diff --git a/dev/run_tutorials.sh b/dev/run_tutorials.sh
new file mode 100644
index 00000000..1abf4207
--- /dev/null
+++ b/dev/run_tutorials.sh
@@ -0,0 +1,52 @@
+#!/usr/bin/bash
+# Copyright (c) Facebook, Inc. and its affiliates. All rights reserved.
+
+# This script is for running some of the tutorials using the nightly build in
+# an isolated environment. It is designed to be run in docker.
+
+# If you run this script in this directory with
+#   sudo docker run --runtime=nvidia -it --rm -v $PWD/../docs/tutorials:/notebooks -v $PWD:/loc pytorch/conda-cuda bash /loc/run_tutorials.sh | tee log.txt
+# it should execute some tutorials with the nightly build and resave them, and
+# save a log in the current directory.
+
+# We use nbconvert. runipy would be an alternative but it currently doesn't
+# work well with plotly.
+
+set -e
+
+conda init bash
+# shellcheck source=/dev/null
+source ~/.bashrc
+conda create -y -n myenv python=3.8 matplotlib ipython ipywidgets nbconvert
+conda activate myenv
+conda install -y -c conda-forge fvcore
+conda install -y -c pytorch pytorch=1.6.0 cudatoolkit=10.1 torchvision
+conda install -y -c pytorch3d-nightly pytorch3d
+pip install plotly scikit-image
+
+for notebook in /notebooks/*.ipynb
+do
+    name=$(basename "$notebook")
+
+    if [[ "$name" == "dataloaders_ShapeNetCore_R2N2.ipynb" ]]
+    then
+        #skip as data not easily available
+        continue
+    fi
+    if [[ "$name" == "render_densepose.ipynb" ]]
+    then
+        #skip as data not easily available
+        continue
+    fi
+
+    #comment the lines which install torch, torchvision and pytorch3d
+    sed -Ei '/(torchvision)|(pytorch3d)/ s/!pip/!#pip/' "$notebook"
+    #Don't let tqdm use widgets
+    sed -i 's/from tqdm.notebook import tqdm/from tqdm import tqdm/' "$notebook"
+
+    echo
+    echo "###   ###   ###"
+    echo "starting $name"
+    time jupyter nbconvert --to notebook --inplace --ExecutePreprocessor.kernel_name=python3 --execute "$notebook" || true
+    echo "ending $name"
+done