chore: bump version to 0.3.20 (#2171)

fix(misc): replace deprecated llama.cpp references (#2170)

feat: Update llama.cpp to ggerganov/llama.cpp@f49e9178767d557a522618b16ce8694f9ddac628 (#2169)

feat(server): add model-load chat_template_kwargs (#2168)

ci: publish release wheels as py3-none (#2166)

fix(ci): publish distinct manylinux and musllinux cpu wheels (#2165)

Bump version to 0.3.19 (#2162)

feat: Update llama.cpp to ggerganov/llama.cpp@c0159f9c1f874da15e94f371d136f5920b4b5335 (#2161)

fix: handle embedding models without KV memory (#2160)

fix(ci): reduce CUDA binary wheel size only including cubins for current arches and one PTX target for forward compatibility (#2158)

chore: bump version (#2157)

feat: expose attention_type parameter in Llama.__init__ (#2143)

fix(ci): docker build workflow (#2156)

fix(ci): cuda wheel workflow (#2155)

fix(ci): release wheel workflow (#2154)