Update.
authorFrançois Fleuret <francois@fleuret.org>
Thu, 20 Jul 2023 22:14:16 +0000 (00:14 +0200)
committerFrançois Fleuret <francois@fleuret.org>
Thu, 20 Jul 2023 22:14:16 +0000 (00:14 +0200)
rpl.py
tasks.py

diff --git a/rpl.py b/rpl.py
index f826fc4..b51edef 100755 (executable)
--- a/rpl.py
+++ b/rpl.py
@@ -105,8 +105,10 @@ def decompose(seq):
     k = 0
     while seq[k] == "<input>":
         o = next_marker(seq, ["<output>"], start=k + 1)
+        if o is None:
+            raise ValueError("Missing output markers (should be correct in the prompt)")
         e = next_marker(seq, ["<input>", "<prog>"], start=o)
-        if o is None or e is None:
+        if e is None:
             raise ValueError(
                 "Missing input/output markers (should be correct in the prompt)"
             )
@@ -133,6 +135,12 @@ def decompose(seq):
     return prog, io
 
 
+def stack_distance(target_stack, result_stack):
+    return abs(len(result_stack) - len(target_stack)) + sum(
+        [0 if x == y else 1 for x, y in zip(result_stack, target_stack)]
+    )
+
+
 def compute_nb_errors(seq):
     prog, io = decompose(seq)
 
@@ -152,9 +160,7 @@ def compute_nb_errors(seq):
         for start_stack, target_stack in io:
             result_stack = rpl_exec(prog, start_stack)
             nb_total += len(target_stack)
-            e = abs(len(result_stack) - len(target_stack)) + sum(
-                [0 if x == y else 1 for x, y in zip(result_stack, target_stack)]
-            )
+            e = stack_distance(target_stack, result_stack)
             nb_errors += e
             stacks.append((start_stack, target_stack, result_stack, e == 0))
 
index 0827a44..da39a83 100755 (executable)
--- a/tasks.py
+++ b/tasks.py
@@ -1182,9 +1182,13 @@ class RPL(Task):
         def compute_nb_errors_output(input, nb_to_log=0):
             result = input.clone()
             k = torch.arange(result.size(1), device=result.device)[None, :]
-            last_output_idx = ((result == self.t_output) * k).max(dim=1, keep_dim=True)
-            first_prog_idx = ((result == self.t_prog) * k).min(dim=1, keep_dim=True)
-            ar_mask = (k > last_output_idx).long() * (k < first_prog_idx)
+            last_output_idx = (
+                ((result == self.t_output) * k).max(dim=1, keepdim=True).values
+            )
+            first_prog_idx = (
+                ((result == self.t_prog) * k).max(dim=1, keepdim=True).values
+            )
+            ar_mask = (k > last_output_idx).long() * (k < first_prog_idx).long()
             result = (1 - ar_mask) * result + ar_mask * self.t_nul
 
             masked_inplace_autoregression(
@@ -1197,25 +1201,20 @@ class RPL(Task):
             )
 
             sum_nb_total, sum_nb_errors = 0, 0
-            for x, y in zip(input, result):
+            for x, y, i, j in zip(input, result, last_output_idx, first_prog_idx):
                 seq = [self.id2token[i.item()] for i in y]
                 sum_nb_total += 1
-                sum_nb_errors += 0 if (x - y).abs().max() == 0 else 1
+                correct = (x - y).abs().max() == 0
+                sum_nb_errors += 0 if correct else 1
                 if nb_to_log > 0:
-                    gt_seq = [self.id2token[i.item()] for i in x]
-                    _, _, gt_prog, _ = rpl.compute_nb_errors(gt_seq)
-                    gt_prog = " ".join([str(x) for x in gt_prog])
-                    prog = " ".join([str(x) for x in prog])
-                    comment = "*" if nb_errors == 0 else "-"
-                    logger(f"{comment} PROG [{gt_prog}] PREDICTED [{prog}]")
-                    for start_stack, target_stack, result_stack, correct in stacks:
-                        comment = "*" if correct else "-"
-                        start_stack = " ".join([str(x) for x in start_stack])
-                        target_stack = " ".join([str(x) for x in target_stack])
-                        result_stack = " ".join([str(x) for x in result_stack])
-                        logger(
-                            f"  {comment} [{start_stack}] -> [{target_stack}] PREDICTED [{result_stack}]"
-                        )
+                    result_stack = [self.id2token[i.item()] for i in y[i : j + 1]]
+                    target_stack = [self.id2token[i.item()] for i in x[i : j + 1]]
+                    comment = "*" if correct else "-"
+                    result_stack = " ".join([str(x) for x in result_stack])
+                    target_stack = " ".join([str(x) for x in target_stack])
+                    logger(
+                        f"output_test {comment} [{target_stack}] PREDICTED [{result_stack}]"
+                    )
                     nb_to_log -= 1
 
             return sum_nb_total, sum_nb_errors
@@ -1227,7 +1226,15 @@ class RPL(Task):
         )
 
         logger(
-            f"accuracy_test {n_epoch} nb_total {test_nb_total} nb_errors {test_nb_errors} accuracy {100.0*(1-test_nb_errors/test_nb_total):.02f}%"
+            f"accuracy_prog_test {n_epoch} nb_total {test_nb_total} nb_errors {test_nb_errors} accuracy {100.0*(1-test_nb_errors/test_nb_total):.02f}%"
+        )
+
+        test_nb_total, test_nb_errors = compute_nb_errors_output(
+            self.test_input[:1000].to(self.device), nb_to_log=10
+        )
+
+        logger(
+            f"accuracy_output_test {n_epoch} nb_total {test_nb_total} nb_errors {test_nb_errors} accuracy {100.0*(1-test_nb_errors/test_nb_total):.02f}%"
         )