Yes, SkyPilot allows you to define a sequence of spot tasks that run one after another through its Spot Pipeline feature. This is particularly useful for running a series of jobs that depend on each other, such as training a model followed by running inference on it. The Spot Pipeline feature enables you to specify different resource requirements for each task, optimizing resource utilization and cost savings while simplifying task management. To define a sequence of spot tasks, you can specify the sequence in a YAML file. Below is an example of how to define a pipeline with two tasks, train and eval, which are executed sequentially:

name: pipeline

---

name: train

resources:
  accelerators: V100:8

file_mounts:
  /checkpoint:
    name: train-eval # NOTE: Fill in your bucket name
    mode: MOUNT

setup: |
  echo setup for training

run: |
  echo run for training
  echo save checkpoints to /checkpoint

---

name: eval

resources:
  accelerators: T4:1

file_mounts:
  /checkpoint:
    name: train-eval # NOTE: Fill in your bucket name
    mode: MOUNT

setup: |
  echo setup for eval

run: |
  echo load trained model from /checkpoint
  echo eval model on test set

To submit this pipeline, you would use the command sky spot launch -n pipeline pipeline.yaml, and you can monitor the status of the pipeline with sky spot queue or sky spot dashboard.