[Feature] TensorSpec.enumerate()

ghstack-source-id: 98d1bd5
Pull Request resolved: #2354

Author: Vincent Moens

test/test_specs.py

@@ -3816,6 +3816,53 @@ def test_non_tensor(self):
         assert not isinstance(non_tensor, MultiOneHot)
 
 
+class TestSpecEnumerate:
+    def test_discrete(self):
+        spec = DiscreteTensorSpec(n=5, shape=(3,))
+        assert (
+            spec.enumerate()
+            == torch.tensor([[0, 0, 0], [1, 1, 1], [2, 2, 2], [3, 3, 3], [4, 4, 4]])
+        ).all()
+
+    def test_one_hot(self):
+        spec = OneHotDiscreteTensorSpec(n=5, shape=(2, 5))
+        assert (
+            spec.enumerate()
+            == torch.tensor(
+                [
+                    [[1, 0, 0, 0, 0], [1, 0, 0, 0, 0]],
+                    [[0, 1, 0, 0, 0], [0, 1, 0, 0, 0]],
+                    [[0, 0, 1, 0, 0], [0, 0, 1, 0, 0]],
+                    [[0, 0, 0, 1, 0], [0, 0, 0, 1, 0]],
+                    [[0, 0, 0, 0, 1], [0, 0, 0, 0, 1]],
+                ],
+                dtype=torch.bool,
+            )
+        ).all()
+
+    def test_multi_discrete(self):
+        spec = MultiDiscreteTensorSpec([3, 4, 5], shape=(2, 3))
+        enum = spec.enumerate()
+        assert enum.shape == torch.Size([60, 2, 3])
+
+    def test_multi_onehot(self):
+        spec = MultiOneHotDiscreteTensorSpec([3, 4, 5], shape=(2, 12))
+        enum = spec.enumerate()
+        assert enum.shape == torch.Size([60, 2, 12])
+
+    def test_composite(self):
+        c = CompositeSpec(
+            {
+                "a": OneHotDiscreteTensorSpec(n=5, shape=(3, 5)),
+                ("b", "c"): DiscreteTensorSpec(n=4, shape=(3,)),
+            },
+            shape=[3],
+        )
+        c_enum = c.enumerate()
+        assert c_enum.shape == torch.Size((20, 3))
+        assert c_enum["b"].shape == torch.Size((20, 3))
+
+
 if __name__ == "__main__":
     args, unknown = argparse.ArgumentParser().parse_known_args()
     pytest.main([__file__, "--capture", "no", "--exitfirst"] + unknown)

torchrl/data/tensor_specs.py

@@ -834,6 +834,16 @@ def contains(self, item: torch.Tensor | TensorDictBase) -> bool:
         """
         return self.is_in(item)
 
+    @abc.abstractmethod
+    def enumerate(self):
+        """Returns all the samples that can be obtained from the TensorSpec.
+
+        The samples will be stacked along the first dimension.
+
+        This method is only implemented for discrete specs.
+        """
+        ...
+
     def project(
         self, val: torch.Tensor | TensorDictBase
     ) -> torch.Tensor | TensorDictBase:
@@ -1271,6 +1281,11 @@ def __eq__(self, other):
                 return False
         return True
 
+    def enumerate(self):
+        return torch.stack(
+            [spec.enumerate() for spec in self._specs], dim=self.stack_dim + 1
+        )
+
     def __len__(self):
         return self.shape[0]
 
@@ -1732,6 +1747,13 @@ def to_numpy(self, val: torch.Tensor, safe: bool = None) -> np.ndarray:
             return np.array(vals).reshape(tuple(val.shape))
         return val
 
+    def enumerate(self):
+        return (
+            torch.eye(self.n, dtype=self.dtype, device=self.device)
+            .expand(*self.shape, self.n)
+            .permute(-2, *range(self.ndimension() - 1), -1)
+        )
+
     def index(self, index: INDEX_TYPING, tensor_to_index: torch.Tensor) -> torch.Tensor:
         if not isinstance(index, torch.Tensor):
             raise ValueError(
@@ -2056,6 +2078,11 @@ def __init__(
             domain=domain,
         )
 
+    def enumerate(self):
+        raise NotImplementedError(
+            f"enumerate is not implemented for spec of class {type(self).__name__}."
+        )
+
     def __eq__(self, other):
         return (
             type(other) == type(self)
@@ -2375,6 +2402,9 @@ def __init__(
             shape=shape, space=None, device=device, dtype=dtype, domain=domain, **kwargs
         )
 
+    def enumerate(self):
+        raise NotImplementedError("Cannot enumerate a NonTensorSpec.")
+
     def to(self, dest: Union[torch.dtype, DEVICE_TYPING]) -> NonTensor:
         if isinstance(dest, torch.dtype):
             dest_dtype = dest
@@ -2611,6 +2641,9 @@ def is_in(self, val: torch.Tensor) -> bool:
     def _project(self, val: torch.Tensor) -> torch.Tensor:
         return torch.as_tensor(val, dtype=self.dtype).reshape(self.shape)
 
+    def enumerate(self):
+        raise NotImplementedError("enumerate cannot be called with continuous specs.")
+
     def expand(self, *shape):
         if len(shape) == 1 and isinstance(shape[0], (tuple, list, torch.Size)):
             shape = shape[0]
@@ -2775,6 +2808,18 @@ def __init__(
         )
         self.update_mask(mask)
 
+    def enumerate(self):
+        nvec = self.nvec
+        enum_disc = self.to_categorical_spec().enumerate()
+        enums = torch.cat(
+            [
+                torch.nn.functional.one_hot(enum_unb, nv).to(self.dtype)
+                for nv, enum_unb in zip(nvec, enum_disc.unbind(-1))
+            ],
+            -1,
+        )
+        return enums
+
     def update_mask(self, mask):
         """Sets a mask to prevent some of the possible outcomes when a sample is taken.
 
@@ -3208,6 +3253,12 @@ def __init__(
         )
         self.update_mask(mask)
 
+    def enumerate(self):
+        arange = torch.arange(self.n, dtype=self.dtype, device=self.device)
+        if self.ndim:
+            arange = arange.view(-1, *(1,) * self.ndim)
+        return arange.expand(self.n, *self.shape)
+
     @property
     def n(self):
         return self.space.n
@@ -3715,6 +3766,29 @@ def __init__(
         self.update_mask(mask)
         self.remove_singleton = remove_singleton
 
+    def enumerate(self):
+        if self.mask is not None:
+            raise RuntimeError(
+                "Cannot enumerate a masked TensorSpec. Submit an issue on github if this feature is requested."
+            )
+        if self.nvec._base.ndim == 1:
+            nvec = self.nvec._base
+        else:
+            # we have to use unique() to isolate the nvec
+            nvec = self.nvec.view(-1, self.nvec.shape[-1]).unique(dim=0).squeeze(0)
+            if nvec.ndim > 1:
+                raise ValueError(
+                    f"Cannot call enumerate on heterogeneous nvecs: unique nvecs={nvec}."
+                )
+        arange = torch.meshgrid(
+            *[torch.arange(n, device=self.device, dtype=self.dtype) for n in nvec],
+            indexing="ij",
+        )
+        arange = torch.stack([arange_.reshape(-1) for arange_ in arange], dim=-1)
+        arange = arange.view(arange.shape[0], *(1,) * (self.ndim - 1), self.shape[-1])
+        arange = arange.expand(arange.shape[0], *self.shape)
+        return arange
+
     def update_mask(self, mask):
         """Sets a mask to prevent some of the possible outcomes when a sample is taken.
 
@@ -3932,6 +4006,8 @@ def to_one_hot(
 
     def to_one_hot_spec(self) -> MultiOneHot:
         """Converts the spec to the equivalent one-hot spec."""
+        if self.ndim > 1:
+            return torch.stack([spec.to_one_hot_spec() for spec in self.unbind(0)])
         nvec = [_space.n for _space in self.space]
         return MultiOneHot(
             nvec,
@@ -4606,6 +4682,33 @@ def clone(self) -> Composite:
             shape=self.shape,
         )
 
+    def enumerate(self):
+        # We are going to use meshgrid to create samples of all the subspecs in here
+        #  but first let's get rid of the batch size, we'll put it back later
+        self_without_batch = self
+        while self_without_batch.ndim:
+            self_without_batch = self_without_batch[0]
+        samples = {key: spec.enumerate() for key, spec in self_without_batch.items()}
+        if samples:
+            idx_rep = torch.meshgrid(
+                *(torch.arange(s.shape[0]) for s in samples.values()), indexing="ij"
+            )
+            idx_rep = tuple(idx.reshape(-1) for idx in idx_rep)
+            samples = {
+                key: sample[idx]
+                for ((key, sample), idx) in zip(samples.items(), idx_rep)
+            }
+            samples = TensorDict(
+                samples, batch_size=idx_rep[0].shape[:1], device=self.device
+            )
+            # Expand
+            if self.ndim:
+                samples = samples.reshape(-1, *(1,) * self.ndim)
+                samples = samples.expand(samples.shape[0], *self.shape)
+        else:
+            samples = TensorDict(batch_size=self.shape, device=self.device)
+        return samples
+
     def empty(self):
         """Create a spec like self, but with no entries."""
         try:
@@ -4856,6 +4959,12 @@ def update(self, dict) -> None:
             self[key] = item
         return self
 
+    def enumerate(self):
+        dim = self.stack_dim
+        return LazyStackedTensorDict.maybe_dense_stack(
+            [spec.enumerate() for spec in self._specs], dim + 1
+        )
+
     def __eq__(self, other):
         if not isinstance(other, StackedComposite):
             return False
@@ -5150,7 +5259,7 @@ def rand(self, shape: torch.Size = None) -> TensorDictBase:
 
 
 @TensorSpec.implements_for_spec(torch.stack)
-def _stack_specs(list_of_spec, dim, out=None):
+def _stack_specs(list_of_spec, dim=0, out=None):
     if out is not None:
         raise NotImplementedError(
             "In-place spec modification is not a feature of torchrl, hence "
@@ -5187,7 +5296,7 @@ def _stack_specs(list_of_spec, dim, out=None):
 
 
 @Composite.implements_for_spec(torch.stack)
-def _stack_composite_specs(list_of_spec, dim, out=None):
+def _stack_composite_specs(list_of_spec, dim=0, out=None):
     if out is not None:
         raise NotImplementedError(
             "In-place spec modification is not a feature of torchrl, hence "