[Kernel] fp4 marlin kernel (#17687)
Signed-off-by: Jinzhen Lin <linjinzhen@hotmail.com>
This commit is contained in:
@@ -304,8 +304,10 @@ class HQQMarlinMethod(LinearMethodBase):
|
||||
|
||||
marlin_out = ops.gptq_marlin_gemm(
|
||||
x,
|
||||
None,
|
||||
layer.marlin_qweight,
|
||||
scales,
|
||||
None,
|
||||
zeros,
|
||||
layer.g_idx,
|
||||
layer.g_idx_sort_indices,
|
||||
@@ -315,7 +317,7 @@ class HQQMarlinMethod(LinearMethodBase):
|
||||
self.output_size_per_partition,
|
||||
self.input_size_per_partition,
|
||||
True, # is_k_full
|
||||
True, # has_zp
|
||||
False, # use atomic add
|
||||
True, # use 32-bit reduce
|
||||
True, # use float zp
|
||||
)
|
||||
|
||||
Reference in New Issue
Block a user