Pinned Loading

1. nanoserve nanoserve Public

   A GPT-2 inference engine built from scratch to measure what a KV-cache actually buys and what it costs in memory.

   Python

2. quick-recognition quick-recognition Public

   This web app uses face-api.js which is built upon tensorflowJS. It has deep learning pre-trained models, used to detect facial landmark and facial recognition and store feature vectors for future m…

   JavaScript 1 1

3. nanograd nanograd Public

   how does backpropagation work in practise?

   Jupyter Notebook

4. finrag finrag Public

   Cited-answer RAG over SEC filings (FinanceBench): retrieval and reranking, abstention, LLM-judge evaluation, vLLM + FastAPI service

   Python

5. fleet-sim fleet-sim Public

   Warehouse robot fleet controller and simulator in Java 17: Hungarian task assignment, windowed cooperative A*, station admission control, charging policies, CBS baseline, independent Python checker

   Java

6. sva-agent sva-agent Public

   LLM-written SystemVerilog Assertions scored by an open-source model checker (EBMC) on NVIDIA's FVEval, calibrated against Jasper

   Python