You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I tried to run DeepSeek V4 Flash with expert parallel with deepep version 1.2.1 and sglang 0.5.12 on H20,but I got the Error, And I found that the recipe does not mention the FP8-FP4 mixed EP running method, and I wonder if sglang will support deepep with FP4 in the future on SM90
reacted with thumbs up emoji reacted with thumbs down emoji reacted with laugh emoji reacted with hooray emoji reacted with confused emoji reacted with heart emoji reacted with rocket emoji reacted with eyes emoji
Uh oh!
There was an error while loading. Please reload this page.
-
I tried to run DeepSeek V4 Flash with expert parallel with deepep version 1.2.1 and sglang 0.5.12 on H20,but I got the Error, And I found that the recipe does not mention the FP8-FP4 mixed EP running method, and I wonder if sglang will support deepep with FP4 in the future on SM90
tvm.error.InternalError: Assertion error (/deepgemm/csrc/apis/../jit_kernels/impls/../heuristics/../../utils/layout.hpp:57):
ab.scalar_type() == kPackedFP4 and arch_major == 10
The Assertion says it need FP4 and SM100 simultaneously, and We expect it can run on SM90
Beta Was this translation helpful? Give feedback.
All reactions