X-Git-Url: https://fleuret.org/cgi-bin/gitweb/gitweb.cgi?a=blobdiff_plain;f=main.py;h=9a3d34633bd88cff4ac1e05ee77989e4d725e7b1;hb=960c93d7c0aea41d180814c46d3a05686a426764;hp=d1f82cf20cd19b539f17c95da6a32d6c47547089;hpb=5703df4c32a0856c8fa4b1ff97810cdc1fb76253;p=picoclvr.git

diff --git a/main.py b/main.py
index d1f82cf..9a3d346 100755
--- 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
@@ -15,7 +12,7 @@ from torch import nn
 from torch.nn import functional as F
 
 import ffutils
-import mygpt, tasks
+import mygpt, tasks, problems
 
 ######################################################################
 
@@ -36,7 +33,7 @@ parser.add_argument(
     "--task",
     type=str,
     default="sandbox",
-    help="sandbox, picoclvr, mnist, maze, snake, stack, expr, rpl, world",
+    help="byheart, learnop, guessop, twotargets, addition, picoclvr, mnist, maze, snake, stack, expr, rpl",
 )
 
 parser.add_argument("--log_filename", type=str, default="train.log", help=" ")
@@ -45,6 +42,10 @@ parser.add_argument("--result_dir", type=str, default=None)
 
 parser.add_argument("--seed", type=int, default=0)
 
+parser.add_argument("--max_percents_of_test_in_train", type=int, default=1)
+
+########################################
+
 parser.add_argument("--nb_epochs", type=int, default=None)
 
 parser.add_argument("--batch_size", type=int, default=None)
@@ -59,6 +60,8 @@ parser.add_argument("--learning_rate", type=float, default=1e-4)
 
 parser.add_argument("--learning_rate_schedule", type=str, default="10: 2e-5,30: 4e-6")
 
+########################################
+
 parser.add_argument("--model", type=str, default="37M")
 
 parser.add_argument("--dim_model", type=int, default=None)
@@ -73,6 +76,8 @@ parser.add_argument("--nb_blocks", type=int, default=None)
 
 parser.add_argument("--dropout", type=float, default=0.1)
 
+########################################
+
 parser.add_argument("--deterministic_synthesis", action="store_true", default=False)
 
 parser.add_argument("--no_checkpoint", action="store_true", default=False)
@@ -82,15 +87,17 @@ 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("--sandbox_level", type=int, default=0)
+parser.add_argument("--rpl_max_input", type=int, default=9)
 
-parser.add_argument("--sandbox_levels_nb_items", type=int, default=25)
+parser.add_argument("--rpl_prog_len", type=int, default=10)
 
-parser.add_argument("--sandbox_levels_len_source", type=int, default=6)
+parser.add_argument("--rpl_nb_runs", type=int, default=8)
 
-parser.add_argument("--sandbox_levels_len_result", type=int, default=8)
+parser.add_argument("--rpl_no_prog", action="store_true", default=False)
 
 ##############################
 # picoclvr options
@@ -323,30 +330,52 @@ picoclvr_pruner_eval = (
 
 ######################################################################
 
-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}")
+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,
+    )
+
+
+elif args.task == "learnop":
+    task = tasks.SandBox(
+        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,
-        # tasks.ProblemAddition(zero_padded=False, inverted_result=False),
+        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 == "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,
@@ -430,6 +459,12 @@ elif args.task == "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,
     )
 
@@ -502,12 +537,12 @@ 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)
@@ -525,34 +560,37 @@ token_probas = token_count / token_count.sum()
 entropy = -torch.xlogy(token_probas, token_probas).sum()
 train_set_perplexity = math.exp(entropy)
 
-##############################
-
+######################################################################
 # A bit of paranoia never hurts
 
-train_examples = {}
 
+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
 
-for input in task.batches(split="train"):
-    assert input.dim() == 2 and input.dtype == torch.int64
-    for x in input:
-        train_examples[x.sum().item()] = x
-
-nb_total, nb_collisions = 0, 0
-for input in task.batches(split="test"):
-    assert input.dim() == 2 and input.dtype == torch.int64
-    for x in input:
-        nb_total += 1
-        y = train_examples.get(x.sum().item())
-        if y is not None:
-            if x.size() == y.size() and (x - y).abs().sum() == 0:
-                nb_collisions += 1
-
-del train_examples
+
+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_collisions*100/nb_total:.02f}% ({nb_collisions}/{nb_total}) of test samples are in the train set"
+    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"
+
 ##############################
 
 if args.learning_rate_schedule == "cos":
@@ -583,11 +621,11 @@ 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):
@@ -641,7 +679,11 @@ 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 = {