Update.
authorFrançois Fleuret <francois@fleuret.org>
Tue, 18 Jul 2023 15:26:17 +0000 (17:26 +0200)
committerFrançois Fleuret <francois@fleuret.org>
Tue, 18 Jul 2023 15:26:17 +0000 (17:26 +0200)
main.py
tasks.py

diff --git a/main.py b/main.py
index 3be3d55..213524e 100755 (executable)
--- a/main.py
+++ b/main.py
@@ -266,7 +266,8 @@ picoclvr_pruner_eval = (
 
 if args.task == "sandbox":
     task = tasks.SandBox(
-        tasks.ProblemByheart(),
+        tasks.ProblemLevel1(),
+        # 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,
index fb85576..332d6c5 100755 (executable)
--- a/tasks.py
+++ b/tasks.py
@@ -67,13 +67,15 @@ class Problem:
     def generate_sequences(self, nb):
         pass
 
-    def log_performance(self, sequences, logger):
-        pass
+    def seq2str(self, seq):
+        return "[NOT IMPLEMENTED]"
+
+
+####################
 
 
-class ProblemByheart(Problem):
-    def __init__(self):
-        nb_seq, len_prompt, len_result = 100, 5, 5
+class ProblemLevel0(Problem):
+    def __init__(self, nb_sentences=100, len_prompt=5, len_result=5):
         self.seq = torch.randint(10, (nb_seq, len_prompt + 1 + len_result))
         self.seq[:, len_prompt] = 10
 
@@ -83,20 +85,104 @@ class ProblemByheart(Problem):
         ar_mask = (ar_mask.cumsum(1) - ar_mask).clamp(max=1)
         return sequences, ar_mask
 
-        # problems = [ProblemByheart()]
-        # nb_common_codes = 100
 
-        # def generate_sequences(nb_samples):
-        # problem_indexes = torch.randint(len(problems), (nb_samples,))
-        # nb_samples_per_problem = torch.one_hot(problem_indexes).sum(0)
-        # print(f"{nb_samples_per_problem}")
-        # all_seq = []
-        # for nb, p in zip(nb_samples_per_problem, problems):
-        # all_seq.append(p.generate_sequences(nb_samples_per_problem[nb]))
-        # return all_seq
+class ProblemLevel1(Problem):
+    def __init__(self, nb_operators=100, len_prompt=5, len_result=8):
+        self.len_prompt = len_prompt
+        self.len_result = len_result
+        self.len_nb_operator = int(math.log(nb_operators) / math.log(10)) + 1
+        self.operators = F.one_hot(
+            torch.rand(nb_operators, len_result, len_prompt).argmax(-1),
+            num_classes=len_prompt,
+        )
+
+    def generate_sequences(self, nb):
+        a = self.len_nb_operator
+        b = a + 1 + self.len_prompt
+        sequences = torch.empty(nb, b + 1 + self.len_result, dtype=torch.int64)
+        nb_operators = torch.randint(self.operators.size(0), (nb,))
+        sequences[:, :a] = (nb_operators[:, None] / 10 ** torch.arange(a)) % 10
+        sequences[:, a] = 10
+        sequences[:, a + 1 : b] = torch.randint(10, (nb, b - a - 1))
+        sequences[:, b] = 11
+
+        o = self.operators[nb_operators]
+        p = sequences[:, a + 1 : b]
+        print(f"{o.size()=} {p.size()=} {sequences[:,b+1:].size()=}")
+        sequences[:, b + 1 :] = o.bmm(p[:, :, None]).squeeze(-1)
+        ar_mask = (sequences == 11).long()
+        ar_mask = (ar_mask.cumsum(1) - ar_mask).clamp(max=1)
+        return sequences, ar_mask
+
+    def seq2str(self, seq):
+        return "".join(self.id2char[x.item()] for x in seq)
 
-        # for strain, stest in zip(train_seq, test_seq):
-        # s = torch.cat((strain, stest), 0)
+
+####################
+
+
+class ProblemAddition(Problem):
+    def __init__(self, nb_digits=10, zero_padded=False, inverted_result=False):
+        self.nb_digits = nb_digits
+        self.zero_padded = zero_padded
+        self.inverted_result = inverted_result
+        self.char2id = dict([(c, n) for n, c in enumerate("0123456789+=$")])
+        self.id2char = dict([(n, c) for c, n in self.char2id.items()])
+
+    def tensorize(self, strings):
+        len_max = max([len(x) for x in strings])
+        return torch.cat(
+            [
+                torch.tensor(
+                    [
+                        [self.char2id[c] for c in s + "$" * (len_max - len(s))]
+                        for s in strings
+                    ]
+                )
+            ],
+            0,
+        )
+
+    def generate_sequences(self, nb):
+        sequences = []
+        for k in range(nb):
+            a, b = torch.randint(10**self.nb_digits, (2,))
+            c = a + b
+            a, b, c = str(a.item()), str(b.item()), str(c.item())
+            if self.zero_padded:
+                a = "0" * (self.nb_digits - len(a)) + a
+                b = "0" * (self.nb_digits - len(b)) + b
+                c = "0" * (self.nb_digits + 1 - len(c)) + c
+            if self.inverted_result:
+                c = c[::-1]
+            sequences.append(f"{a}+{b}={c}$")
+
+        sequences = self.tensorize(sequences)
+        ar_mask = (sequences == self.char2id["="]).long()
+        ar_mask = (ar_mask.cumsum(1) - ar_mask).clamp(max=1)
+        return sequences, ar_mask
+
+    def seq2str(self, seq):
+        return "".join(self.id2char[x.item()] for x in seq)
+
+
+# class ProblemUnion(Problem):
+# problems = [ProblemByheart()]
+# nb_common_codes = 100
+
+# def generate_sequences(nb_samples):
+# problem_indexes = torch.randint(len(problems), (nb_samples,))
+# nb_samples_per_problem = torch.one_hot(problem_indexes).sum(0)
+# print(f"{nb_samples_per_problem}")
+# all_seq = []
+# for nb, p in zip(nb_samples_per_problem, problems):
+# all_seq.append(p.generate_sequences(nb_samples_per_problem[nb]))
+# return all_seq
+
+# for strain, stest in zip(train_seq, test_seq):
+# s = torch.cat((strain, stest), 0)
+
+####################
 
 
 class SandBox(Task):
@@ -114,11 +200,21 @@ class SandBox(Task):
 
         self.batch_size = batch_size
         self.device = device
+        self.problem = problem
 
-        self.train_input, self.train_ar_mask = problem.generate_sequences(
+        self.train_input, self.train_ar_mask = self.problem.generate_sequences(
             nb_train_samples
         )
-        self.test_input, self.test_ar_mask = problem.generate_sequences(nb_test_samples)
+        self.test_input, self.test_ar_mask = self.problem.generate_sequences(
+            nb_test_samples
+        )
+
+        self.train_input, self.train_ar_mask = self.train_input.to(
+            device
+        ), self.train_ar_mask.to(device)
+        self.test_input, self.test_ar_mask = self.test_input.to(
+            device
+        ), self.test_ar_mask.to(device)
 
         self.nb_codes = max(self.train_input.max(), self.test_input.max()) + 1
 
@@ -147,10 +243,12 @@ class SandBox(Task):
         return self.nb_codes
 
     def produce_results(
-        self, n_epoch, model, result_dir, logger, deterministic_synthesis
+        self, n_epoch, model, result_dir, logger, deterministic_synthesis, nmax=1000
     ):
-        def compute_accuracy(input, ar_mask):
+        def compute_accuracy(input, ar_mask, logger=None):
+            input, ar_mask = input[:nmax], ar_mask[:nmax]
             result = input.clone() * (1 - ar_mask)
+
             masked_inplace_autoregression(
                 model,
                 self.batch_size,
@@ -161,6 +259,15 @@ class SandBox(Task):
                 device=self.device,
             )
 
+            if logger is not None:
+                for sp, st in zip(result[:10], input[:10]):
+                    logger(
+                        f"test_sequences {n_epoch} prediction   {self.problem.seq2str(sp)}"
+                    )
+                    logger(
+                        f"               {n_epoch} ground truth {self.problem.seq2str(st)}"
+                    )
+
             nb_total = ar_mask.sum().item()
             nb_correct = ((result == input).long() * ar_mask).sum().item()
 
@@ -175,7 +282,7 @@ class SandBox(Task):
         )
 
         test_nb_total, test_nb_correct = compute_accuracy(
-            self.test_input, self.test_ar_mask
+            self.test_input, self.test_ar_mask, logger
         )
 
         logger(
@@ -1119,8 +1226,6 @@ class World(Task):
             device_storage=device_storage,
         )
 
-        print(f"{train_action_seq.size()=}")
-
         train_frame_seq = self.frame2seq(train_frames).to(device_storage)
         test_frame_seq = self.frame2seq(test_frames).to(device_storage)
 
@@ -1132,7 +1237,7 @@ class World(Task):
         self.nb_codes = nb_frame_codes + nb_action_codes
 
         train_frame_seq = train_frame_seq.reshape(train_frame_seq.size(0) // 2, 2, -1)
-        print(f"{train_action_seq.device=} {nb_frame_codes.device=}")
+
         train_action_seq += nb_frame_codes
         self.train_input = torch.cat(
             (train_frame_seq[:, 0, :], train_action_seq, train_frame_seq[:, 1, :]), 1
@@ -1191,7 +1296,6 @@ class World(Task):
             (seq_start[:, None, :], seq_end[:, None, :], seq_predicted[:, None, :]), 1
         )
         result = result.reshape(-1, result.size(-1))
-        print(f"{result.size()=}")
 
         frames = self.seq2frame(result)
         image_name = os.path.join(result_dir, f"world_result_{n_epoch:04d}.png")