HackerTrans
TopNewTrendsCommentsPastAskShowJobs

thecellardoor

no profile record

Submissions

Show HN: FastDS – Open-Source Machine Learning Version Control. Fast and Easy

github.com
1 points·by thecellardoor·vor 5 Jahren·0 comments

Millisecond Latency MLOps and the CERN for ML: An Interview with HuggingFace CTO

youtube.com
2 points·by thecellardoor·vor 5 Jahren·0 comments

A Comparison of ML Experiment Tracking Tools

dagshub.com
2 points·by thecellardoor·vor 5 Jahren·1 comments

A solution to the dark matter problem?

phys.org
3 points·by thecellardoor·vor 5 Jahren·2 comments

Stop treating data as static. Datasets should behave like Git repositories

dagshub.com
4 points·by thecellardoor·vor 5 Jahren·0 comments

comments

thecellardoor
·vor 5 Jahren·discuss
As someone who looks at GitHub accounts as a resume, I can tell you that improvement over time is amazing to see. Assuming it’s clear what’s most interesting or recent.
thecellardoor
·vor 5 Jahren·discuss
Seems like life goals
thecellardoor
·vor 5 Jahren·discuss
Does this support distributed nodes within the DAG or does it assume everything runs on a single node?
thecellardoor
·vor 5 Jahren·discuss
Does the Pseudocode example work for complicated functions too? Or is it limited to really basic stuff?

Cool ideas
thecellardoor
·vor 5 Jahren·discuss
There are so many options and this post tries to make some sense of the mess.
thecellardoor
·vor 5 Jahren·discuss
This seems really interesting to me, but wanted a critical second opinion. Seems like a much more elegant solution than the alternatives
thecellardoor
·vor 5 Jahren·discuss
This is oddly addictive. Just typing in stuff and see what colors it gives me :)
thecellardoor
·vor 6 Jahren·discuss
I think that by definition this is a tradeoff. Most times you talk to data scientists that want a fully automated end-2-end solution, that doesn't require they change anything about their current workflow, and that any future modifications to their workflow would be supported as well.

That is magical thinking. I prefer best of breed solutions that integrate nicely with other best of breed solutions every day. That way if a tool doesn't suit you tomorrow, you can relatively easily swap it out for something better
thecellardoor
·vor 6 Jahren·discuss
Congrats on the launch. To me this feels like DVC but with a slightly more convenient python API, and without the pipeline which might be really great. How are you organizing experiments though? Without Git it seems too easy to have experiments become unusable. And filtering according to accuracy is not really a solution..
thecellardoor
·vor 6 Jahren·discuss
Looks really interesting! How do you connect the data and models in my project to the code, so you know what to merge?

Also, any way to easily connect to an existing project on GitHub/GitLab?