Feat/glm5 2 main - #523

Draft

wooway777 wants to merge 9 commits into

mainfrom

feat/glm5-2-main

Commits on Aug 1, 2026

pepe: glm 5.2 trial
wooway777
committed
ee55795
pepe: optimize deepseek v2 for iluvatar
wooway777
committed
9332baa
pepe: glm 5.2 service
wooway777
committed
df4e9c1
pepe: adjust decode graph batch size
wooway777
committed
0be8042
pepe: support pp in bench.py
wooway777
committed
06fb828
fix: adapt glm runtime to current InfiniCore APIs
wooway777
committed
a9d316b
fix: preserve prefix caching with GLM scheduling
wooway777
committed
c857f22
fix: adapt GLM to current cache and weight loading APIs
wooway777
committed
d764b71
fix: preserve Mamba support in dynamic decode graphs
wooway777
committed
5929d37