#include <kernel_func.hpp>
|
| static constexpr uint32_t | layer_size = 3 |
| | layer_size = 3
|
| |
| static constexpr uint32_t | sequence_length = 2 |
| | sequence_length = 64
|
| |
| static constexpr uint32_t | batch_size = 512 |
| | batch_size = 512
|
| |
| static constexpr uint32_t | input_size = 384 |
| | input_size = 384
|
| |
| static constexpr uint32_t | hidden_size = 704 |
| | hidden_size = 688;
|
| |
| static constexpr uint32_t | wg_tile_m = 64 |
| | launch config
|
| |
| static constexpr uint32_t | wg_tile_n = 128 |
| |
| static constexpr uint32_t | sg_tile_m = 16 |
| |
| static constexpr uint32_t | sg_tile_n = 16 |
| |
| static constexpr uint32_t | sg_tile_k = 32 |
| |
◆ dtype_acc
◆ dtype_in
◆ batch_size
| constexpr uint32_t gru_config_t::batch_size = 512 |
|
staticconstexpr |
◆ hidden_size
| constexpr uint32_t gru_config_t::hidden_size = 704 |
|
staticconstexpr |
◆ input_size
| constexpr uint32_t gru_config_t::input_size = 384 |
|
staticconstexpr |
◆ layer_size
| constexpr uint32_t gru_config_t::layer_size = 3 |
|
staticconstexpr |
◆ sequence_length
| constexpr uint32_t gru_config_t::sequence_length = 2 |
|
staticconstexpr |
◆ sg_tile_k
| constexpr uint32_t gru_config_t::sg_tile_k = 32 |
|
staticconstexpr |
◆ sg_tile_m
| constexpr uint32_t gru_config_t::sg_tile_m = 16 |
|
staticconstexpr |
◆ sg_tile_n
| constexpr uint32_t gru_config_t::sg_tile_n = 16 |
|
staticconstexpr |
◆ wg_tile_m
| constexpr uint32_t gru_config_t::wg_tile_m = 64 |
|
staticconstexpr |
◆ wg_tile_n
| constexpr uint32_t gru_config_t::wg_tile_n = 128 |
|
staticconstexpr |