Contributors

Contributors

This fork is developed and tested by people who contribute code, reviews, benchmarks, documentation, and access to hardware.

Project lead

GenerelSchwerz

Main developer and maintainer. Leads the fork's architecture, implementation, integration, releases, documentation, and performance validation.

Major contributors

PiggiDragon

Major code and testing contributor. Merged contributions include:

• Qwen3.8 MTP and DSpark validation and fixes (PR #3)
• Native quantized KV FlashAttention (PR #4)
• CUDA build and configuration fixes (PR #34)
• Benchmark controls for KV placement and workspace testing (PR #70)
• Correct multi-stream scheduler range copies (PR #77)
• Broader architecture coverage for rollback, quantization, and parallel sequences (PR #94)

PiggiDragon also maintains ongoing work on host-KV transport, multi-GPU placement, speculative decoding, and native quantized FlashAttention. See all submitted pull requests.

Machine donors

The following people donated machines or remote machine access to GenerelSchwerz for testing that the project could not cover locally.

Contributor	Hardware	What it enabled
Add contributor	CPU, GPU, RAM, and operating system	Validation or benchmark work

Only add people here with their permission. Describe the hardware and the work it enabled so the credit remains useful.

Contributing

Code, testing, reproducible benchmark reports, documentation corrections, and platform-specific validation are all useful. Open a pull request against the relevant fork branch and include enough evidence for someone else to reproduce the result.