Edit model card

HBERTv1_emb_compress_48_L12_H64_A2

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

  • Loss: 6.4079
  • Accuracy: 0.1285

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: 1e-05
  • train_batch_size: 96
  • eval_batch_size: 96
  • seed: 10
  • distributed_type: multi-GPU
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_steps: 10000
  • num_epochs: 5

Training results

Training Loss Epoch Step Validation Loss Accuracy
8.6554 0.16 10000 8.5846 0.0483
7.2331 0.33 20000 7.2280 0.0542
7.0014 0.49 30000 6.9927 0.0677
6.8699 0.66 40000 6.8637 0.0856
6.7777 0.82 50000 6.7726 0.0922
6.7091 0.98 60000 6.7101 0.0974
6.6626 1.15 70000 6.6620 0.1015
6.6279 1.31 80000 6.6255 0.1040
6.5917 1.47 90000 6.5948 0.1068
6.5691 1.64 100000 6.5695 0.1094
6.5486 1.8 110000 6.5460 0.1122
6.5246 1.97 120000 6.5275 0.1144
6.5069 2.13 130000 6.5115 0.1162
6.5001 2.29 140000 6.4962 0.1180
6.4785 2.46 150000 6.4822 0.1197
6.4706 2.62 160000 6.4714 0.1212
6.4612 2.79 170000 6.4610 0.1225
6.4485 2.95 180000 6.4530 0.1233
6.4477 3.11 190000 6.4441 0.1243
6.4373 3.28 200000 6.4395 0.1251
6.4351 3.44 210000 6.4322 0.1259
6.4273 3.6 220000 6.4264 0.1262
6.4153 3.77 230000 6.4219 0.1269
6.4188 3.93 240000 6.4182 0.1274
6.4128 4.1 250000 6.4150 0.1278
6.4189 4.26 260000 6.4121 0.1280
6.4102 4.42 270000 6.4112 0.1282
6.4105 4.59 280000 6.4087 0.1285
6.4065 4.75 290000 6.4067 0.1287
6.4082 4.92 300000 6.4070 0.1285

Framework versions

  • Transformers 4.33.2
  • Pytorch 1.14.0a0+410ce96
  • Datasets 2.14.5
  • Tokenizers 0.13.3
Downloads last month
2
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_emb_compress_48_L12_H64_A2

Evaluation results

  • Accuracy on gokuls/wiki_book_corpus_complete_processed_bert_dataset
    self-reported
    0.129