Update.

[picoclvr.git] / main.py

diff --git a/main.py b/main.py
index 2339dcf..0f2cb61 100755 (executable)
--- a/main.py
+++ b/main.py
@@ -186,6 +186,8 @@ parser.add_argument("--greed_T", type=int, default=25)
 
 parser.add_argument("--greed_nb_walls", type=int, default=5)
 
+parser.add_argument("--greed_nb_coins", type=int, default=2)
+
 ######################################################################
 
 args = parser.parse_args()
@@ -625,6 +627,7 @@ elif args.task == "greed":
         width=args.greed_width,
         T=args.greed_T,
         nb_walls=args.greed_nb_walls,
+        nb_coins=args.greed_nb_coins,
         logger=log_string,
         device=device,
     )
@@ -700,8 +703,6 @@ if args.task == "expr" and args.expr_input_file is not None:
 
 ######################################################################
 
-nb_epochs = args.nb_epochs if args.nb_epochs > 0 else nb_epochs_default
-
 # Compute the entropy of the training tokens
 
 token_count = 0
@@ -770,7 +771,7 @@ log_string(f"learning_rate_schedule {learning_rate_schedule}")
 
 nb_samples_seen = 0
 
-if nb_epochs_finished >= nb_epochs:
+if nb_epochs_finished >= args.nb_epochs:
     task.produce_results(
         n_epoch=nb_epochs_finished,
         model=model,
@@ -781,7 +782,7 @@ if nb_epochs_finished >= nb_epochs:
 
 time_pred_result = None
 
-for n_epoch in range(nb_epochs_finished, nb_epochs):
+for n_epoch in range(nb_epochs_finished, args.nb_epochs):
     learning_rate = learning_rate_schedule[n_epoch]
 
     log_string(f"learning_rate {learning_rate}")