"--task",
type=str,
default="twotargets",
- help="byheart, learnop, guessop, twotargets, addition, picoclvr, mnist, maze, snake, stack, expr, rpl",
+ help="byheart, learnop, guessop, twotargets, addition, picoclvr, mnist, maze, snake, stack, expr, rpl, grid",
)
parser.add_argument("--log_filename", type=str, default="train.log", help=" ")
##############################
# rpl options
-parser.add_argument("--rpl_nb_starting_values", type=int, default=5)
+parser.add_argument("--rpl_nb_starting_values", type=int, default=3)
parser.add_argument("--rpl_max_input", type=int, default=9)
-parser.add_argument("--rpl_prog_len", type=int, default=10)
+parser.add_argument("--rpl_prog_len", type=int, default=8)
-parser.add_argument("--rpl_nb_runs", type=int, default=8)
+parser.add_argument("--rpl_nb_runs", type=int, default=5)
parser.add_argument("--rpl_no_prog", action="store_true", default=False)
+##############################
+# grid options
+
+parser.add_argument("--grid_size", type=int, default=6)
+
##############################
# picoclvr options
"nb_test_samples": 10000,
},
"rpl": {
- "model": "352M",
+ "model": "122M",
"nb_epochs": 50,
- "batch_size": 10,
- "nb_train_samples": 2500000,
+ "batch_size": 5,
+ "nb_train_samples": 1000000,
"nb_test_samples": 10000,
},
"world": {
"nb_train_samples": 25000,
"nb_test_samples": 1000,
},
+ "grid": {
+ "model": "37M",
+ "nb_epochs": 25,
+ "batch_size": 25,
+ "nb_train_samples": 250000,
+ "nb_test_samples": 10000,
+ },
}
if args.task in default_task_args:
device=device,
)
+elif args.task == "grid":
+ task = tasks.Grid(
+ nb_train_samples=args.nb_train_samples,
+ nb_test_samples=args.nb_test_samples,
+ batch_size=args.batch_size,
+ size=args.grid_size,
+ logger=log_string,
+ device=device,
+ )
+
elif args.task == "world":
task = tasks.World(
nb_train_samples=args.nb_train_samples,