Pinned Loading

1. kibitzer kibitzer Public

   chess model that treats board states as tokens. learns rules from millions of games, studies the legends, then plays itself to get better.

   Python 2

2. HyperKuvid-Labs/energy_throttling_llms HyperKuvid-Labs/energy_throttling_llms Public

   Energy-aware speculative decoding for EAGLE3 + SGLang, using GPU telemetry to find batch-specific configs that maximize throughput without thermal throttling.

   Python 2

3. transformer-lab transformer-lab Public

   Plug-and-play PyTorch transformer research framework. Swap attention, FFN, norm, positional, connection, projection, optimizer, scheduler, loss, and dataset via Hydra config groups. Encoder-decoder…

   Python 14 1

4. HyperKuvid-Labs/DGAT HyperKuvid-Labs/DGAT Public

   Dependency Graph as a Tool

   C++ 6 1

5. HyperKuvid-Labs/alpha-stack HyperKuvid-Labs/alpha-stack Public

   Intelligent agent that converts natural language prompts into production-ready multi-file codebases with automatic dependency resolution, Docker validation, and iterative error correction.

   Python 4 2

6. learn_cuda learn_cuda Public

   Learning cuda (aka gpu programming)

   Cuda 2