timshel1

Joined 4 years ago

Karma

5

Submissions

4

Member since

Apr 3

llm researcher

View on Hacker News

Recent submissions

Comment · 8 months ago

Modded-nanogpt is also much more data efficient than vanilla napogpt, even if some of the individual optimizations trade off higher throughput for worse data efficiency.

Reducing VRAM Footprint in PPO and GRPO Using Selective Log-Softmax tylerromero.com

1 timshel1 1 year ago 0 comments

An Extension to Badge Active Learning for Variable-Sized Batches tylerromero.com

1 timshel1 1 year ago 0 comments

Direct Preference Optimization Explained In-Depth tylerromero.com

1 timshel1 2 years ago 0 comments