HH =>A brain.Event:2R. ,tensorboard.summary.writer.event_file_writer1`/# >A*  hp_metricΩ5 >A* _hparams_/experimentJ hparams  " hidden_dim " transformer_dim " transformer_dim_scale " initial_embedding_dim " position_embedding_dim " num_embedding_layers " num_encoder_layers " num_branch_embedding_layers " num_branch_encoder_layers " num_jet_embedding_layers " num_jet_encoder_layers " num_detector_layers " num_regression_layers " num_classification_layers " split_symmetric_attention " num_attention_heads " transformer_activation " skip_connections "& "initial_embedding_skip_connections " linear_block_type " transformer_type " linear_activation " normalization " masking " linear_prelu_activation " use_pairwise_interactions " num_pairwise_features " pairwise_embedding_dim " pairwise_input_source " use_moe " num_experts " num_experts_per_tok " moe_loss_scale " event_info_file " training_file " validation_file " testing_file " normalize_features " limit_to_num_jets " balance_particles " balance_jets " balance_classifications " partial_events " dataset_limit " dataset_randomization " train_validation_split " limit_val_batches " batch_size " num_dataloader_workers " mask_sequence_vectors " combine_pair_loss " optimizer " learning_rate " focal_gamma " combinatorial_scale " learning_rate_warmup_epochs " learning_rate_cycles " assignment_loss_scale " detection_loss_scale " kl_loss_scale " regression_loss_scale " classification_loss_scale " use_mass_classification " balance_losses " l2_penalty " gradient_clip " dropout " epochs " num_gpu " verbose_output " usable_gpus " trial_time " trial_output_dir *  hp_metric;:R : >A*  _hparams_/session_start_infoJ  hparams  kl_loss_scale  & num_branch_encoder_layers @ # trial_output_dir ./test_output  validation_file " mask_sequence_vectors ?  learning_rate -C6:? & split_symmetric_attention ? " transformer_dim_scale @  testing_file  use_moe  ! learning_rate_cycles ? " num_pairwise_features @  normalize_features ?  hidden_dim 6@  gradient_clip $@ $ balance_classifications   transformer_dim 6@  moe_loss_scale {Gz?  num_encoder_layers @  dataset_limit ? & num_classification_layers @  linear_block_typeGRU num_attention_heads @ num_experts_per_tok @ num_detector_layers @ " initial_embedding_dim (@  usable_gpus  transformer_typeGated  linear_activationgelu  combine_pair_lossmin ( num_branch_embedding_layers @ ! num_embedding_layers $@ $ linear_prelu_activation ? & use_pairwise_interactions  " dataset_randomization  # position_embedding_dim (@  limit_to_num_jets  J event_info_file75/data/spatop/event_files/v11/tt_hadronic_v7_full.yaml  l2_penalty -C6*? " assignment_loss_scale ? ! detection_loss_scale ? " num_regression_layers @  focal_gamma  / "initial_embedding_skip_connections ? ( learning_rate_warmup_epochs ? % num_jet_embedding_layers   epochs @@  balance_jets  # pairwise_embedding_dim  @  trial_time # num_dataloader_workers  @ combinatorial_scale  " regression_loss_scale  & classification_loss_scale   optimizerAdamW  partial_events ? H training_file75/data/spatop/tpm70_wpm30_FB350_training/all_merged.h5  masking Filling  balance_particles  # train_validation_split ffffff? # num_jet_encoder_layers @  verbose_output   pairwise_input_source  balance_losses ?  dropout ?  limit_val_batches ? transformer_activationgelu  num_experts  @  skip_connections ?  batch_size @ $ use_mass_classification   normalization  LayerNorm  num_gpu ?{_<ȷR <>A*1 / _hparams_/session_end_infoJ  hparams"C