[Kernel] Add non-gated support for NVFP4 CUTLASS MoE (#37320)
Signed-off-by: mgoin <mgoin64@gmail.com>
This commit is contained in:
+2
-1
@@ -262,7 +262,8 @@ void get_cutlass_moe_mm_data(
|
||||
torch::Tensor& problem_sizes1, torch::Tensor& problem_sizes2,
|
||||
torch::Tensor& input_permutation, torch::Tensor& output_permutation,
|
||||
const int64_t num_experts, const int64_t n, const int64_t k,
|
||||
const std::optional<torch::Tensor>& blockscale_offsets);
|
||||
const std::optional<torch::Tensor>& blockscale_offsets,
|
||||
const bool is_gated);
|
||||
|
||||
void get_cutlass_moe_mm_problem_sizes_from_expert_offsets(
|
||||
const torch::Tensor& expert_first_token_offset,
|
||||
|
||||
Reference in New Issue
Block a user