Leonccaa /
ReVolta
Stable, high-throughput inference server for multi-GPU NVIDIA V100 (SM70): Qwen3.8 Flash-Next with AWQ / NVFP4, multimodal input, CPU/disk KV offload, and MTP or DFlash2 speculative decoding.
Updated 43 min ago
Why now
Version v0.1.4 came out on October 27.
+22 stars gained this week
Stars gained each day, last 30 days
Oct 4Oct 19Today
This week Before
In short
Needs four or more V100 32 GB GPUs on NVLink or one PCIe switch with P2P, Linux x86-64, and an NVIDIA CUDA 12.8 driver. It serves an OpenAI-compatible API via a vllm-style command.
Summary written by AI from the repository’s description and README.
Leonccaa / ReVolta
View on GitHub(opens in new tab)