Hardware Auto-Setup for Examples (#22319)

* Add initial remote hardware auto-setup docs

* Fix a few typos and clarify some language

* Add missing dependency

* Update self-hosted launch script with Sylvain's comments.

* Formatting.

* Trigger CI

* Style

examples/README.md

@@ -94,3 +94,41 @@ Alternatively, you can switch your cloned 🤗 Transformers to a specific versio
 git checkout tags/v3.5.1
 ```
 and run the example command as usual afterward.
+
+## Running the Examples on Remote Hardware with Auto-Setup
+
+[run_on_remote.py](./run_on_remote.py) is a script that launches any example on remote self-hosted hardware, 
+with automatic hardware and environment setup. It uses [Runhouse](https://github.com/run-house/runhouse) to launch 
+on self-hosted hardware (e.g. in your own cloud account or on-premise cluster) but there are other options 
+for running remotely as well. You can easily customize the example used, command line arguments, dependencies, 
+and type of compute hardware, and then run the script to automatically launch the example.
+
+You can refer to 
+[hardware setup](https://runhouse-docs.readthedocs-hosted.com/en/main/rh_primitives/cluster.html#hardware-setup)
+for more information about hardware and dependency setup with Runhouse, or this
+[Colab tutorial](https://colab.research.google.com/drive/1sh_aNQzJX5BKAdNeXthTNGxKz7sM9VPc) for a more in-depth 
+walkthrough.
+
+You can run the script with the following commands:
+
+```bash
+# First install runhouse:
+pip install runhouse
+
+# For an on-demand V100 with whichever cloud provider you have configured:
+python run_on_remote.py \
+    --example pytorch/text-generation/run_generation.py \
+    --model_type=gpt2 \
+    --model_name_or_path=gpt2 \
+    --prompt "I am a language model and"
+
+# For byo (bring your own) cluster:
+python run_on_remote.py --host <cluster_ip> --user <ssh_user> --key_path <ssh_key_path> \
+  --example <example> <args>
+
+# For on-demand instances
+python run_on_remote.py --instance <instance> --provider <provider> \
+  --example <example> <args>
+```
+
+You can also adapt the script to your own needs.
\ No newline at end of file

examples/run_on_remote.py

+#!/usr/bin/env python
+# coding=utf-8
+# Copyright 2021 The HuggingFace Inc. team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import argparse
+import shlex
+import runhouse as rh
+
+if __name__ == "__main__":
+    # Refer to https://runhouse-docs.readthedocs-hosted.com/en/main/rh_primitives/cluster.html#hardware-setup for cloud access
+    # setup instructions, if using on-demand hardware
+
+    # If user passes --user <user> --host <host> --key_path <key_path> <example> <args>, fill them in as BYO cluster
+    # If user passes --instance <instance> --provider <provider> <example> <args>, fill them in as on-demand cluster
+    # Throw an error if user passes both BYO and on-demand cluster args
+    # Otherwise, use default values
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--user", type=str, default="ubuntu")
+    parser.add_argument("--host", type=str, default="localhost")
+    parser.add_argument("--key_path", type=str, default=None)
+    parser.add_argument("--instance", type=str, default="V100:1")
+    parser.add_argument("--provider", type=str, default="cheapest")
+    parser.add_argument("--use_spot", type=bool, default=False)
+    parser.add_argument("--example", type=str, default="pytorch/text-generation/run_generation.py")
+    args, unknown = parser.parse_known_args()
+    if args.host != "localhost":
+        if args.instance != "V100:1" or args.provider != "cheapest":
+            raise ValueError("Cannot specify both BYO and on-demand cluster args")
+        cluster = rh.cluster(
+            name="rh-cluster", ips=[args.host], ssh_creds={"ssh_user": args.user, "ssh_private_key": args.key_path}
+        )
+    else:
+        cluster = rh.cluster(
+            name="rh-cluster", instance_type=args.instance, provider=args.provider, use_spot=args.use_spot
+        )
+    example_dir = args.example.rsplit("/", 1)[0]
+
+    # Set up remote environment
+    cluster.install_packages(["pip:./"])  # Installs transformers from local source
+    # Note transformers is copied into the home directory on the remote machine, so we can install from there
+    cluster.run([f"pip install -r transformers/examples/{example_dir}/requirements.txt"])
+    cluster.run(["pip install torch --upgrade --extra-index-url https://download.pytorch.org/whl/cu117"])
+
+    # Run example. You can bypass the CLI wrapper and paste your own code here.
+    cluster.run([f'python transformers/examples/{args.example} {" ".join(shlex.quote(arg) for arg in unknown)}'])
+
+    # Alternatively, we can just import and run a training function (especially if there's no wrapper CLI):
+    # from my_script... import train
+    # reqs = ['pip:./', 'torch', 'datasets', 'accelerate', 'evaluate', 'tqdm', 'scipy', 'scikit-learn', 'tensorboard']
+    # launch_train_gpu = rh.function(fn=train,
+    #                                system=gpu,
+    #                                reqs=reqs,
+    #                                name='train_bert_glue')
+    #
+    # We can pass in arguments just like we would to a function:
+    # launch_train_gpu(num_epochs = 3, lr = 2e-5, seed = 42, batch_size = 16
+    #                  stream_logs=True)