mirror of
https://github.com/HChaZZY/Stockfish.git
synced 2025-12-24 19:16:49 +08:00
Skeleton for new evaluate learner
This commit is contained in:
@@ -54,6 +54,12 @@ namespace Eval::NNUE {
|
||||
const std::string& seed,
|
||||
SynchronizedRegionLogger::Region& out) {
|
||||
|
||||
#if defined (OPENBLAS_VERSION)
|
||||
openblas_set_num_threads(1);
|
||||
#elif defined (INTEL_MKL_VERSION)
|
||||
mkl_set_num_threads(1);
|
||||
#endif
|
||||
|
||||
out << "INFO (initialize_training): Initializing NN training for "
|
||||
<< get_architecture_string() << std::endl;
|
||||
|
||||
@@ -199,39 +205,62 @@ namespace Eval::NNUE {
|
||||
|
||||
bool collect_stats = verbose;
|
||||
|
||||
std::vector<double> abs_eval_diff_sum_local(thread_pool.size(), 0.0);
|
||||
std::vector<double> abs_discrete_eval_sum_local(thread_pool.size(), 0.0);
|
||||
std::vector<double> gradient_norm_local(thread_pool.size(), 0.0);
|
||||
|
||||
while (examples.size() >= batch_size) {
|
||||
std::vector<Example> batch(examples.end() - batch_size, examples.end());
|
||||
examples.resize(examples.size() - batch_size);
|
||||
|
||||
const auto network_output = trainer->propagate(thread_pool, batch);
|
||||
|
||||
const auto network_output = trainer->step_start(thread_pool, batch);
|
||||
std::vector<LearnFloatType> gradients(batch.size());
|
||||
for (std::size_t b = 0; b < batch.size(); ++b) {
|
||||
const auto shallow = static_cast<Value>(round<std::int32_t>(
|
||||
batch[b].sign * network_output[b] * kPonanzaConstant));
|
||||
const auto discrete = batch[b].sign * batch[b].discrete_nn_eval;
|
||||
const auto& psv = batch[b].psv;
|
||||
const double gradient =
|
||||
batch[b].sign * calc_grad(shallow, (Value)psv.score, psv.game_result, psv.gamePly);
|
||||
gradients[b] = static_cast<LearnFloatType>(gradient * batch[b].weight);
|
||||
|
||||
thread_pool.for_each_index_chunk_with_workers(
|
||||
std::size_t(0), batch.size(),
|
||||
[&](Thread& th, std::size_t offset, std::size_t count) {
|
||||
const auto thread_id = th.thread_idx();
|
||||
|
||||
trainer->propagate(th, offset, count);
|
||||
|
||||
for (std::size_t b = offset; b < offset + count; ++b) {
|
||||
const auto shallow = static_cast<Value>(round<std::int32_t>(
|
||||
batch[b].sign * network_output[b] * kPonanzaConstant));
|
||||
const auto discrete = batch[b].sign * batch[b].discrete_nn_eval;
|
||||
const auto& psv = batch[b].psv;
|
||||
const double gradient =
|
||||
batch[b].sign * calc_grad(shallow, (Value)psv.score, psv.game_result, psv.gamePly);
|
||||
gradients[b] = static_cast<LearnFloatType>(gradient * batch[b].weight);
|
||||
|
||||
|
||||
// The discrete eval will only be valid before first backpropagation,
|
||||
// that is only for the first batch.
|
||||
// Similarily we want only gradients from one batch.
|
||||
if (collect_stats)
|
||||
{
|
||||
abs_eval_diff_sum += std::abs(discrete - shallow);
|
||||
abs_discrete_eval_sum += std::abs(discrete);
|
||||
gradient_norm += std::abs(gradient);
|
||||
// The discrete eval will only be valid before first backpropagation,
|
||||
// that is only for the first batch.
|
||||
// Similarily we want only gradients from one batch.
|
||||
if (collect_stats)
|
||||
{
|
||||
abs_eval_diff_sum_local[thread_id] += std::abs(discrete - shallow);
|
||||
abs_discrete_eval_sum_local[thread_id] += std::abs(discrete);
|
||||
gradient_norm_local[thread_id] += std::abs(gradient);
|
||||
}
|
||||
}
|
||||
|
||||
trainer->backpropagate(th, gradients.data(), offset, count);
|
||||
}
|
||||
}
|
||||
);
|
||||
thread_pool.wait_for_workers_finished();
|
||||
|
||||
trainer->backpropagate(thread_pool, gradients.data(), learning_rate);
|
||||
trainer->step_end(thread_pool, learning_rate);
|
||||
|
||||
collect_stats = false;
|
||||
}
|
||||
|
||||
if (verbose)
|
||||
{
|
||||
abs_eval_diff_sum = std::accumulate(abs_eval_diff_sum_local.begin(), abs_eval_diff_sum_local.end(), 0.0);
|
||||
abs_discrete_eval_sum = std::accumulate(abs_discrete_eval_sum_local.begin(), abs_discrete_eval_sum_local.end(), 0.0);
|
||||
gradient_norm = std::accumulate(gradient_norm_local.begin(), gradient_norm_local.end(), 0.0);
|
||||
}
|
||||
|
||||
if (verbose) {
|
||||
const double avg_abs_eval_diff = abs_eval_diff_sum / batch_size;
|
||||
const double avg_abs_discrete_eval = abs_discrete_eval_sum / batch_size;
|
||||
|
||||
Reference in New Issue
Block a user