v0.1.5

v0.1.5 Release

Features

• Support Prefix Caching (APC) in vLLM (#274, #275, #283)
• Support APC in SGLang with Mem Bound (#282)
• Support Pipeline Parallelism (#246)
• Support MLA model in vLLM (#261)
• Support GPT-OSS in vLLM (#259)
• Support GPT-OSS in SGLang v0.5.9 (#268)
• Support vLLM v0.19.0
• Support SGLang v0.5.10

Bug Fixes

• Fix RadixCache Patch version issue in SGLang (#296)
• Fix bug when enable APC with GPT-OSS (#292)
• Fix async scheduler support on MP backend (#288)
• Fix reshape KV cache tensors in vLLM patches (#285)
• Fix ENOENT on the socket path when tp=1 (#276)

Maintenance

• Update default cached tokens (#290)
• Refactor: merge alloc_mla_kv_cache into alloc_kv_cache for SGLang interfaces (#278)
• Update scripts to adapt to new vLLM versions (#270)

Documentation

• Update README for recent changes (#293)
• Add Sardeenz demo link (#287)