[None][fix] Use legal MXF8 scale-vector encoding

Signed-off-by: Mingyang Hao <200044211+mingyangHao@users.noreply.github.com>

Author: mingyangHao

tensorrt_llm/_torch/cute_dsl_kernels/blackwell/dynamic_mxf8_mainloop.py

@@ -11,11 +11,7 @@
 from cutlass.cute.arch.nvvm_wrappers import nvvm
 from cutlass.cutlass_dsl import Boolean, Int32, Uint32, dsl_user_op
 
-_MXF8_SCALE_VEC_SIZE = getattr(
-    _nvvm_raw.Tcgen05MMAScaleVecSize,
-    "BLOCK32",
-    _nvvm_raw.Tcgen05MMAScaleVecSize.X1,
-)
+_MXF8_SCALE_VEC_SIZE = _nvvm_raw.Tcgen05MMAScaleVecSize.X1
 
 
 def _align16(x):