DEV Community

Rahul Kumar profile picture

Rahul Kumar

ML engineer, on-device AI, open-source contributor. I measure what actually happens, not what the label says: in open source, in trading research, on one GPU. Real numbers, failures left in.

Keras to TFLite on Android: wrong normalization and a backwards top-5

Keras to TFLite on Android: wrong normalization and a backwards top-5

Comments
4 min read

Want to connect with Rahul Kumar?

Create an account to connect with Rahul Kumar. You can also sign in below to proceed if you already have an account.

Already have an account? Sign in
Two Android ML tests that could never fail, and a TFLite check that can

Two Android ML tests that could never fail, and a TFLite check that can

Comments
5 min read
Why a Keras model fails to convert to TFLite: 11 known cases, 5 traps

Why a Keras model fails to convert to TFLite: 11 known cases, 5 traps

Comments
6 min read
A hot-expert cache made a 181 GiB model faster on one 96 GiB GPU

A hot-expert cache made a 181 GiB model faster on one 96 GiB GPU

Comments 1
7 min read
Keras #22561 speedups: compiled Keras on PyTorch now generates text 11.6x faster

Keras #22561 speedups: compiled Keras on PyTorch now generates text 11.6x faster

Comments
7 min read
Why two out of three open-source projects don't get a Porchlamp score

Why two out of three open-source projects don't get a Porchlamp score

Comments
2 min read
Where to send your first open-source pull request (the projects that actually answer)

Where to send your first open-source pull request (the projects that actually answer)

Comments 2
3 min read
Which ecosystems leave your pull request hanging? npm and PyPI more than Rust, Go or Java

Which ecosystems leave your pull request hanging? npm and PyPI more than Rust, Go or Java

Comments
2 min read
I measured silence: 212,390 outside pull requests across 11,000 GitHub projects

I measured silence: 212,390 outside pull requests across 11,000 GitHub projects

11
Comments 3
3 min read
loading...