parser.add_argument("--nb_gpts", type=int, default=5)
-parser.add_argument("--max_fail_to_validate", type=int, default=2)
+parser.add_argument("--max_fail_to_validate", type=int, default=3)
-parser.add_argument("--accuracy_to_make_c_quizzes", type=float, default=0.98)
+parser.add_argument("--accuracy_to_make_c_quizzes", type=float, default=0.95)
parser.add_argument("--proba_understands", type=float, default=0.95)
-parser.add_argument("--proba_not_understands", type=float, default=0.5)
+parser.add_argument("--proba_not_understands", type=float, default=0.1)
parser.add_argument("--temperature_hot", type=float, default=1.5)
parser.add_argument("--temperature_cold", type=float, default=1)
-parser.add_argument("--prompt_noise", type=float, default=0.0)
+parser.add_argument("--prompt_noise", type=float, default=0.05)
parser.add_argument("--nb_averaging_rounds", type=int, default=3)
parser.add_argument(
"--grids_world_tasks",
type=str,
- default=None,
- help="A comma-separated subset of: " + grids_tasks + ", or None for all.",
+ default="replace_color,translate,grow,frame",
+ help="A comma-separated subset of: " + grids_tasks + ".",
)
parser.add_argument(
"--grids_science_tasks",
type=str,
default=None,
- help="A comma-separated subset of: " + grids_tasks + ", or None for all.",
+ help="A comma-separated subset of: " + grids_tasks + ", or None.",
)
######################################################################
"model": "37M",
"batch_size": 25,
"inference_batch_size": 50,
- "nb_train_samples": 100000,
- "nb_test_samples": 10000,
+ "nb_train_samples": 40000,
+ "nb_test_samples": 1000,
}
for k, v in default_args.items():
# pass
+warnings.warn("*********** novel procedure!!! **********", RuntimeWarning)
+
c_quizzes_procedure = [
+ # (("f_B", "f_A", "A", "B"), (1, 0, 0, 0), model_transformer_hot),
+ # (("f_B", "f_A", "A", "B"), (0, 1, 1, 1), model_transformer_cold),
+ # (("A", "f_A", "B", "f_B"), (0, 0, 0, 1), model_transformer_cold),
(("f_B", "f_A", "A", "B"), (1, 0, 0, 0), model_transformer_hot),
- (("f_B", "f_A", "A", "B"), (0, 1, 1, 1), model_transformer_cold),
- (("A", "f_A", "B", "f_B"), (0, 0, 0, 1), model_transformer_cold),
- (("f_A", "A", "f_B", "B"), (0, 0, 0, 1), model_transformer_cold),
+ (("f_B", "f_A", "A", "B"), (0, 1, 1, 0), model_transformer_cold),
+ (("A", "f_A", "B", "f_B"), (0, 0, 1, 1), model_transformer_cold),
+ # (("f_A", "A", "f_B", "B"), (0, 0, 0, 1), model_transformer_cold),
# (("f_B", "f_A", "A", "B"), (0, 0, 1, 1), model_transformer_cold),
# (("A", "f_A", "B", "f_B"), (0, 0, 0, 1), model_transformer_cold),
]
),
)
+ print(model)
+ exit(0)
+
+
######################################################################
current_epoch = 0