in sparse_autoencoder/kernels.py [0:0]
def triton_sparse_transpose_dense_matmul_kernel(
coo_indices_ptr,
coo_values_ptr,
dense_ptr,
out_ptr,
stride_da,
stride_db,
B,
N,
AK,
BLOCK_SIZE_AK: tl.constexpr,
BLOCK_SIZE_B: tl.constexpr,