Skip to content

Add dual activation #357

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
kevlu8 merged 4 commits into from
Aug 9, 2026
Merged

Add dual activation #357

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
43afa6c
Add dual activation
kevlu8 Aug 9, 2026
19b514d
Expand L1 (actually L2) biases
kevlu8 Aug 9, 2026
dd7f68a
Fix bug with l2b size
kevlu8 Aug 9, 2026
777b0e0
Minor cleanups
kevlu8 Aug 9, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
15 changes: 10 additions & 5 deletions engine/nnue/network.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -41,7 +41,7 @@ void Network::load() {

for (int i = 0; i < NBUCKETS; i++) {
for (int j = 0; j < L3_SIZE; j++) {
for (int k = 0; k < L2_SIZE; k++) {
for (int k = 0; k < L2_SIZE * 2; k++) {
memcpy(&l2_weights[i][k][j], ptr, 4);
ptr += 4;
}
Expand All @@ -55,6 +55,7 @@ void Network::load() {
ptr += sizeof(output_weights);

memcpy(&output_biases, ptr, sizeof(output_biases));
ptr += sizeof(output_biases);
}

int calculate_index(Square sq, PieceType pt, bool side, bool perspective, int nbucket) {
Expand All @@ -79,7 +80,7 @@ int32_t nnue_eval(const Network &net, const Accumulator &stm, const Accumulator

alignas(64) uint8_t l1[L1_SIZE];
alignas(64) int32_t l2i[L2_SIZE];
alignas(64) float l2[L2_SIZE];
alignas(64) float l2[L2_SIZE * 2];
alignas(64) float l3[L3_SIZE];

// Pairwise mul
Expand Down Expand Up @@ -122,26 +123,30 @@ int32_t nnue_eval(const Network &net, const Accumulator &stm, const Accumulator

// Convert l2 into a proper float array
for (int i = 0; i < L2_SIZE; i += FLOATS_PER_VEC) {
// using the raw values, do dual activation
// val is activated with CReLU and val2 with CSReLU
ivec i_val = simd::load_ivec((ivec *)&l2i[i]);
fvec val = simd::cvt_i32_f32(i_val);

fvec bias = simd::load_fvec(&net.l1_biases[nbucket][i]);

val = simd::fma_f32(val, div, bias);

val = simd::clamp_f32(val, f_zero, f_clip);
fvec val2 = simd::mul_f32(val, val); // note that order is changed for dual activation (CSReLU)

val = simd::mul_f32(val, val);
val = simd::clamp_f32(val, f_zero, f_clip);
val2 = simd::clamp_f32(val2, f_zero, f_clip);

simd::store_f32(&l2[i], val);
simd::store_f32(&l2[i + L2_SIZE], val2);
}

for (int i = 0; i < L3_SIZE; i += FLOATS_PER_VEC * L2_UNROLL) {
fvec sums[L2_UNROLL];
for (int j = 0; j < L2_UNROLL; j++)
sums[j] = simd::load_fvec(&net.l2_biases[nbucket][i + j * FLOATS_PER_VEC]);

for (int j = 0; j < L2_SIZE; j++) {
for (int j = 0; j < L2_SIZE * 2; j++) {
fvec val = simd::broadcast_f32(l2[j]);

for (int k = 0; k < L2_UNROLL; k++) {
Expand Down
2 changes: 1 addition & 1 deletion engine/nnue/network.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -52,7 +52,7 @@ struct alignas(32) Network {
int8_t l1_weights[NBUCKETS][L2_SIZE][L1_SIZE];
float l1_biases[NBUCKETS][L2_SIZE];

float l2_weights[NBUCKETS][L2_SIZE][L3_SIZE];
float l2_weights[NBUCKETS][L2_SIZE * 2][L3_SIZE];
float l2_biases[NBUCKETS][L3_SIZE];

float output_weights[NBUCKETS][L3_SIZE];
Expand Down
Loading