parser.add_argument(
"--task",
type=str,
- default="sandbox",
- help="sandbox, picoclvr, mnist, maze, snake, stack, expr, rpl, world",
+ default="twotargets",
+ help="byheart, learnop, guessop, degradation, twotargets, addition, picoclvr, mnist, maze, snake, stack, expr, rpl, grid, qmlp",
)
parser.add_argument("--log_filename", type=str, default="train.log", help=" ")
########################################
-parser.add_argument("--nb_epochs", type=int, default=None)
+parser.add_argument("--nb_epochs", type=int, default=25)
parser.add_argument("--batch_size", type=int, default=None)
########################################
-parser.add_argument("--model", type=str, default="37M")
+parser.add_argument("--model", type=str, default=None)
parser.add_argument("--dim_model", type=int, default=None)
##############################
# 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)
##############################
-# sandbox options
+# grid options
-parser.add_argument("--sandbox_level", type=int, default=0)
-
-parser.add_argument("--sandbox_levels_nb_items", type=int, default=25)
-
-parser.add_argument("--sandbox_levels_len_source", type=int, default=6)
-
-parser.add_argument("--sandbox_levels_len_result", type=int, default=8)
+parser.add_argument("--grid_size", type=int, default=6)
##############################
# picoclvr options
##############################
# Maze options
-parser.add_argument("--maze_height", type=int, default=23)
+parser.add_argument("--maze_height", type=int, default=13)
-parser.add_argument("--maze_width", type=int, default=39)
+parser.add_argument("--maze_width", type=int, default=21)
-parser.add_argument("--maze_nb_walls", type=int, default=45)
+parser.add_argument("--maze_nb_walls", type=int, default=15)
##############################
# Snake options
-parser.add_argument("--snake_height", type=int, default=6)
+parser.add_argument("--snake_height", type=int, default=9)
-parser.add_argument("--snake_width", type=int, default=8)
+parser.add_argument("--snake_width", type=int, default=12)
parser.add_argument("--snake_nb_colors", type=int, default=5)
parser.add_argument("--expr_input_file", type=str, default=None)
##############################
-# World options
+# Misc
-parser.add_argument("--world_vqae_nb_epochs", type=int, default=25)
+parser.add_argument("--degradation_hard", action="store_true", default=False)
######################################################################
######################################################################
default_task_args = {
- "sandbox": {
- "nb_epochs": 50,
+ "addition": {
+ "model": "352M",
"batch_size": 25,
- "nb_train_samples": 100000,
+ "nb_train_samples": 250000,
"nb_test_samples": 10000,
},
- "picoclvr": {
- "nb_epochs": 25,
+ "byheart": {
+ "model": "37M",
+ "batch_size": 25,
+ "nb_train_samples": 50000,
+ "nb_test_samples": 10000,
+ },
+ "expr": {
+ "model": "352M",
+ "batch_size": 25,
+ "nb_train_samples": 2500000,
+ "nb_test_samples": 10000,
+ },
+ "grid": {
+ "model": "37M",
"batch_size": 25,
"nb_train_samples": 250000,
"nb_test_samples": 10000,
},
- "mnist": {
- "nb_epochs": 25,
+ "qmlp": {
+ "model": "37M",
"batch_size": 10,
- "nb_train_samples": 250000,
+ "nb_train_samples": 100000,
+ "nb_test_samples": 1000,
+ },
+ "guessop": {
+ "model": "352M",
+ "batch_size": 25,
+ "nb_train_samples": 1000000,
+ "nb_test_samples": 10000,
+ },
+ "learnop": {
+ "model": "37M",
+ "batch_size": 25,
+ "nb_train_samples": 50000,
"nb_test_samples": 10000,
},
"maze": {
- "nb_epochs": 25,
+ "model": "37M",
"batch_size": 5,
+ "nb_train_samples": 100000,
+ "nb_test_samples": 10000,
+ },
+ "picoclvr": {
+ "model": "37M",
+ "batch_size": 25,
"nb_train_samples": 250000,
"nb_test_samples": 10000,
},
+ "rpl": {
+ "model": "352M",
+ "batch_size": 5,
+ "nb_train_samples": 2500000,
+ "nb_test_samples": 10000,
+ },
"snake": {
- "nb_epochs": 5,
+ "model": "37M",
"batch_size": 25,
"nb_train_samples": 250000,
"nb_test_samples": 10000,
},
"stack": {
- "nb_epochs": 5,
+ "model": "37M",
"batch_size": 25,
"nb_train_samples": 100000,
"nb_test_samples": 1000,
},
- "expr": {
- "nb_epochs": 40,
+ "twotargets": {
+ "model": "37M",
"batch_size": 25,
- "nb_train_samples": 1000000,
+ "nb_train_samples": 50000,
"nb_test_samples": 10000,
},
- "rpl": {
- "nb_epochs": 40,
+ "degradation": {
+ "model": "37M",
"batch_size": 25,
"nb_train_samples": 100000,
"nb_test_samples": 10000,
},
- "world": {
- "nb_epochs": 10,
- "batch_size": 25,
- "nb_train_samples": 25000,
- "nb_test_samples": 1000,
+ "mnist": {
+ "model": "37M",
+ "batch_size": 10,
+ "nb_train_samples": 60000,
+ "nb_test_samples": 10000,
},
}
######################################################################
-if args.task == "sandbox":
- if args.sandbox_level == 0:
- problem = problems.ProblemLevel0(
- nb_sentences=args.sandbox_levels_nb_items,
- len_prompt=args.sandbox_levels_len_source,
- len_result=args.sandbox_levels_len_result,
- )
- elif args.sandbox_level == 1:
- problem = problems.ProblemLevel1(
- nb_operators=args.sandbox_levels_nb_items,
- len_source=args.sandbox_levels_len_source,
- len_result=args.sandbox_levels_len_result,
- )
- elif args.sandbox_level == 2:
- problem = problems.ProblemLevel2(
- len_source=args.sandbox_levels_len_source,
- len_result=args.sandbox_levels_len_result,
- )
- else:
- raise ValueError(f"Unknown sandbox level {args.sandbox_level}")
+if args.task == "byheart":
+ task = tasks.SandBox(
+ problem=problems.ProblemByHeart(),
+ nb_train_samples=args.nb_train_samples,
+ nb_test_samples=args.nb_test_samples,
+ batch_size=args.batch_size,
+ logger=log_string,
+ device=device,
+ )
+ args.max_percents_of_test_in_train = -1
+elif args.task == "learnop":
task = tasks.SandBox(
- # problem,
- # problems.ProblemAddition(zero_padded=False, inverted_result=False),
- # problems.ProblemLenId(len_max=args.sandbox_levels_len_source),
- problems.ProblemTwoTargets(len_total=16, len_targets=4),
+ problem=problems.ProblemLearnOperator(),
+ nb_train_samples=args.nb_train_samples,
+ nb_test_samples=args.nb_test_samples,
+ batch_size=args.batch_size,
+ logger=log_string,
+ device=device,
+ )
+
+
+elif args.task == "guessop":
+ task = tasks.SandBox(
+ problem=problems.ProblemGuessOperator(),
+ nb_train_samples=args.nb_train_samples,
+ nb_test_samples=args.nb_test_samples,
+ batch_size=args.batch_size,
+ logger=log_string,
+ device=device,
+ )
+
+
+elif args.task == "twotargets":
+ task = tasks.SandBox(
+ problem=problems.ProblemTwoTargets(),
+ nb_train_samples=args.nb_train_samples,
+ nb_test_samples=args.nb_test_samples,
+ batch_size=args.batch_size,
+ logger=log_string,
+ device=device,
+ )
+
+elif args.task == "degradation":
+ task = tasks.SandBox(
+ problem=problems.ProblemDegradation(hard=args.degradation_hard),
+ nb_train_samples=args.nb_train_samples,
+ nb_test_samples=args.nb_test_samples,
+ batch_size=args.batch_size,
+ logger=log_string,
+ device=device,
+ )
+
+elif args.task == "addition":
+ task = tasks.SandBox(
+ problem=problems.ProblemAddition(),
nb_train_samples=args.nb_train_samples,
nb_test_samples=args.nb_test_samples,
batch_size=args.batch_size,
device=device,
)
-elif args.task == "world":
- task = tasks.World(
+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 == "qmlp":
+ task = tasks.QMLP(
nb_train_samples=args.nb_train_samples,
nb_test_samples=args.nb_test_samples,
batch_size=args.batch_size,
- vqae_nb_epochs=args.world_vqae_nb_epochs,
+ result_dir=args.result_dir,
logger=log_string,
device=device,
)
######################################################################
# A bit of paranoia never hurts
+if args.max_percents_of_test_in_train >= 0:
+
+ def subsets_as_tuples(batches, cs):
+ s = set()
+ for batch in batches:
+ for x in batch:
+ s.add(tuple([v.item() for v in x]))
+ if len(s) == cs:
+ yield s
+ s = set()
+ yield s
+
+ nb_test, nb_in_train = 0, 0
+ for test_subset in subsets_as_tuples(task.batches(split="test"), 25000):
+ in_train = set()
+ for train_subset in subsets_as_tuples(task.batches(split="train"), 25000):
+ in_train.update(test_subset.intersection(train_subset))
+ nb_in_train += len(in_train)
+ nb_test += len(test_subset)
+
+ log_string(
+ f"data_check {nb_in_train*100/nb_test:.02f}% ({nb_in_train}/{nb_test}) of test samples are in the train set"
+ )
-def subsets_as_tuples(batches, cs):
- s = set()
- for batch in batches:
- for x in batch:
- s.add(tuple([v.item() for v in x]))
- if len(s) == cs:
- yield s
- s = set()
- yield s
-
-
-nb_test, nb_in_train = 0, 0
-for test_subset in subsets_as_tuples(task.batches(split="test"), 25000):
- in_train = set()
- for train_subset in subsets_as_tuples(task.batches(split="train"), 25000):
- in_train.update(test_subset.intersection(train_subset))
- nb_in_train += len(in_train)
- nb_test += len(test_subset)
-
-log_string(
- f"data_check {nb_in_train*100/nb_test:.02f}% ({nb_in_train}/{nb_test}) of test samples are in the train set"
-)
-
-assert (
- nb_in_train <= args.max_percents_of_test_in_train * nb_test / 100
-), f"More than {args.max_percents_of_test_in_train}% of test samples are in the train set"
+ assert (
+ nb_in_train <= args.max_percents_of_test_in_train * nb_test / 100
+ ), f"More than {args.max_percents_of_test_in_train}% of test samples are in the train set"
##############################