Skypilot + Verda Cloud training examples

train.yml

#
# Example job to run on Verda Cloud (formerly DataCrunch).
# This will launch a cluster which will be autodeleted in 120 minutes and will run the training job on it.
# 
# $ mkdir -p ${HOME}/.verda 
# $ echo { "client_id": "YOUR_CLIENT_ID", "client_secret": "YOUR_CLIENT_SECRET" } > ${HOME}/.verda/config.json
# $ pip install skypilot[verda]
# $ sky launch -i 120 train.yaml
#
name: minGPT-ddp

resources:
    cpus: 4+
    # Use A100 80GB x 8 node from Verda Cloud
    accelerators: A100-80GB:8

run: |
    set -e
    git clone --depth 1 https://github.com/pytorch/examples || true
    cd examples/distributed/minGPT-ddp
    git pull
    uv venv --python 3.11
    uv pip install -r requirements.txt "numpy<2" "torch" "torchvision" "torchrun" --extra-index-url https://download.pytorch.org/whl/cu126
    export LOGLEVEL=INFO
    echo "Starting minGPT-ddp training"
    uv run torchrun --nproc_per_node=$SKYPILOT_NUM_GPUS_PER_NODE mingpt/main.py

Example job to run on Verda Cloud (formerly DataCrunch).

This will launch a cluster which will be autodeleted in 120 minutes and will run the training job on it.

mkdir -p ${HOME}/.verda 
echo { "client_id": "YOUR_CLIENT_ID", "client_secret": "YOUR_CLIENT_SECRET" } > ${HOME}/.verda/config.json
git clone https://gist.github.com/huksley/56e9dbbbb1dba9dd947f6efe305c8722 skyverda
cd skyverda
uv venv --python 3.11 
uv pip install skypilot[verda]
uv run sky launch train.yml