Skip training if play_only is True. Fixes #58.

PiperOrigin-RevId: 728284290 Change-Id: I1ff15d308005512cd314b6ab145a7f09cfc0c0fc
google-deepmind · Feb 18, 2025 · 1fdd805 · 1fdd805
1 parent 6430d0d
commit 1fdd805
Showing 1 changed file with 2 additions and 0 deletions.
diff --git a/learning/train_jax_ppo.py b/learning/train_jax_ppo.py
@@ -163,6 +163,8 @@ def main(argv):
 
   if _NUM_TIMESTEPS.present:
     ppo_params.num_timesteps = _NUM_TIMESTEPS.value
+  if _PLAY_ONLY.present:
+    ppo_params.num_timesteps = 0
   if _NUM_EVALS.present:
     ppo_params.num_evals = _NUM_EVALS.value
   if _REWARD_SCALING.present: