[BugFix] Fix slicesampler terminated/truncated signaling

test/test_rb.py

@@ -18,6 +18,8 @@
 import torch
 
 from _utils_internal import get_default_devices, make_tc
+
+from mocking_classes import CountingEnv
 from packaging import version
 from packaging.version import parse
 from tensordict import (
@@ -30,7 +32,6 @@
 )
 from torch import multiprocessing as mp
 from torch.utils._pytree import tree_flatten, tree_map
-
 from torchrl.collectors import RandomPolicy, SyncDataCollector
 from torchrl.collectors.utils import split_trajectories
 from torchrl.data import (
@@ -2025,8 +2026,11 @@ def test_slice_sampler(
             assert too_short
 
         assert len(trajs_unique_id) == 4
+        done = info[("next", "done")]
+        assert done.view(num_slices, -1)[:, -1].all()
         truncated = info[("next", "truncated")]
-        assert truncated.view(num_slices, -1)[:, -1].all()
+        terminated = info[("next", "terminated")]
+        assert (truncated | terminated).view(num_slices, -1)[:, -1].all()
 
     @pytest.mark.parametrize("sampler", [SliceSampler, SliceSamplerWithoutReplacement])
     def test_slice_sampler_at_capacity(self, sampler):
@@ -2166,8 +2170,10 @@ def test_slice_sampler_without_replacement(
             trajs_unique_id = trajs_unique_id.union(
                 cur_episodes,
             )
-        truncated = info[("next", "truncated")]
-        assert truncated.view(num_slices, -1)[:, -1].all()
+        done = info[("next", "done")]
+        assert done.view(num_slices, -1)[:, -1].all()
+        done_recon = info[("next", "truncated")] | info[("next", "terminated")]
+        assert done_recon.view(num_slices, -1)[:, -1].all()
 
     def test_slicesampler_strictlength(self):
 
@@ -2792,6 +2798,43 @@ def test_rb_multidim_collector(
             print(f"rb {rb}")  # noqa: T201
             raise
 
+    @pytest.mark.parametrize("strict_length", [True, False])
+    def test_done_slicesampler(self, strict_length):
+        env = SerialEnv(
+            3,
+            [
+                lambda: CountingEnv(max_steps=31),
+                lambda: CountingEnv(max_steps=32),
+                lambda: CountingEnv(max_steps=33),
+            ],
+        )
+        full_action_spec = CountingEnv(max_steps=32).full_action_spec
+        policy = lambda td: td.update(
+            full_action_spec.zero((3,)).apply_(lambda x: x + 1)
+        )
+        rb = TensorDictReplayBuffer(
+            storage=LazyTensorStorage(200, ndim=2),
+            sampler=SliceSampler(
+                slice_len=32,
+                strict_length=strict_length,
+                truncated_key=("next", "truncated"),
+            ),
+            batch_size=128,
+        )
+
+        for i in range(50):
+            r = env.rollout(50, policy=policy, break_when_any_done=False)
+            r["next", "done"][:, -1] = 1
+            rb.extend(r)
+
+            sample = rb.sample()
+
+            assert sample["next", "done"].sum() == 128 // 32, (
+                i,
+                sample["next", "done"].sum(),
+            )
+            assert (split_trajectories(sample)["next", "done"].sum(-2) == 1).all()
+
 
 if __name__ == "__main__":
     args, unknown = argparse.ArgumentParser().parse_known_args()

torchrl/data/replay_buffers/samplers.py

@@ -1160,9 +1160,9 @@ def _get_index(
             terminated = torch.zeros_like(truncated)
             if traj_terminated.any():
                 if isinstance(seq_length, int):
-                    truncated.view(num_slices, -1)[traj_terminated] = 1
+                    terminated.view(num_slices, -1)[traj_terminated, -1] = 1
                 else:
-                    truncated[(seq_length.cumsum(0) - 1)[traj_terminated]] = 1
+                    terminated[(seq_length.cumsum(0) - 1)[traj_terminated]] = 1
             truncated = truncated & ~terminated
             done = terminated | truncated
             return index.to(torch.long).unbind(-1), {
@@ -1726,9 +1726,9 @@ def sample(self, storage: Storage, batch_size: int) -> Tuple[torch.Tensor, dict]
             terminated = torch.zeros_like(truncated)
             if traj_terminated.any():
                 if isinstance(seq_length, int):
-                    truncated.view(num_slices, -1)[traj_terminated] = 1
+                    terminated.view(num_slices, -1)[:, traj_terminated] = 1
                 else:
-                    truncated[(seq_length.cumsum(0) - 1)[traj_terminated]] = 1
+                    terminated[(seq_length.cumsum(0) - 1)[traj_terminated]] = 1
             truncated = truncated & ~terminated
             done = terminated | truncated