Hey team, Aim 3.5 featuring TensorBoard logs support is now available!! We are on a mission to democratize MLOps tools. Thanks to the awesome Aim community for the help and contributions.
Here is what’s new:
- Aim supports TensorBoard logs
- Track system params, CLI ,Env, Executable, Git, Installed packages
- Matplotlib figure tracking and visualization
- Ability move runs between Aim repos
TensorBoard logs support
It’s been one of the most highly requested features by the Aim community. This feature was available for Aim 2.x and folks used to love this it. However, we had to drop it due to the backend changes.
Now its back! Better than it was before and this is how it works:
$ cd /path/to/.aim $ aim convert tf --logdir ~/tensorflow/logdir
This command will scan then convert the
image type logs from your directory into Aim runs.
Read more about how it works here.
Tracking Env info, git info with Aim
Tracking your ENV variables, CLI argument, git info, etc could be a lot of details to care about.
Now there is a way to enable Aim to track the environment info automatically and they will be available as params.
It takes a small tweak to enable that:
run = Run(log_system_params=True)
Then once tracked, you can search experiments based on these values too:
run.__system_params.git_info.branch == 'feature/testing'
More on this feature find out here.
A special UI for these tracked data is to be shipped with the next version.
Tracking Matplotlib figures
Starting Aim 3.5 you can also track Matplotlib figures with Aim. During research (especially with Jupyter Notebooks) Matplotlib is very helpful in rendering intermediate images for analysis.
Now you can track all such figures on Aim (both as Matplotlib figure and as an image). When tracking as an image, you can query and compare them too at scale on the Images Explorer.
Moving Runs between Aim repos
We have added a CLI command to move runs between folders.
This will allow to easily move your best runs from a draft scratch project to your main one with one command. Here is how it works:
aim mv --destination /new/path/to/.aim <my_run_hash_1> ...
For more info, check the runs subcommand docs here.
We have been incredibly lucky to get help and contributions from the amazing Aim community. It’s humbling and inspiring.
And don’t forget to leave Aim a star on GitHub for support 🙌.