MCPcopy Create free account
hub / github.com/DeepGraphLearning/graphvite / train

Method train

include/core/solver.h:1511–1522  ·  view source on GitHub ↗

* @brief Train embeddings with samples in the sample block * @param _head_partition_id id of head partition * @param _tail_partition_id id of tail partition */

Source from the content-addressed store, hash-verified

1509 * @param _tail_partition_id id of tail partition
1510 */
1511 void train(int _head_partition_id, int _tail_partition_id) {
1512 CUDA_CHECK(cudaSetDevice(device_id));
1513 load_partition(_head_partition_id, _tail_partition_id);
1514
1515 auto &samples = solver->sample_pools[solver->pool_id][head_partition_id][tail_partition_id];
1516 log_frequency = solver->log_frequency;
1517 for (int i = 0; i < solver->positive_reuse; i++)
1518 for (int j = 0; j < solver->episode_size; j++) {
1519 batch.copy(&samples[j * batch_size], batch_size * kSampleSize);
1520 train_batch(solver->batch_id++);
1521 }
1522 }
1523
1524 /** Train a single batch */
1525 virtual void train_batch(int batch_id) {

Callers

nothing calls this directly

Calls 1

copyMethod · 0.80

Tested by

no test coverage detected