HH |ȓA brain.Event:2R. ,tensorboard.summary.writer.event_file_writer7S `/# \|ȓA*  hp_metricbZ5 L|ȓA* _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_metric3 A |ȓA*  _hparams_/session_start_infoJ  hparams  balance_jets   optimizerAdamW  use_moe   combine_pair_lossmin  transformer_typeGated  testing_file  dataset_limit ? & use_pairwise_interactions ?  l2_penalty -C6*? transformer_activationgelu  validation_file  masking Filling  moe_loss_scale {Gz?  num_experts  @ " num_regression_layers @ " transformer_dim_scale @ # trial_output_dir ./test_output % num_jet_embedding_layers   linear_activationgelu ! detection_loss_scale ? & num_classification_layers @  pairwise_input_sourceall num_experts_per_tok @ ! learning_rate_cycles ?  limit_val_batches ? $ use_mass_classification   num_gpu ? " regression_loss_scale  / "initial_embedding_skip_connections ? num_detector_layers @  kl_loss_scale   normalization  LayerNorm # position_embedding_dim P@ ! num_embedding_layers $@ " mask_sequence_vectors ?  trial_time J event_info_file75/data/spatop/event_files/v11/tt_hadronic_v7_full.yaml " num_pairwise_features @ $ linear_prelu_activation ?  linear_block_typeGRU & num_branch_encoder_layers @  balance_losses ? # train_validation_split ffffff?  usable_gpus  epochs @  skip_connections ? combinatorial_scale  H training_file75/data/spatop/tpm70_wpm30_FB350_training/all_merged.h5 num_attention_heads  @  num_encoder_layers @  limit_to_num_jets   dropout ?  gradient_clip $@  verbose_output  $ balance_classifications   focal_gamma  ( num_branch_embedding_layers @ # num_jet_encoder_layers @ & split_symmetric_attention ? ( learning_rate_warmup_epochs ? " dataset_randomization   transformer_dim `@ # num_dataloader_workers  @ " initial_embedding_dim @@  learning_rate -C6:? & classification_loss_scale  # pairwise_embedding_dim  @  partial_events ?  batch_size p@  balance_particles   hidden_dim `@  normalize_features ? " assignment_loss_scale ?P<ȷR |ȓA*1 / _hparams_/session_end_infoJ  hparams"zZ