diff --git a/.github/scripts/m1_script.sh b/.github/scripts/m1_script.sh
index fd374bf79..cddc40035 100644
--- a/.github/scripts/m1_script.sh
+++ b/.github/scripts/m1_script.sh
@@ -1,3 +1,3 @@
 #!/bin/bash
 
-export TENSORDICT_BUILD_VERSION=0.4.0
+export TENSORDICT_BUILD_VERSION=0.5.0
diff --git a/.github/workflows/wheels.yml b/.github/workflows/wheels.yml
index c67317d42..4ab4adb81 100644
--- a/.github/workflows/wheels.yml
+++ b/.github/workflows/wheels.yml
@@ -38,7 +38,7 @@ jobs:
         run: |
           export PATH="/opt/python/${{ matrix.python_version[1] }}/bin:$PATH"
           python3 -mpip install wheel
-          TENSORDICT_BUILD_VERSION=0.4.0 python3 setup.py bdist_wheel
+          TENSORDICT_BUILD_VERSION=0.5.0 python3 setup.py bdist_wheel
           # NB: wheels have the linux_x86_64 tag so we rename to manylinux1
           # find . -name 'dist/*whl' -exec bash -c ' mv $0 ${0/linux/manylinux1}' {} \;
       # pytorch/pytorch binaries are also manylinux_2_17 compliant but they
@@ -84,7 +84,7 @@ jobs:
         shell: bash
         run: |
             python3 -mpip install wheel setuptools
-            TENSORDICT_BUILD_VERSION=0.4.0 python3 setup.py bdist_wheel
+            TENSORDICT_BUILD_VERSION=0.5.0 python3 setup.py bdist_wheel
       - name: Upload wheel for the test-wheel job
         uses: actions/upload-artifact@v2
         with:
diff --git a/benchmarks/distributed/dataloading.py b/benchmarks/distributed/dataloading.py
index 5d1777b57..0c8b03cdf 100644
--- a/benchmarks/distributed/dataloading.py
+++ b/benchmarks/distributed/dataloading.py
@@ -28,8 +28,7 @@
 import torch
 import tqdm
 
-from tensordict import MemoryMappedTensor
-from tensordict.prototype import tensorclass
+from tensordict import MemoryMappedTensor, tensorclass
 from torch import multiprocessing as mp, nn
 from torch.distributed import rpc
 from torch.utils.data import DataLoader
diff --git a/benchmarks/tensorclass/test_tensorclass_speed.py b/benchmarks/tensorclass/test_tensorclass_speed.py
index 0c8bf8c7d..452f6b86a 100644
--- a/benchmarks/tensorclass/test_tensorclass_speed.py
+++ b/benchmarks/tensorclass/test_tensorclass_speed.py
@@ -9,7 +9,7 @@
 import pytest
 import torch
 
-from tensordict.prototype import tensorclass
+from tensordict import tensorclass
 
 
 @tensorclass
diff --git a/benchmarks/tensorclass/test_torch_functions.py b/benchmarks/tensorclass/test_torch_functions.py
index 0dc3c560c..7ed717872 100644
--- a/benchmarks/tensorclass/test_torch_functions.py
+++ b/benchmarks/tensorclass/test_torch_functions.py
@@ -6,7 +6,7 @@
 import pytest
 import torch
 
-from tensordict.prototype import tensorclass
+from tensordict import tensorclass
 
 
 @tensorclass
diff --git a/tensordict/_lazy.py b/tensordict/_lazy.py
index 4d08ce4cd..70c456931 100644
--- a/tensordict/_lazy.py
+++ b/tensordict/_lazy.py
@@ -27,6 +27,7 @@
     Tuple,
     Type,
 )
+from warnings import warn
 
 import numpy as np
 
@@ -1695,8 +1696,18 @@ def _apply_nest(
                 )
                 for i, (td, *oth) in enumerate(zip(self.tensordicts, *others))
             ]
-        if filter_empty and all(r is None for r in results):
-            return
+        if all(r is None for r in results):
+            if filter_empty is None:
+                warn(
+                    "Your resulting tensordict has no leaves but you did not specify filter_empty=True. "
+                    "This now returns None (filter_empty=True). "
+                    "To silence this warning, set filter_empty to the desired value in your call to `apply`. "
+                    "This warning will be removed in v0.6.",
+                    category=DeprecationWarning,
+                )
+                return
+            elif filter_empty:
+                return
         if not inplace:
             out = type(self)(
                 *results,
diff --git a/tensordict/_td.py b/tensordict/_td.py
index d2df70d11..d1f5aacff 100644
--- a/tensordict/_td.py
+++ b/tensordict/_td.py
@@ -1237,12 +1237,13 @@ def make_result(names=names, batch_size=batch_size):
             # After we introduce the new behaviour, we will have to consider what happens
             # to empty tensordicts by default: will they disappear or stay?
             warn(
-                "Your resulting tensordict has no leaves but you did not specify filter_empty=False. "
-                "Currently, this returns an empty tree (filter_empty=True), but from v0.5 it will return "
-                "a None unless filter_empty=False. "
-                "To silence this warning, set filter_empty to the desired value in your call to `apply`.",
+                "Your resulting tensordict has no leaves but you did not specify filter_empty=True. "
+                "This now returns None (filter_empty=True). "
+                "To silence this warning, set filter_empty to the desired value in your call to `apply`. "
+                "This warning will be removed in v0.6.",
                 category=DeprecationWarning,
             )
+            return
         if result is None:
             result = make_result()
 
diff --git a/tensordict/functional.py b/tensordict/functional.py
index 23367f998..ea11d0423 100644
--- a/tensordict/functional.py
+++ b/tensordict/functional.py
@@ -141,7 +141,8 @@ def pad_sequence(
         )
     if batch_first is not None:
         warnings.warn(
-            "The batch_first argument is deprecated and will be removed in a future release. The output will always be batch_first.",
+            "The batch_first argument is deprecated and will be removed in v0.6. "
+            "The output will always be batch_first.",
             category=DeprecationWarning,
         )
 
diff --git a/tensordict/nn/common.py b/tensordict/nn/common.py
index cc53aaef7..77862465e 100644
--- a/tensordict/nn/common.py
+++ b/tensordict/nn/common.py
@@ -1163,7 +1163,8 @@ def forward(
                 # special rule for TensorDictModule that states that the second arg is
                 # likely to be the module params.
                 warnings.warn(
-                    "tensordict_out will be deprecated soon.",
+                    "tensordict_out will be deprecated in v0.6. "
+                    "Make sure you have removed any such arg by then.",
                     category=DeprecationWarning,
                 )
             if len(args):
diff --git a/tensordict/nn/probabilistic.py b/tensordict/nn/probabilistic.py
index bf8ac049c..1066f4a3c 100644
--- a/tensordict/nn/probabilistic.py
+++ b/tensordict/nn/probabilistic.py
@@ -64,7 +64,7 @@ def _insert_interaction_mode_deprecation_warning(
 ) -> Callable[[str, Warning, int], None]:
     return warn(
         (
-            f"{prefix}interaction_mode is deprecated for naming clarity. "
+            f"{prefix}interaction_mode is deprecated for naming clarity and will be removed in v0.6. "
             f"Please use {prefix}interaction_type with InteractionType enum instead."
         ),
         DeprecationWarning,
@@ -413,7 +413,7 @@ def log_prob(self, tensordict):
     @property
     def SAMPLE_LOG_PROB_KEY(self):
         warnings.warn(
-            "SAMPLE_LOG_PROB_KEY will be deprecated soon."
+            "SAMPLE_LOG_PROB_KEY will be deprecated in v0.6."
             "Use 'obj.log_prob_key' instead",
             category=DeprecationWarning,
         )
diff --git a/tensordict/prototype/__init__.py b/tensordict/prototype/__init__.py
index 352ee63d0..a30550412 100644
--- a/tensordict/prototype/__init__.py
+++ b/tensordict/prototype/__init__.py
@@ -3,10 +3,7 @@
 # This source code is licensed under the MIT license found in the
 # LICENSE file in the root directory of this source tree.
 from tensordict.prototype.fx import symbolic_trace
-from tensordict.prototype.tensorclass import is_tensorclass, tensorclass
 
 __all__ = [
-    "is_tensorclass",
     "symbolic_trace",
-    "tensorclass",
 ]
diff --git a/tensordict/prototype/tensorclass.py b/tensordict/prototype/tensorclass.py
deleted file mode 100644
index bf1b958e5..000000000
--- a/tensordict/prototype/tensorclass.py
+++ /dev/null
@@ -1,23 +0,0 @@
-import warnings
-from functools import wraps
-
-from tensordict._td import is_tensorclass as is_tensorclass_true  # no_qa
-from tensordict.tensorclass import tensorclass as tensorclass_true  # no_qa
-
-
-@wraps(tensorclass_true)
-def tensorclass(*args, **kwargs):  # noqa: D103
-    warnings.warn(
-        "tensorclass is not a prototype anymore and can be imported directly from tensordict root.",
-        category=DeprecationWarning,
-    )
-    return tensorclass_true(*args, **kwargs)
-
-
-@wraps(is_tensorclass_true)
-def is_tensorclass(*args, **kwargs):  # noqa: D103
-    warnings.warn(
-        "is_tensorclass is not a prototype anymore and can be imported directly from tensordict root.",
-        category=DeprecationWarning,
-    )
-    return is_tensorclass_true(*args, **kwargs)
diff --git a/tensordict/utils.py b/tensordict/utils.py
index 6a1ff2db7..554dbecd3 100644
--- a/tensordict/utils.py
+++ b/tensordict/utils.py
@@ -715,7 +715,7 @@ def _get_item(tensor: Tensor, index: IndexType) -> Tensor:
                 if index.dtype is torch.bool:
                     warnings.warn(
                         "Indexing a tensor with a nested list of boolean values is "
-                        "going to be deprecated as this functionality is not supported "
+                        "going to be deprecated in v0.6 as this functionality is not supported "
                         f"by PyTorch. (follows error: {err})",
                         category=DeprecationWarning,
                     )
diff --git a/tutorials/sphinx_tuto/tensorclass_fashion.py b/tutorials/sphinx_tuto/tensorclass_fashion.py
index b69d4a60c..bfcd6265b 100644
--- a/tutorials/sphinx_tuto/tensorclass_fashion.py
+++ b/tutorials/sphinx_tuto/tensorclass_fashion.py
@@ -16,8 +16,7 @@
 import torch
 import torch.nn as nn
 
-from tensordict import MemoryMappedTensor
-from tensordict.prototype import tensorclass
+from tensordict import MemoryMappedTensor, tensorclass
 from torch.utils.data import DataLoader
 from torchvision import datasets
 from torchvision.transforms import ToTensor
diff --git a/tutorials/sphinx_tuto/tensorclass_imagenet.py b/tutorials/sphinx_tuto/tensorclass_imagenet.py
index 5899748f1..ba3023189 100644
--- a/tutorials/sphinx_tuto/tensorclass_imagenet.py
+++ b/tutorials/sphinx_tuto/tensorclass_imagenet.py
@@ -37,8 +37,7 @@
 import torch.nn as nn
 import tqdm
 
-from tensordict import MemoryMappedTensor
-from tensordict.prototype import tensorclass
+from tensordict import MemoryMappedTensor, tensorclass
 from tensordict.utils import strtobool
 from torch.utils.data import DataLoader
 from torchvision import datasets, transforms
diff --git a/version.txt b/version.txt
index 1d0ba9ea1..8f0916f76 100644
--- a/version.txt
+++ b/version.txt
@@ -1 +1 @@
-0.4.0
+0.5.0