cannot run glm4.7-flash on 4*32GB V100

[RNGMARTIN]

cannot run glm4.7-flash on 4*32GB V100 by lmdeploy pytroch

report error:OutOfResources: out of resource: shared memory, Required: 102400, Hardware limit: 98304.

[RunningLeon]

@RNGMARTIN Hi, it may be due to the large head_dim in MLA. Sadly, we don't have v100 on hands, pls. check if this the reason.

[valentijnvenus]

Eager to know too