distilBERT-finetuned-resumes-sections

This model is a fine-tuned version of Geotrend/distilbert-base-en-fr-cased on an unknown dataset. It achieves the following results on the evaluation set:

  • Loss: 0.0369
  • F1: 0.9652
  • Roc Auc: 0.9808
  • Accuracy: 0.9621

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: 2e-05
  • train_batch_size: 8
  • eval_batch_size: 8
  • seed: 42
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • num_epochs: 20

Training results

Training Loss Epoch Step Validation Loss F1 Roc Auc Accuracy
0.0509 1.0 1173 0.0331 0.9439 0.9659 0.9356
0.024 2.0 2346 0.0274 0.9550 0.9750 0.9493
0.0148 3.0 3519 0.0290 0.9493 0.9712 0.9446
0.0089 4.0 4692 0.0324 0.9492 0.9714 0.9442
0.0071 5.0 5865 0.0317 0.9540 0.9732 0.9476
0.0064 6.0 7038 0.0324 0.9527 0.9742 0.9484
0.0036 7.0 8211 0.0320 0.9574 0.9766 0.9540
0.0042 8.0 9384 0.0367 0.9528 0.9732 0.9493
0.0052 9.0 10557 0.0342 0.9563 0.9757 0.9531
0.0027 10.0 11730 0.0294 0.9629 0.9800 0.9595
0.0017 11.0 12903 0.0355 0.9605 0.9778 0.9582
0.0022 12.0 14076 0.0338 0.9627 0.9792 0.9591
0.0012 13.0 15249 0.0358 0.9609 0.9780 0.9591
0.0011 14.0 16422 0.0360 0.9618 0.9791 0.9604
0.0009 15.0 17595 0.0358 0.9648 0.9807 0.9625
0.0007 16.0 18768 0.0373 0.9627 0.9794 0.9595
0.0006 17.0 19941 0.0397 0.9597 0.9774 0.9574
0.0008 18.0 21114 0.0369 0.9652 0.9808 0.9621
0.0007 19.0 22287 0.0377 0.9646 0.9801 0.9621
0.0005 20.0 23460 0.0381 0.9639 0.9797 0.9616

Framework versions

  • Transformers 4.21.1
  • Pytorch 1.12.1+cu113
  • Datasets 2.4.0
  • Tokenizers 0.12.1
Downloads last month
505
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.

Spaces using has-abi/distilBERT-finetuned-resumes-sections 3