Popular repositories Loading

1. Decoupled-multi-agent-traffic-coordination Decoupled-multi-agent-traffic-coordination Public

   Solo-built Unity (C#) multi-agent navigation for cars and drones - from-scratch 2D ORCA; best drone performance of all groups

   C#

2. gpt2-lora-int8 gpt2-lora-int8 Public

   From-scratch PyTorch LoRA + int8 weight quantisation on GPT-2 (124M): LoRA matches full fine-tuning at 0.24% of params; int8 cuts inference memory ~20%.

   Python

3. RL-for-Train-Speed-Profile RL-for-Train-Speed-Profile Public

   SAC (Ray RLlib, PyTorch) agent that learns a train's speed profile in the Open Rails simulator for punctuality and speed-limit compliance - KTH bachelor thesis

   Python