Specifically, it's a genius way to store large files in git repos directly on any object storage without custom application servers like git-lfs or rewriting git from scratch...
At DagsHub [0], we've integrated directly with DVC for a looong time, so teams can use it with added features like visualizing and labeling datasets managing and models, running experiments collaboratively, and tracking everything (code, data, models, etc.) all in one place.
Just wanted to share that for those already using or considering DVC—there are some options to use it as a building block in a more end-to-end toolchain.
Hi! (DagsHub Co-Founder here) We worked with the Colab team to create DagsHub Storage Buckets – It's S3-compatible, user-friendly, and address some common pains we've seen in data management, specifically in working with Drive in Colab.(It also works great outside of Colab btw)Would love it if you could check it out and let me know what you think...
We were working on a project with a domain expert recently and felt there could be learnings that are relevant to the community we could share.In our case, the domain expert was a dentist who reached out to us to help him create a machine learning model that would segment teeth in panoramic X-rays. He had some data pre-labeled, but the vast majority of his dataset was unlabeled.Since labeling...
Hey HN! I'm working on a talk/blog on a related topic and was curious to see what the case is for people here. I think each case is a bit different so I don't want to constrain people too much.If you deploy ML to production in your group/team/company – what does production mean for you?Examples: - "We run a model once a week that predicts some stuff and stores it in a table, then the customer...
As Data Scientists, we sometimes have interactive outputs from Jupyter notebook cells. Up to now, most places you'd save notebooks, e.g. GitHub/GitLab, only showed a static image instead of the original rich output, and you'd have to open the notebook in Colab, or a similar tool in order to interact with your outputs.
This limits sharability and creates a lossy experience.
We wanted to fix this, so now when you commit a notebook to DagsHub (or connect a GitHub project with such a notebook), you see the interactive outputs as they should be – interactive.
This means that things like EDA and dashboarding tools are much cooler and more useful.
Oh, and you don't need to change your code or download or import anything.
This might be a semantic argument, but what you describe is replication. Imagine every scientist would say "In my experiment, I perfectly predicted this and that, oh but no one else would ever be able to run that experiment again, so just trust me, ok?"
Replication/reproducibility isn't about logging every configuration, machine or particle. It's about being able to run the same method and get the same result. If that isn't the case, how do we know the predictions are correct?
There are a lot of great articles about measuring the performance of data annotator’s agreement on labels, like this one https://towardsdatascience.com/the-definite-guide-for-creating-an-academic-level-dataset-with-industry-requirements-and-6db446a26cb2.I see mentions in a lot of places of Cohen’s Kappa/Krippendorf’s alpha, Fleischer’s Kappa, Comparing to predefined ground truth, etc.If you’re...
TL;DR: We announced that we'd support the Reproducibility Challenge $500 per paper reproduced, and we're announcing the award winners for the Spring '21 edition, as well as our support for the Fall '21 edition of the challenge. Check out the awesome papers below---Hey HN! Creator of DagsHub here. We really care about reproducibility. That's why, a while back, we announced our support for the...
Hey HN! For Hacktoberfest this year, we wanted to do something that was geared towards the ML community, and we decided to create an open-source catalog of audio datasets.The response has been truly amazing! We received 40 dataset contributions, which are now publicly available, and viewable on DagsHub. They cover various tasks, languages, and sizes, and you can use them all for your...
Hey HN community! Creator of DagsHub here. Hacktoberfest 2021 is well underway, but there's still a lot of time left, and I was missing some opportunities to contribute to the community on the ML/DS fronts.We've decided to support Hacktoberfest by creating an open-source catalog of datasets in the audio domain. The idea is to have a bunch of audio datasets, which will be completely open-source,...
Hey, HN! I'm one of the creators of DAGsHub(https://dagshub.com). As some of you might know, there's an ML Reproducibility Challenge(https://paperswithcode.com/rc2020) happening now, organized by Papers with Code, with some of the leading ML conferences. If you haven't heard about it, or didn't consider joining, I think that it provides an awesome opportunity for a number of reasons:- It's a...
Hey HN, Creator of DAGsHub here!I wanted to share something cool that we've been working on. You might've heard about DVC (dvc.org) and MLflow (mlflow.org). They are open-source projects for ML that are widely adopted, each for its own specialty.DVC is great for versioning your data. MLflow is used for a bunch of stuff (it's actually multiple tools combined into one) – but mostly for...
Also, thanks! I just discovered unDraw thanks to this!
As many others have said there are already great UIs to help you manage these things retroactively, but there is probably no good substitute to properly planning your work and executing on that plan (if order is your top priority)