google-deepmind / google-deepmind/deepmind-research
Long rollout feature
Open
- Dominant language
- Jupyter Notebook
- Stars
- 15.2k
- Forks
- 2.9k
- PR merge metrics
- No merged PRs in 30d
Description
I am trying to create a long rollout with 1500 steps on a trained model with _cylinder_flow_. I found a variable called `num_steps` in the _cfd_eval.py_ file and changed it to 1500 but it fails at this line `scalar_data, traj_data = sess.run([scalar_op, traj_ops])` of _run_model.py_. I was wondering what changes were made to the code to create the long rollouts as shown with the flag example. Also an argument to specify the rollout_steps would be very useful as a new feature.
Contributor guide
Assessment
This issue has not been assessed yet.