fix links that should be references

Author: mercury0100

docs/source/learn/core_notebooks/dims_module.ipynb

@@ -34,9 +34,9 @@
     ":::\n",
     "\n",
     "**Related API reference:**  \n",
-    "- [Dims — distributions](https://www.pymc.io/projects/docs/en/stable/api/dims/distributions.html)  \n",
-    "- [Dims — transforms](https://www.pymc.io/projects/docs/en/stable/api/dims/transforms.html)  \n",
-    "- [Dims — math](https://www.pymc.io/projects/docs/en/stable/api/dims/math.html)\n",
+    "- {doc}`Dims — distributions </api/dims/distributions>`\n",
+    "- {doc}`Dims — transforms </api/dims/transforms>`\n",
+    "- {doc}`Dims — math </api/dims/math>`\n",
     "\n",
     "## A simple example\n",
     "\n",
@@ -315,13 +315,13 @@
     "\n",
     "There are some notable differences:\n",
     "\n",
-    "1. [`ZeroSumNormal`](https://www.pymc.io/projects/docs/en/stable/api/dims/distributions.html#pymc.distributions.ZeroSumNormal) takes a `core_dims` argument instead of `n_zerosum_axes`. This tells PyMC which of the `dims` that define the distribution are constrained to be zero-summed. All distributions that take non-scalar parameters now require a `core_dims` argument. Previously, they were assumed to be right-aligned by the user (see {doc}`dimensionality`). Now you don't have to worry about the order of dimensions in your model, only their meaning.\n",
+    "1. {class}`~pymc.distributions.ZeroSumNormal` takes a `core_dims` argument instead of `n_zerosum_axes`. This tells PyMC which of the `dims` that define the distribution are constrained to be zero-summed. All distributions that take non-scalar parameters now require a `core_dims` argument. Previously, they were assumed to be right-aligned by the user (see {doc}`dimensionality`). Now you don't have to worry about the order of dimensions in your model, only their meaning.\n",
     "\n",
     "2. The `trial_preference` computation aligns dimensions for broadcasting automatically. Note we use {func}`pymc.dims.Deterministic` (not {func}`pymc.Deterministic`), which automatically propagates the `dims` to the model object.\n",
     "\n",
     "3. The `softmax` operation specifies the `dim` argument, not the positional axis. *Note:* The parameter is called `dim` and not `core_dims` because we try to stay as close as possible to the Xarray API (which uses `dim` throughout). We make an exception for distributions because they already have the `dims` argument.\n",
     "\n",
-    "4. [`Categorical`](https://www.pymc.io/projects/docs/en/stable/api/dims/distributions.html#pymc.distributions.Categorical) observed variables, like `ZeroSumNormal`, require a `core_dims` argument to specify which dimension corresponds to the probability vector. Previously, it was necessary to place this dimension explicitly on the rightmost axis — not anymore!\n",
+    "4. {class}`~pymc.distributions.Categorical` observed variables, like `ZeroSumNormal`, require a `core_dims` argument to specify which dimension corresponds to the probability vector. Previously, it was necessary to place this dimension explicitly on the rightmost axis — not anymore!\n",
     "\n",
     "5. Even when `dims` are not specified for either `trial_preference` or `response`, PyMC automatically infers them.\n",
     "\n",
@@ -891,7 +891,7 @@
     "\n",
     " * All vector arguments (and observed values) must have known dims. An error is raised otherwise.\n",
     "\n",
-    " * Distributions with non-scalar inputs will require a `core_dims` argument. The meaning of the `core_dims` argument will be denoted in the docstrings of each distribution. For example, for the [`MvNormal`](https://www.pymc.io/projects/docs/en/stable/api/dims/generated/pymc.dims.MvNormal.html#pymc.dims.MvNormal), the `core_dims` are the two dimensions of the covariance matrix, one (and only one) of which must also be present in the mean parameter. The shared `core_dim` is the one that persists in the output. Sometimes the order of `core_dims` will be important!\n",
+    " * Distributions with non-scalar inputs will require a `core_dims` argument. The meaning of the `core_dims` argument will be denoted in the docstrings of each distribution. For example, for the {class}`~pymc.distributions.MvNormal`, the `core_dims` are the two dimensions of the covariance matrix, one (and only one) of which must also be present in the mean parameter. The shared `core_dim` is the one that persists in the output. Sometimes the order of `core_dims` will be important!\n",
     "\n",
     " * `dims` accept ellipsis, and variables are transposed to match the user-specified `dims` argument.\n",
     "\n",