HH 拃OȓA brain.Event:2R. ,tensorboard.summary.writer.event_file_writerh`/# vOȓA*  hp_metricoUg5 ^OȓA* _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} A OȓA*  _hparams_/session_start_infoJ  hparams " num_regression_layers @  balance_jets   kl_loss_scale   limit_to_num_jets  # position_embedding_dim P@ ! learning_rate_cycles ? " num_pairwise_features @  testing_file num_experts_per_tok @  linear_block_typeGRU  num_experts  @  usable_gpus  hidden_dim `@  dropout ? # num_jet_encoder_layers @  l2_penalty -C6*? " transformer_dim_scale @  transformer_dim `@ $ linear_prelu_activation ?  partial_events ?  linear_activationgelu # num_dataloader_workers  @  learning_rate -C6:? $ balance_classifications   batch_size p@  moe_loss_scale {Gz? " mask_sequence_vectors ? ( learning_rate_warmup_epochs ? & num_branch_encoder_layers @ & num_classification_layers @  num_gpu ? num_detector_layers @ # trial_output_dir ./test_output / "initial_embedding_skip_connections ? " dataset_randomization  ! detection_loss_scale ?  focal_gamma   limit_val_batches ?  masking Filling  epochs @ $ use_mass_classification   skip_connections ? ( num_branch_embedding_layers @  combine_pair_lossmin & classification_loss_scale  # pairwise_embedding_dim  @  balance_losses ? combinatorial_scale   validation_file  pairwise_input_sourceall " initial_embedding_dim @@ ! num_embedding_layers $@  balance_particles  " regression_loss_scale  J event_info_file75/data/spatop/event_files/v11/tt_hadronic_v7_full.yaml  num_encoder_layers @  use_moe  % num_jet_embedding_layers  H training_file75/data/spatop/tpm70_wpm30_FB350_training/all_merged.h5 & split_symmetric_attention ? & use_pairwise_interactions ?  verbose_output   normalization  LayerNorm  normalize_features ? " assignment_loss_scale ?  gradient_clip $@ # train_validation_split ffffff?  optimizerAdamW  dataset_limit ? num_attention_heads  @ transformer_activationgelu  trial_time  transformer_typeGated]9<ȷR ȪOȓA*1 / _hparams_/session_end_infoJ  hparams"\>