Nah just filling in a bunch of hyperparam results isn't interesting — you need to actually try to win! I mean try to get onto this leaderboard (doesn't matter if you make it or not, but you'll see what actually leads to fast accurate training): https://
github.com/fastai/imagene
tte?tab=readme-ov-file#imagenette-leaderboard
…
@jeremyphoward
-
Win ImageNette leaderboard: practical hyperparameter training optimization
By
–
-
Imagenette Speed-Running: Building Intuition in ML Optimization
By
–
You should maybe try doing some imagenette speed-running — it's a good way to get an intuitive understanding of these issues.
-
Batch Size Impact on Model Training Updates and Results
By
–
We covered this in some of our earlier courses – lower batch sizes provide more updates, which should give better results for a fixed # epochs.
-
Chart Misinterpretation: Fixed Index vs Year-over-Year Change
By
–
The poster misread the chart. They thought it shows a change in velocity. But that would only be true if this was YoY % change. It’s actually to a fixed 2025 index, so it doesn’t show a change in velocity.
-
Deep Learning Accessibility Prediction: A Decade Later Review
By
–
Just came across this old GigaOM article. I guessed in 2013 it might take ~10 years until "packages like word2vec can make deep learning even for relatively unsophisticated users". Not bad, I think! 😀 https://
web.archive.org/web/2022070705
2644/https://old.gigaom.com/2013/08/16/were-on-the-cusp-of-deep-learning-for-the-masses-you-can-thank-google-later/
… -
Nomad Coder Side Gig Opportunities for Developers
By
–
If you're interested in doing the nomad coder side gig thing, then listen to this man please!:
-
Natural Flow Over Control in AI-Assisted Development
By
–
When you're working well with AI, you have it gently helping you go with your natural flow. It's *your* nature that is the foundation, not the AI's. "Vibe coding" today isn't like this at all. It's a battle against the AI to control it under your will.
-
Wu Wei: Working with AI Without Forcing
By
–
Also there's the later "therefore" part which corresponds more closely to "the way of code": "是以聖人處無為之事，行不言之教". "無為" is *not* "without labor". It's "without forcing things". I think *this* can teach us a lot about how to work with AI in fact.
-
CIFAR-10 Speed Run Championship Experience
By
–
at one point I was the cifar 10 speed run champ, but that was a long time ago!
-
xAI Expanding Capacity Infrastructure Updates
By
–
actually maybe that's not right – I know xai is adding a lot of capacity so maybe I'm out of date there…