- University of Kansas, Lawrence, Kansas, 66049
- https://pavansaipendry.dev/
- in/pavansaireddypendry
- https://scholar.google.com/citations?user=AVS2fb4AAAAJ&hl=en
Pinned Loading
-
-
gated-attention-repro
gated-attention-repro PublicReproduction of the NeurIPS 2025 Best Paper 'Gated Attention for LLMs' — baseline vs gated on TinyShakespeare & enwik8, with results and analysis.
Python
-
highway-rl
highway-rl PublicDeep RL self-driving agent — PPO on highway-env. ~80km/h, 0% collisions. Live TensorBoard video, behavioral eval, constrained-RL speed limits.
Python
-
nsa-mini
nsa-mini PublicFrom-scratch Native Sparse Attention (DeepSeek, arXiv:2502.11089): PyTorch reference + Triton kernels, benchmarked vs FlashAttention-2 and validated on A100. 22x faster window kernel at 64k; qualit…
Python
-
-
reasoning-slm
reasoning-slm PublicA math/code reasoning small language model trained end-to-end on my own stack: data pipeline -> pretraining (gated + NSA sparse attention) -> SFT -> GRPO RL -> served on a from-scratch inference en…
Python
If the problem persists, check the GitHub status page or contact support.