Update examples

Signed-off-by: Emmanuel Ferdman <emmanuelferdman@gmail.com>

Author: emmanuel-ferdman

docs/examples/bert.ipynb

@@ -12,7 +12,7 @@
     "\n",
     "We demonstrate how to train it (to ~60% accuracy), or how to load some pretrained weights (to ~80% accuracy).\n",
     "\n",
-    "This example is available as a Jupyter notebook [here](https://github.com/patrick-kidger/equinox/blob/main/examples/bert.ipynb).\n",
+    "This example is available as a Jupyter notebook [here](https://github.com/patrick-kidger/equinox/blob/main/docs/examples/bert.ipynb).\n",
     "\n",
     "Authors: [Jannis Bulian](mailto:jbulian@google.com), [Jasmijn Bastings](mailto:bastings@google.com)\n",
     "\n",
@@ -514,7 +514,7 @@
     "\n",
     "Here we initialize the model with the official parameters of Tiny-BERT, i.e. a 2-layer BERT model with a hidden size of 128.\n",
     "\n",
-    "In this case, we assume that we've downloaded the [checkpoint](https://github.com/patrick-kidger/equinox/blob/main/examples/bert_checkpoint.eqx) provided in the examples directory. Initializing the model with pretrained weights will give you much better training performance."
+    "In this case, we assume that we've downloaded the [checkpoint](https://github.com/patrick-kidger/equinox/blob/main/docs/examples/bert_checkpoint.eqx) provided in the examples directory. Initializing the model with pretrained weights will give you much better training performance."
    ]
   },
   {
@@ -556,7 +556,7 @@
    "outputs": [],
    "source": [
     "# Download the checkpoint from\n",
-    "# https://github.com/patrick-kidger/equinox/blob/main/examples/bert_checkpoint.eqx\n",
+    "# https://github.com/patrick-kidger/equinox/blob/main/docs/examples/bert_checkpoint.eqx\n",
     "classifier_chkpt = eqx.tree_deserialise_leaves(\"bert_checkpoint.eqx\", classifier)"
    ]
   },

docs/examples/frozen_layer.ipynb

@@ -9,7 +9,7 @@
     "\n",
     "In this example, we demonstrate how to only train some parameters and freeze the rest.\n",
     "\n",
-    "This example is available as a Jupyter notebook [here](https://github.com/patrick-kidger/equinox/blob/main/examples/frozen_layer.ipynb)."
+    "This example is available as a Jupyter notebook [here](https://github.com/patrick-kidger/equinox/blob/main/docs/examples/frozen_layer.ipynb)."
    ]
   },
   {

docs/examples/init_apply.ipynb

@@ -13,7 +13,7 @@
     "\n",
     "Equinox can be made to fit with this style very easily, like so.\n",
     "\n",
-    "This example is available as a Jupyter notebook [here](https://github.com/patrick-kidger/equinox/blob/main/examples/init_apply.ipynb)."
+    "This example is available as a Jupyter notebook [here](https://github.com/patrick-kidger/equinox/blob/main/docs/examples/init_apply.ipynb)."
    ]
   },
   {

docs/examples/mnist.ipynb

@@ -15,7 +15,7 @@
     "\n",
     "The JAX ecosystem is build around a number of libraries, that each do a single thing. So in addition to Equinox (for model building), this example also uses [Optax](https://github.com/deepmind/optax) to train the network, and [jaxtyping](https://github.com/google/jaxtyping) to provide type annotations.\n",
     "\n",
-    "This example is available as a Jupyter notebook [here](https://github.com/patrick-kidger/equinox/blob/main/examples/mnist.ipynb).\n",
+    "This example is available as a Jupyter notebook [here](https://github.com/patrick-kidger/equinox/blob/main/docs/examples/mnist.ipynb).\n",
     "\n",
     "!!! FAQ \"What's the difference between JAX and Equinox?\"\n",
     "\n",

docs/examples/score_based_diffusion.ipynb

@@ -27,7 +27,7 @@
     "\n",
     "- Uses an [MLP-Mixer](https://arxiv.org/abs/2105.01601) to parameterise the score model $s_\\theta$. ([See here](./unet.ipynb) for a U-Net implementation that could be substituted.)\n",
     "\n",
-    "This example is available as a Jupyter notebook [here](https://github.com/patrick-kidger/equinox/blob/main/examples/score_based_diffusion.ipynb).\n",
+    "This example is available as a Jupyter notebook [here](https://github.com/patrick-kidger/equinox/blob/main/docs/examples/score_based_diffusion.ipynb).\n",
     "\n",
     "!!! warning\n",
     "\n",

docs/examples/stateful.ipynb

@@ -13,7 +13,7 @@
     "\n",
     "See also the [stateful API reference](../api/nn/stateful.md).\n",
     "\n",
-    "This example is available as a Jupyter notebook [here](https://github.com/patrick-kidger/equinox/blob/main/examples/stateful.ipynb)."
+    "This example is available as a Jupyter notebook [here](https://github.com/patrick-kidger/equinox/blob/main/docs/examples/stateful.ipynb)."
    ]
   },
   {

docs/examples/train_rnn.ipynb

@@ -11,7 +11,7 @@
     "\n",
     "Here, we'll train an RNN to classify clockwise vs anticlockwise spirals.\n",
     "\n",
-    "This example is available as a Jupyter notebook [here](https://github.com/patrick-kidger/equinox/blob/main/examples/train_rnn.ipynb)."
+    "This example is available as a Jupyter notebook [here](https://github.com/patrick-kidger/equinox/blob/main/docs/examples/train_rnn.ipynb)."
    ]
   },
   {

docs/examples/unet.ipynb

@@ -11,7 +11,7 @@
     "\n",
     "This version is intended for use in a [score-based diffusion](score_based_diffusion.ipynb), so it accepts a `t` argument.\n",
     "\n",
-    "This example is available as a Jupyter notebook [here](https://github.com/patrick-kidger/equinox/blob/main/examples/unet.ipynb)\n",
+    "This example is available as a Jupyter notebook [here](https://github.com/patrick-kidger/equinox/blob/main/docs/examples/unet.ipynb)\n",
     "\n",
     "Author: Ben Walker ([https://github.com/Benjamin-Walker](https://github.com/Benjamin-Walker))"
    ]