Skip to content

Commit

Permalink
Hardware Auto-Setup for Examples (#22319)
Browse files Browse the repository at this point in the history
* 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
  • Loading branch information
dongreenberg authored Mar 27, 2023
1 parent 738944c commit f02e3a2
Show file tree
Hide file tree
Showing 2 changed files with 107 additions and 0 deletions.
38 changes: 38 additions & 0 deletions examples/README.md
Original file line number Diff line number Diff line change
Expand Up @@ -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.
69 changes: 69 additions & 0 deletions examples/run_on_remote.py
Original file line number Diff line number Diff line change
@@ -0,0 +1,69 @@
#!/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)

0 comments on commit f02e3a2

Please sign in to comment.