Edit on GitHub


Make an in-code checkpoint.

def make_checkpoint()


from dvc.api import make_checkpoint

while True:
    # ... write a stage output


To track successive steps in a longer experiment, you can write your code so it registers checkpoints with DVC during runtime (similar to a logger). This function can be called by the code in stages executed by dvc exp run.

make_checkpoint() does the following:

  1. Check that the DVC_ROOT env var is set. It means this code is being executed via dvc exp run, and it contains the location to the correct .dvc/ directory for this experiment (which can vary when -j is used)
  2. Creates an empty $DVC_ROOT/.dvc/tmp/DVC_CHECKPOINT signal file so DVC knows that a checkpoint should be captured now.
  3. Blocks the execution of any further code (that changes the state of the workspace) until the signal file is deleted, which means that DVC has finished caching all the data, calculating hashes, etc. (see dvc commit).

๐Ÿ’ก Note that for non-Python code, the way to register checkpoints with DVC is to implement the steps above yourself.

Note that the stage definition in dvc.yaml should contain at least one output with the checkpoint: true value set for DVC to register its checkpoints. See Checkpoints for details.

Example: Every 100th iteration

Let's consider the following dvc.yaml file:

    cmd: python iterate.py
      - int.txt:
          checkpoint: true

The code in iterate.py will execute continuously increment an integer number saved in int.txt (starting at 0). At 0 and every 100 loops, it makes a checkpoint for dvc experiments:

import os

from dvc.api import make_checkpoint

while True:
    if os.path.exists("int.txt"):
        with open("int.txt", "r") as fd:
            i_ = int(fd.read()) + 1
        i_ = 0

    # ... do something meaningful

    with open("int.txt", "w") as fd:

    if i_ % 100 == 0:

Since checkpoint outputs in effect implement a circular dependency, dvc repro does not support running this stage. Let's execute the stage with dvc exp run instead, and interrupt the process manually moments later:

$ dvc exp run
Running stage 'every100':
> python iterate.py
Generating lock file 'dvc.lock'
Updating lock file 'dvc.lock'
Checkpoint experiment iteration 'd832784'.
Updating lock file 'dvc.lock'
Checkpoint experiment iteration '6f5009b'.
Updating lock file 'dvc.lock'
Checkpoint experiment iteration '75ff5e0'.

Reproduced experiment(s): exp-8a3bd
Experiment results have been applied to your workspace.

In this example we killed the process (with Ctrl + C) after 3 checkpoints (at 0, 100, and 200 i_). The cache will contain those 3 versions of int.txt.

$ cat int.txt
$ ls .dvc/cache
36  cf  f8

dvc exp show will display these checkpoints as an experiment branch:

$ dvc exp show
โ”ƒ Experiment    โ”ƒ Created      โ”ƒ
โ”‚ workspace     โ”‚ -            โ”‚
โ”‚ master        โ”‚ Feb 10, 2021 โ”‚
โ”‚ โ”‚ โ•“ exp-8a3bd โ”‚ 02:07 PM     โ”‚
โ”‚ โ”‚ โ•Ÿ 75ff5e0   โ”‚ 01:54 PM     โ”‚
โ”‚ โ”‚ โ•Ÿ 6f5009b   โ”‚ 01:54 PM     โ”‚
โ”‚ โ”œโ”€โ•จ d832784   โ”‚ 01:54 PM     โ”‚
# Press q to exit this screen.

If we use dvc exp run again, the process will start from 200 (since that's what the workspace reflects).

See dvc experiments for more details on managing experiments.


๐Ÿ› Found an issue? Let us know! Or fix it:

Edit on GitHub

โ“ Have a question? Join our chat, we will help you:

Discord Chat