Update.

[picoclvr.git] / main.py

diff --git a/main.py b/main.py
index efcc0dd..1b0d39a 100755 (executable)
--- a/main.py
+++ b/main.py
@@ -5,9 +5,6 @@
 
 # Written by Francois Fleuret <francois@fleuret.org>
 
-# torch.backends.cuda.matmul.allow_tf23
-# torch.autocast(torch.bfloat16)
-
 import math, sys, argparse, time, tqdm, os
 
 import torch, torchvision
@@ -36,7 +33,7 @@ parser.add_argument(
     "--task",
     type=str,
     default="sandbox",
-    help="sandbox, picoclvr, mnist, maze, snake, stack, expr, world",
+    help="sandbox, picoclvr, mnist, maze, snake, stack, expr, rpl, world",
 )
 
 parser.add_argument("--log_filename", type=str, default="train.log", help=" ")
@@ -82,7 +79,20 @@ parser.add_argument("--overwrite_results", action="store_true", default=False)
 parser.add_argument("--checkpoint_name", type=str, default="checkpoint.pth")
 
 ##############################
-# picoclvr options
+# rpl options
+
+parser.add_argument("--rpl_nb_starting_values", type=int, default=5)
+
+parser.add_argument("--rpl_max_input", type=int, default=9)
+
+parser.add_argument("--rpl_prog_len", type=int, default=10)
+
+parser.add_argument("--rpl_nb_runs", type=int, default=8)
+
+parser.add_argument("--rpl_no_prog", action="store_true", default=False)
+
+##############################
+# sandbox options
 
 parser.add_argument("--sandbox_level", type=int, default=0)
 
@@ -206,6 +216,12 @@ default_task_args = {
         "nb_train_samples": 1000000,
         "nb_test_samples": 10000,
     },
+    "rpl": {
+        "nb_epochs": 40,
+        "batch_size": 25,
+        "nb_train_samples": 100000,
+        "nb_test_samples": 10000,
+    },
     "world": {
         "nb_epochs": 10,
         "batch_size": 25,
@@ -419,6 +435,20 @@ elif args.task == "expr":
         device=device,
     )
 
+elif args.task == "rpl":
+    task = tasks.RPL(
+        nb_train_samples=args.nb_train_samples,
+        nb_test_samples=args.nb_test_samples,
+        batch_size=args.batch_size,
+        nb_starting_values=args.rpl_nb_starting_values,
+        max_input=args.rpl_max_input,
+        prog_len=args.rpl_prog_len,
+        nb_runs=args.rpl_nb_runs,
+        no_prog=args.rpl_no_prog,
+        logger=log_string,
+        device=device,
+    )
+
 elif args.task == "world":
     task = tasks.World(
         nb_train_samples=args.nb_train_samples,
@@ -488,12 +518,12 @@ else:
 
 if args.task == "expr" and args.expr_input_file is not None:
     task.produce_results(
-        nb_epochs_finished,
-        model,
-        args.result_dir,
-        log_string,
-        args.deterministic_synthesis,
-        args.expr_input_file,
+        n_epoch=nb_epochs_finished,
+        model=model,
+        result_dir=args.result_dir,
+        logger=log_string,
+        deterministic_synthesis=args.deterministic_synthesis,
+        input_file=args.expr_input_file,
     )
 
     exit(0)
@@ -569,11 +599,11 @@ nb_samples_seen = 0
 
 if nb_epochs_finished >= nb_epochs:
     task.produce_results(
-        nb_epochs_finished,
-        model,
-        args.result_dir,
-        log_string,
-        args.deterministic_synthesis,
+        n_epoch=nb_epochs_finished,
+        model=model,
+        result_dir=args.result_dir,
+        logger=log_string,
+        deterministic_synthesis=args.deterministic_synthesis,
     )
 
 for n_epoch in range(nb_epochs_finished, nb_epochs):
@@ -627,7 +657,11 @@ for n_epoch in range(nb_epochs_finished, nb_epochs):
         )
 
         task.produce_results(
-            n_epoch, model, args.result_dir, log_string, args.deterministic_synthesis
+            n_epoch=n_epoch,
+            model=model,
+            result_dir=args.result_dir,
+            logger=log_string,
+            deterministic_synthesis=args.deterministic_synthesis,
         )
 
     checkpoint = {