Skip to content

GitLab

Commit da80a2e3 authored by guangzlu's avatar guangzlu
Browse files

changed fwd instance parameters

parent 88980945
Hide whitespace changes
Inline Side-by-side
Showing with 14 additions and 14 deletions
example/32_batched_gemm_scale_softmax_gemm/batched_multihead_attention_forward.cpp
View file @ da80a2e3
...@@ -105,9 +105,9 @@ using DeviceGemmInstance = ...@@ -105,9 +105,9 @@ using DeviceGemmInstance =
256, 256,
128, // MPerBlock 128, // MPerBlock
128, // NPerBlock 128, // NPerBlock
64, // KPerBlock 32, // KPerBlock
64, // Gemm1NPerBlock 64, // Gemm1NPerBlock
64, // Gemm1KPerBlock 32, // Gemm1KPerBlock
8, // AK1 8, // AK1
8, // BK1 8, // BK1
2, // B1K1 2, // B1K1
...@@ -130,11 +130,11 @@ using DeviceGemmInstance = ...@@ -130,11 +130,11 @@ using DeviceGemmInstance =
8, 8,
8, 8,
true, true,
S<8, 32, 1>, // B1BlockTransfer S<16, 16, 1>, // B1BlockTransfer
S<0, 2, 1>, S<0, 2, 1>,
S<0, 2, 1>, S<0, 2, 1>,
1, 1,
2, 4,
2, 2,
false, false,
1, // CShuffleMXdlPerWavePerShuffle 1, // CShuffleMXdlPerWavePerShuffle
......
example/32_batched_gemm_scale_softmax_gemm/batched_multihead_attention_forward_bf16.cpp
View file @ da80a2e3
...@@ -101,9 +101,9 @@ using DeviceGemmInstance = ...@@ -101,9 +101,9 @@ using DeviceGemmInstance =
256, 256,
128, // MPerBlock 128, // MPerBlock
128, // NPerBlock 128, // NPerBlock
64, // KPerBlock 32, // KPerBlock
64, // Gemm1NPerBlock 64, // Gemm1NPerBlock
64, // Gemm1KPerBlock 32, // Gemm1KPerBlock
8, // AK1 8, // AK1
8, // BK1 8, // BK1
2, // B1K1 2, // B1K1
...@@ -130,7 +130,7 @@ using DeviceGemmInstance = ...@@ -130,7 +130,7 @@ using DeviceGemmInstance =
S<0, 2, 1>, S<0, 2, 1>,
S<0, 2, 1>, S<0, 2, 1>,
1, 1,
2, 4,
2, 2,
false, false,
1, // CShuffleMXdlPerWavePerShuffle 1, // CShuffleMXdlPerWavePerShuffle
......
example/32_batched_gemm_scale_softmax_gemm/grouped_multihead_attention_forward.cpp
View file @ da80a2e3
...@@ -105,9 +105,9 @@ using DeviceGemmInstance = ...@@ -105,9 +105,9 @@ using DeviceGemmInstance =
256, 256,
128, // MPerBlock 128, // MPerBlock
128, // NPerBlock 128, // NPerBlock
64, // KPerBlock 32, // KPerBlock
64, // Gemm1NPerBlock 64, // Gemm1NPerBlock
64, // Gemm1KPerBlock 32, // Gemm1KPerBlock
8, // AK1 8, // AK1
8, // BK1 8, // BK1
2, // B1K1 2, // B1K1
...@@ -130,11 +130,11 @@ using DeviceGemmInstance = ...@@ -130,11 +130,11 @@ using DeviceGemmInstance =
8, 8,
8, 8,
true, true,
S<8, 32, 1>, // B1BlockTransfer S<16, 16, 1>, // B1BlockTransfer
S<0, 2, 1>, S<0, 2, 1>,
S<0, 2, 1>, S<0, 2, 1>,
1, 1,
2, 4,
2, 2,
false, false,
1, // CShuffleMXdlPerWavePerShuffle 1, // CShuffleMXdlPerWavePerShuffle
......
example/32_batched_gemm_scale_softmax_gemm/grouped_multihead_attention_forward_bf16.cpp
View file @ da80a2e3
...@@ -101,9 +101,9 @@ using DeviceGemmInstance = ...@@ -101,9 +101,9 @@ using DeviceGemmInstance =
256, 256,
128, // MPerBlock 128, // MPerBlock
128, // NPerBlock 128, // NPerBlock
64, // KPerBlock 32, // KPerBlock
64, // Gemm1NPerBlock 64, // Gemm1NPerBlock
64, // Gemm1KPerBlock 32, // Gemm1KPerBlock
8, // AK1 8, // AK1
8, // BK1 8, // BK1
2, // B1K1 2, // B1K1
...@@ -130,7 +130,7 @@ using DeviceGemmInstance = ...@@ -130,7 +130,7 @@ using DeviceGemmInstance =
S<0, 2, 1>, S<0, 2, 1>,
S<0, 2, 1>, S<0, 2, 1>,
1, 1,
2, 4,
2, 2,
false, false,
1, // CShuffleMXdlPerWavePerShuffle 1, // CShuffleMXdlPerWavePerShuffle
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment