Pinned Loading

1. mega-kernel mega-kernel Public

   Mega Kernel is an experimental CUDA inference engine for accelerating batch-one autoregressive decoding of Qwen3-0.6B. Instead of executing the model as hundreds of separate framework operations, i…

   Cuda 1

2. rl-env rl-env Public

   This env is an ultra hard rl enviornment designed for training LM's on fixing large training pipelines , this one implements and adverserial version of a broken unsloth env where the model is taske…

   Python

3. RL-MAX RL-MAX Public

   Decoupled rl enviornment training with zero2 and hot reload of weights with vllm

   Python 4

4. Triton_Kernels- Triton_Kernels- Public

   Kernels Triton

   Jupyter Notebook 1

5. lidar-model lidar-model Public

   Sensor Fusion for Lidar LM - Implements a custom lidar encoder and sensor fusion into a language model

   Python

6. hint-guided-cast hint-guided-cast Public

   Accelerate code for hint guided cast , a variant of grpo where advantage estimation is done from a frozen model itself as a teacher with guided hints , for per token advantage that differs

   Python