AIPython★ 473⑂ 10 forksApache-2.0
SigmanticAI/apex-inference-chip
FPGA Inference Chip
The project demonstrates the implementation of one transformer layer of Qwen2.5-0.5B in RTL on FPGA, achieving bit‑exact match with the golden model. Measured performance is 0.56 token per second — a 140× speedup over baseline.
// KEY FEATURES
- One transformer layer of Qwen2.5-0.5B is implemented in RTL on FPGA
- Bit‑exact accuracy matches the golden model Qwen2.5-0.5B
- Measured speed is 0.56 token/s — 140× speedup vs baseline
- Full reproducibility and proof chain are available in the repository
New repositories every 30 minutes
REDDYX AI scans GitHub 24/7 and ships the best AI/ML/Web3 projects to Telegram.
Join on Telegram