timshel1

3 ms·

1. 1.▲
   by timshel1 7mo 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.
2. 2.▲
   Reducing VRAM Footprint in PPO and GRPO Using Selective Log-Softmax(tylerromero.com)

   1 points by timshel1 2y ago|0 comments
3. 3.▲
   An Extension to Badge Active Learning for Variable-Sized Batches(tylerromero.com)

   1 points by timshel1 2y ago|0 comments
4. 4.▲
   Direct Preference Optimization Explained In-Depth(tylerromero.com)

   1 points by timshel1 2y ago|0 comments