hBERTv1_new_pretrain_w_init_48_stsb

This model is a fine-tuned version of gokuls/bert_12_layer_model_v1_complete_training_new_wt_init_48 on the GLUE STSB dataset. It achieves the following results on the evaluation set:

  • Loss: 0.9800
  • Pearson: 0.7515
  • Spearmanr: 0.7472
  • Combined Score: 0.7493

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

  • learning_rate: 4e-05
  • train_batch_size: 128
  • eval_batch_size: 128
  • seed: 10
  • distributed_type: multi-GPU
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • num_epochs: 50

Training results

Training Loss Epoch Step Validation Loss Pearson Spearmanr Combined Score
2.5456 1.0 45 2.2706 0.1246 0.1141 0.1194
2.0514 2.0 90 2.0613 0.5266 0.5198 0.5232
1.3837 3.0 135 1.1984 0.6853 0.6942 0.6897
1.0297 4.0 180 1.6176 0.6869 0.6961 0.6915
0.8064 5.0 225 1.1444 0.7476 0.7445 0.7460
0.604 6.0 270 1.2754 0.7422 0.7450 0.7436
0.4818 7.0 315 1.1407 0.7687 0.7673 0.7680
0.3905 8.0 360 1.1860 0.7560 0.7604 0.7582
0.3476 9.0 405 0.9800 0.7515 0.7472 0.7493
0.2819 10.0 450 1.0156 0.7521 0.7507 0.7514
0.2418 11.0 495 1.0174 0.7516 0.7480 0.7498
0.2068 12.0 540 1.2367 0.7530 0.7523 0.7527
0.1863 13.0 585 1.0073 0.7491 0.7468 0.7480
0.1929 14.0 630 1.0470 0.7517 0.7505 0.7511

Framework versions

  • Transformers 4.29.2
  • Pytorch 1.14.0a0+410ce96
  • Datasets 2.12.0
  • Tokenizers 0.13.3
Downloads last month
10
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.

Dataset used to train gokuls/hBERTv1_new_pretrain_w_init_48_stsb

Evaluation results