Focus on machine learning, NOT infrastructure.
From the creators of PyTorch Lightning.

---

Website • Runs • Interactive Sessions • Docs • Run Examples • Join our Slack • PyTorch Lightning

---

Grid.AI

Grid.AI lets you seamlessly train hundreds of machine learning models on the cloud from your laptop.

Note: This repository will host the opensource components of Grid once they are ready. In the meantime please use this to request features and to get help.

---

Grid Philosophy

Grid follows the same philosophy that we use for PyTorch Lightning:

• Let the user focus on machine learning not engineering.
• Remove all the infrastructure engineering so the user can focus.
• Focused on reproducibility

Key features

• Runs: Run (and sweep) any private or public Github repository.
• Sessions: Interactive machines (with multiple GPUs) optimized for development.
• Datastores: Low-latency, high-performance, auto-versioned datasets.
• Realtime costs + much more

Create a free account

Visit Grid.AI

Runs

Run (and sweep) any private or public Github repository on the cloud.

# install
pip install lightning-grid

# login
grid login

# clone the repo
git clone https://github.com/williamFalcon/cifar5.git
cd cifar5

# start run!
grid run --instance_type 8_v100_32gb project/lit_image_classifier.py --gpus 1 --learning_rate "uniform(1e-5, 1e-1, 8)"

The command above generates 8 experiments

(flash) ➜  project git:(master) grid status gifted-seahorse-83
✔ Fetching experiment status ... done!
┏━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━━━━━━━┳━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━┓
┃ Experiment              ┃                          Command ┃ Status ┃    Duration ┃ gpus ┃        learning_rate ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━━━━━━━╇━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━┩
│ gifted-seahorse-83-exp7 │ /project/lit_image_classifier.py │ queued │ 0d-00:00:21 │    1 │  0.07858944673974376 │
│ gifted-seahorse-83-exp6 │ /project/lit_image_classifier.py │ queued │ 0d-00:00:21 │    1 │   0.0492955135740864 │
│ gifted-seahorse-83-exp5 │ /project/lit_image_classifier.py │ queued │ 0d-00:00:21 │    1 │ 0.005299488022082154 │
│ gifted-seahorse-83-exp4 │ /project/lit_image_classifier.py │ queued │ 0d-00:00:21 │    1 │  0.09214628765569398 │
│ gifted-seahorse-83-exp3 │ /project/lit_image_classifier.py │ queued │ 0d-00:00:21 │    1 │  0.04288827588691595 │
│ gifted-seahorse-83-exp2 │ /project/lit_image_classifier.py │ queued │ 0d-00:00:21 │    1 │  0.02618640734272168 │
│ gifted-seahorse-83-exp1 │ /project/lit_image_classifier.py │ queued │ 0d-00:00:21 │    1 │  0.03269913908273296 │
│ gifted-seahorse-83-exp0 │ /project/lit_image_classifier.py │ queued │ 0d-00:00:21 │    1 │  0.06192848686653007 │
└─────────────────────────┴──────────────────────────────────┴────────┴─────────────┴──────┴──────────────────────┘

Interactive Sessions

Sessions allow you to develop models interactively with direct ssh, VSCode and Jupyter Notebook access already built-in.

Or via the CLI

grid session create --instance_type 8_v100_32gb

---

Run Examples

Vision

3D Image classification
Image classificatin (ImageNet)
Object detection (Coco)

NLP

GPT 10+ B Params
Text classification

Self-supervised learning

SIMCLR

---

Community

The Grid community is the place to chat about Deep Learning, machine learning and anything about scaling up production or research projects! Join our slack

Contribute to our documentation https://github.com/gridai/grid-docs

Asking for help

If you have any questions here are some things you can do:

1. Read the docs.
2. Start a discussion.
3. Create a Github issue.
4. Send us a message on slack

Enterprise Features

For enterprise features, upgrades and support, register your interest here