Update.

diff --git a/escape.py b/escape.py
index 7fcafaa..aa87cdd 100755 (executable)
--- a/escape.py
+++ b/escape.py
@@ -150,14 +150,16 @@ def episodes2str(states, actions, rewards, unicode=False, ansi_colors=False):
     result = hline
 
     for n in range(states.size(0)):
+
+        def state_symbol(v):
+            v = v.item()
+            return "?" if v < 0 or v >= len(symbols) else symbols[v]
+
         for i in range(states.size(2)):
             result += (
                 vert
                 + vert.join(
-                    [
-                        "".join([symbols[v.item()] for v in row])
-                        for row in states[n, :, i]
-                    ]
+                    ["".join([state_symbol(v) for v in row]) for row in states[n, :, i]]
                 )
                 + vert
                 + "\n"
@@ -166,8 +168,9 @@ def episodes2str(states, actions, rewards, unicode=False, ansi_colors=False):
         result += (vert + thin_hori * states.size(-1)) * states.size(1) + vert + "\n"
 
         def status_bar(a, r):
-            a = "ISNEW"[a.item()]
-            r = "" if r == 0 else f"{r.item()}"
+            a = a.item()
+            a = "ISNEW"[a] if a >= 0 and a < 5 else "?"
+            r = "?" if r < -1 or r > 2 else ("" if r == 0 else f"{r.item()}")
             return a + " " * (states.size(-1) - len(a) - len(r)) + r
 
         result += (

diff --git a/tasks.py b/tasks.py
index 8b22cad..dba6e13 100755 (executable)
--- a/tasks.py
+++ b/tasks.py
@@ -1881,13 +1881,15 @@ class Escape(Task):
 
         self.batch_size = batch_size
         self.device = device
+        self.height = height
+        self.width = width
 
         states, actions, rewards = escape.generate_episodes(
             nb_train_samples + nb_test_samples, height, width, T
         )
         seq = escape.episodes2seq(states, actions, rewards)
-        self.train_input = seq[:nb_train_samples]
-        self.test_input = seq[nb_train_samples:]
+        self.train_input = seq[:nb_train_samples].to(self.device)
+        self.test_input = seq[nb_train_samples:].to(self.device)
 
         self.nb_codes = max(self.train_input.max(), self.test_input.max()) + 1
 
@@ -1915,7 +1917,30 @@ class Escape(Task):
     def produce_results(
         self, n_epoch, model, result_dir, logger, deterministic_synthesis, nmax=1000
     ):
-        pass
+        result = self.test_input[:100].clone()
+        ar_mask = (
+            torch.arange(result.size(1), device=result.device)
+            > self.height * self.width + 2
+        ).long()[None, :]
+        ar_mask = ar_mask.expand_as(result)
+        result *= 1 - ar_mask  # paraaaaanoiaaaaaaa
+
+        masked_inplace_autoregression(
+            model,
+            self.batch_size,
+            result,
+            ar_mask,
+            deterministic_synthesis,
+            device=self.device,
+        )
+
+        s, a, r = escape.seq2episodes(result, self.height, self.width)
+        str = escape.episodes2str(s, a, r, unicode=True, ansi_colors=True)
+
+        filename = os.path.join(result_dir, f"test_seq_{n_epoch:04d}.txt")
+        with open(filename, "w") as f:
+            f.write(str)
+            logger(f"wrote {filename}")
 
 
 ######################################################################