462

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

Merged
merged 4 commits into from
Jul 23, 2025
Merged

Fix DQN w RNN tutorial #3462

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
92e1abc
Fix DQN w RNN tutorial
svekars Jul 14, 2025
5e85c0a
Update intermediate_source/dqn_with_rnn_tutorial.py
svekars Jul 14, 2025
f4d1b1c
Update
svekars Jul 17, 2025
0ce4e16
bump torchrl and tensordict req (#3474)
vmoens Jul 22, 2025
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
4 changes: 2 additions & 2 deletions .ci/docker/requirements.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -29,8 +29,8 @@ tensorboard
jinja2==3.1.3
pytorch-lightning
torchx
torchrl==0.7.2
tensordict==0.7.2
torchrl==0.9.2
tensordict==0.9.1
# For ax_multiobjective_nas_tutorial.py
ax-platform>=0.4.0,<0.5.0
nbformat>=5.9.2
Expand Down
14 changes: 10 additions & 4 deletions intermediate_source/dqn_with_rnn_tutorial.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -342,7 +342,9 @@
# will return a new instance of the LSTM (with shared weights) that will
# assume that the input data is sequential in nature.
#
policy = Seq(feature, lstm.set_recurrent_mode(True), mlp, qval)
from torchrl.modules import set_recurrent_mode

policy = Seq(feature, lstm, mlp, qval)

######################################################################
# Because we still have a couple of uninitialized parameters we should
Expand Down Expand Up @@ -389,7 +391,10 @@
# For the sake of efficiency, we're only running a few thousands iterations
# here. In a real setting, the total number of frames should be set to 1M.
#
collector = SyncDataCollector(env, stoch_policy, frames_per_batch=50, total_frames=200, device=device)

collector = SyncDataCollector(
env, stoch_policy, frames_per_batch=50, total_frames=200, device=device
)
rb = TensorDictReplayBuffer(
storage=LazyMemmapStorage(20_000), batch_size=4, prefetch=10
)
Expand Down Expand Up @@ -422,7 +427,8 @@
rb.extend(data.unsqueeze(0).to_tensordict().cpu())
for _ in range(utd):
s = rb.sample().to(device, non_blocking=True)
loss_vals = loss_fn(s)
with set_recurrent_mode(True):
loss_vals = loss_fn(s)
loss_vals["loss"].backward()
optim.step()
optim.zero_grad()
Expand Down Expand Up @@ -464,5 +470,5 @@
#
# Further Reading
# ---------------
#
#
# - The TorchRL documentation can be found `here <https://pytorch.org/rl/>`_.
Loading