from torch import nn
from torch.nn import functional as F
+import ffutils
import mygpt, tasks
######################################################################
parser.add_argument("--learning_rate_schedule", type=str, default="10: 2e-5,30: 4e-6")
-parser.add_argument("--dim_model", type=int, default=512)
+parser.add_argument("--model", type=str, default="37M")
-parser.add_argument("--dim_keys", type=int, default=64)
+parser.add_argument("--dim_model", type=int, default=None)
-parser.add_argument("--dim_hidden", type=int, default=2048)
+parser.add_argument("--dim_keys", type=int, default=None)
-parser.add_argument("--nb_heads", type=int, default=8)
+parser.add_argument("--dim_hidden", type=int, default=None)
-parser.add_argument("--nb_blocks", type=int, default=12)
+parser.add_argument("--nb_heads", type=int, default=None)
+
+parser.add_argument("--nb_blocks", type=int, default=None)
parser.add_argument("--dropout", type=float, default=0.1)
##############################
# picoclvr 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)
+
+##############################
+# picoclvr options
+
parser.add_argument("--picoclvr_nb_colors", type=int, default=5)
parser.add_argument("--picoclvr_height", type=int, default=12)
######################################################################
-default_args = {
+default_task_args = {
"sandbox": {
- "nb_epochs": 10,
+ "nb_epochs": 50,
"batch_size": 25,
- "nb_train_samples": 25000,
+ "nb_train_samples": 100000,
"nb_test_samples": 10000,
},
"picoclvr": {
},
}
-if args.task in default_args:
- for k, v in default_args[args.task].items():
+if args.task in default_task_args:
+ for k, v in default_task_args[args.task].items():
+ if getattr(args, k) is None:
+ setattr(args, k, v)
+
+######################################################################
+
+default_model_args = {
+ "17K": {
+ "dim_model": 32,
+ "dim_keys": 32,
+ "dim_hidden": 32,
+ "nb_heads": 2,
+ "nb_blocks": 2,
+ },
+ "37M": {
+ "dim_model": 512,
+ "dim_keys": 64,
+ "dim_hidden": 2048,
+ "nb_heads": 8,
+ "nb_blocks": 12,
+ },
+ "122M": {
+ "dim_model": 768,
+ "dim_keys": 64,
+ "dim_hidden": 2048,
+ "nb_heads": 8,
+ "nb_blocks": 24,
+ },
+ "352M": {
+ "dim_model": 1024,
+ "dim_keys": 64,
+ "dim_hidden": 2048,
+ "nb_heads": 8,
+ "nb_blocks": 48,
+ },
+}
+
+if args.model in default_model_args:
+ for k, v in default_model_args[args.model].items():
if getattr(args, k) is None:
setattr(args, k, v)
+else:
+ raise ValueError(f"Unknown model {args.model}")
######################################################################
######################################################################
if args.task == "sandbox":
+ if args.sandbox_level == 0:
+ problem = tasks.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 = tasks.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 = tasks.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}")
+
task = tasks.SandBox(
+ problem,
+ # tasks.ProblemAddition(zero_padded=False, inverted_result=False),
nb_train_samples=args.nb_train_samples,
nb_test_samples=args.nb_test_samples,
batch_size=args.batch_size,