Hugging Face's logo

Lyreck
/
deberta-v3-PII-ONNX

Token Classification
ONNX
deberta-v2
Generated from Trainer
Model card Files
xet
 Community
deberta-v3-PII-ONNX / README.md
Lyreck's picture
Lyreck
Update README.md
ff65eae verified
preview code
|
Raw
History Blame Contribute Delete
16.6 kB
metadata 
license: mit
base_model: microsoft/deberta-v3-base
tags:
  - generated_from_trainer
metrics:
  - precision
  - recall
  - f1
  - accuracy
model-index:
  - name: deberta-v3-base-pii-en
    results: []
pipeline_tag: token-classification
widget:
  - text: My name is Yoni Go and I live in Israel. My phone number is 054-1234567
inference:
  parameters:
    aggregation_strategy: first

Export of deberta-v3-base-pii-en

This is an onnx-converted version of yonigo/deberta-v3-base-pii-en. This is a test because onnx exports from here and there produce non-working models.

[IMPORTANT] I learned that deberta-V3 is not supported by default for ONNX conversion, apparently because it uses disentangled attention, which calculates relative position embeddings on the fly. When this graph is exported to ONNX and run inside a browser's WASM engine, certain operators can fail silently. In a nutshell, converting a DeBERTa v3 model to ONNX requires a bit more work than for other models like RoBERTa.

deberta-v3-base-pii-en

This model is a fine-tuned version of microsoft/deberta-v3-base on English samples from ai4privacy/pii-masking-300k.

Usage:

from transformers import pipeline

pipe = pipeline("token-classification", model="yonigo/deberta-v3-base-pii-en", aggregation_strategy="first")
pipe("My name is Yoni Go and I live in Israel. My phone number is 054-1234567")

training code git

Training results

Training Loss Epoch Step Validation Loss Bod F1 Building F1 Cardissuer F1 City F1 Country F1 Date F1 Driverlicense F1 Email F1 Geocoord F1 Givenname1 F1 Givenname2 F1 Idcard F1 Ip F1 Lastname1 F1 Lastname2 F1 Lastname3 F1 Pass F1 Passport F1 Postcode F1 Secaddress F1 Sex F1 Socialnumber F1 State F1 Street F1 Tel F1 Time F1 Title F1 Username F1 Precision Recall F1 Accuracy
0.2437 1.0695 1000 0.1168 0.9421 0.8791 0.0 0.8847 0.8841 0.8507 0.8617 0.9746 0.7903 0.5186 0.0 0.7928 0.9609 0.5720 0.0 0.0 0.9128 0.7991 0.8952 0.7145 0.8960 0.8583 0.8807 0.8816 0.9170 0.9390 0.7071 0.8946 0.8053 0.8619 0.8326 0.9736
0.0841 2.1390 2000 0.0731 0.9605 0.9633 0.0 0.9526 0.9399 0.8957 0.9035 0.9819 0.9245 0.7832 0.5095 0.9001 0.9664 0.6905 0.3578 0.0 0.9417 0.8973 0.9628 0.9651 0.9592 0.9111 0.9699 0.9631 0.9419 0.9631 0.9382 0.9388 0.8858 0.9315 0.9081 0.9826
0.0592 3.2086 3000 0.0544 0.9675 0.9787 0.0 0.9630 0.9524 0.9192 0.9337 0.9844 0.9457 0.8391 0.7142 0.9139 0.9862 0.7777 0.5887 0.2644 0.9573 0.9166 0.9743 0.9682 0.9680 0.9437 0.9787 0.9420 0.9698 0.9674 0.9516 0.9491 0.9168 0.9492 0.9327 0.9874
0.0436 4.2781 4000 0.0488 0.9673 0.9821 0.0 0.9679 0.9709 0.9187 0.9487 0.9836 0.9722 0.8580 0.7335 0.9322 0.9912 0.7998 0.6667 0.5722 0.9432 0.9371 0.9791 0.9778 0.9705 0.9548 0.9831 0.9701 0.9680 0.9673 0.9543 0.9584 0.9331 0.9529 0.9429 0.9888
0.037 5.3476 5000 0.0518 0.9653 0.9811 0.0 0.9671 0.9660 0.9052 0.9392 0.9859 0.9745 0.8469 0.7616 0.9225 0.9873 0.8108 0.7059 0.6450 0.9578 0.9437 0.9772 0.9774 0.9715 0.9511 0.9827 0.9645 0.9681 0.9639 0.9617 0.9556 0.9283 0.9574 0.9426 0.9883
0.028 6.4171 6000 0.0488 0.9624 0.9842 0.0 0.9709 0.9732 0.9112 0.9437 0.9869 0.9767 0.8614 0.7818 0.9322 0.9860 0.8266 0.7344 0.7080 0.9518 0.9509 0.9797 0.9802 0.9768 0.9564 0.9831 0.9756 0.9717 0.9714 0.9610 0.9537 0.9383 0.9577 0.9479 0.9891
0.0238 7.4866 7000 0.0483 0.9625 0.9844 0.0 0.9705 0.9732 0.9144 0.9360 0.9804 0.9814 0.8654 0.7707 0.9328 0.9885 0.8234 0.7253 0.6873 0.9504 0.9372 0.9787 0.9750 0.9753 0.9523 0.9848 0.9755 0.9717 0.9730 0.9677 0.9567 0.9379 0.9554 0.9466 0.9893
0.0197 8.5561 8000 0.0517 0.9651 0.9857 0.0 0.9735 0.9735 0.9100 0.9579 0.9858 0.9679 0.8630 0.7748 0.9375 0.9858 0.8229 0.7259 0.6764 0.9496 0.9571 0.9800 0.9780 0.9753 0.9543 0.9829 0.9769 0.9763 0.9725 0.9637 0.9628 0.9409 0.9579 0.9493 0.9892
0.0164 9.6257 9000 0.0536 0.9642 0.9859 0.0 0.9707 0.9628 0.9175 0.9533 0.9857 0.9814 0.8617 0.7674 0.9377 0.9869 0.8193 0.7331 0.7110 0.9471 0.9535 0.9818 0.9756 0.9758 0.9602 0.9829 0.9746 0.9762 0.9710 0.9631 0.9596 0.9396 0.9581 0.9488 0.9893
0.0148 10.6952 10000 0.0545 0.9676 0.9849 0.0 0.9728 0.9741 0.9351 0.9563 0.9833 0.9791 0.8693 0.7877 0.9351 0.9863 0.8294 0.7536 0.7332 0.9609 0.9523 0.9808 0.9809 0.9775 0.9514 0.9837 0.9791 0.9713 0.9707 0.9624 0.9602 0.9435 0.9588 0.9511 0.9897
0.0115 11.7647 11000 0.0546 0.9661 0.9849 0.0 0.9757 0.9661 0.9133 0.9579 0.9800 0.9769 0.8661 0.7935 0.9439 0.9894 0.8292 0.7485 0.7126 0.9513 0.9607 0.9793 0.9815 0.9770 0.9581 0.9851 0.9803 0.9711 0.9645 0.9672 0.9588 0.9413 0.9597 0.9504 0.9896
0.0101 12.8342 12000 0.0573 0.9634 0.9861 0.0 0.9742 0.9693 0.9234 0.9574 0.9850 0.9837 0.8602 0.7854 0.9391 0.9898 0.8220 0.7470 0.7056 0.9515 0.9586 0.9834 0.9803 0.9787 0.9617 0.9841 0.9773 0.9753 0.9691 0.9649 0.9594 0.9459 0.9560 0.9509 0.9898
0.0084 13.9037 13000 0.0597 0.9657 0.9861 0.0 0.9761 0.9733 0.9136 0.9542 0.9828 0.9813 0.8672 0.7989 0.9418 0.9889 0.8326 0.7458 0.7409 0.9556 0.9573 0.9815 0.9797 0.9772 0.9616 0.9866 0.9810 0.9784 0.9644 0.9658 0.9609 0.9467 0.9568 0.9517 0.9897
0.0065 14.9733 14000 0.0621 0.9684 0.9859 0.0 0.9726 0.9741 0.9277 0.9539 0.9789 0.9814 0.8696 0.7879 0.9348 0.9868 0.8368 0.7542 0.7456 0.9487 0.9543 0.9805 0.9809 0.9780 0.9582 0.9863 0.9801 0.9763 0.9716 0.9629 0.9580 0.9439 0.9590 0.9514 0.9896
0.0059 16.0428 15000 0.0613 0.9679 0.9874 0.0 0.9770 0.9694 0.9347 0.9621 0.9786 0.9791 0.8723 0.7857 0.9403 0.9891 0.8414 0.7594 0.7371 0.9508 0.9595 0.9813 0.9797 0.9775 0.9562 0.9856 0.9790 0.9805 0.9725 0.9677 0.9554 0.9457 0.9609 0.9532 0.9901
0.005 17.1123 16000 0.0639 0.9693 0.9839 0.0 0.9781 0.9735 0.9264 0.9631 0.9827 0.9791 0.8731 0.7996 0.9437 0.9869 0.8406 0.7714 0.7593 0.9547 0.9559 0.9813 0.9809 0.9782 0.9502 0.9849 0.9810 0.9795 0.9731 0.9654 0.9617 0.9460 0.9616 0.9537 0.9901
0.0038 18.1818 17000 0.0651 0.9681 0.9869 0.0 0.9785 0.9747 0.9311 0.9606 0.9831 0.9837 0.8749 0.7899 0.9366 0.9889 0.8331 0.7520 0.7230 0.9582 0.9596 0.9805 0.9802 0.9784 0.9609 0.9858 0.9805 0.9800 0.9756 0.9663 0.9614 0.9494 0.9586 0.9540 0.9902
0.0035 19.2513 18000 0.0716 0.9661 0.9857 0.0 0.9791 0.9715 0.9319 0.9607 0.9829 0.9791 0.8707 0.8026 0.9385 0.9859 0.8354 0.7557 0.7374 0.9564 0.9580 0.9795 0.9803 0.9767 0.9563 0.9871 0.9823 0.9750 0.9745 0.9654 0.9574 0.9450 0.9610 0.9529 0.9896
0.0023 20.3209 19000 0.0682 0.9686 0.9857 0.0 0.9789 0.9755 0.9310 0.9621 0.9850 0.9837 0.8777 0.7974 0.9430 0.9880 0.8424 0.7600 0.7545 0.9566 0.9628 0.9813 0.9773 0.9765 0.9620 0.9863 0.9813 0.9743 0.9742 0.9660 0.9567 0.9474 0.9622 0.9548 0.9901
0.002 21.3904 20000 0.0727 0.9696 0.9857 0.0 0.9759 0.9742 0.9315 0.9636 0.9814 0.9814 0.8791 0.8011 0.9427 0.9898 0.8383 0.7556 0.7419 0.9588 0.9575 0.9826 0.9756 0.9756 0.9519 0.9853 0.9802 0.9733 0.9749 0.9618 0.9569 0.9459 0.9614 0.9536 0.9900
0.002 22.4599 21000 0.0756 0.9690 0.9859 0.0 0.9770 0.9752 0.9225 0.9626 0.9829 0.9814 0.8734 0.7850 0.9417 0.9878 0.8312 0.7560 0.7405 0.9570 0.9591 0.9805 0.9814 0.9768 0.9614 0.9858 0.9795 0.9758 0.9700 0.9643 0.9596 0.9452 0.9610 0.9530 0.9898
0.0014 23.5294 22000 0.0746 0.9694 0.9874 0.0 0.9779 0.9749 0.9309 0.9653 0.9869 0.9814 0.8739 0.8019 0.9439 0.9884 0.8365 0.7684 0.7559 0.9570 0.9587 0.9808 0.9797 0.9777 0.9570 0.9861 0.9807 0.9758 0.9732 0.9658 0.9625 0.9476 0.9621 0.9548 0.9902
0.0012 24.5989 23000 0.0762 0.9696 0.9864 1.0 0.9784 0.9761 0.9389 0.9666 0.9844 0.9814 0.8718 0.7860 0.9440 0.9880 0.8296 0.7611 0.7513 0.9579 0.9621 0.9826 0.9797 0.9777 0.9584 0.9856 0.9810 0.9757 0.9754 0.9673 0.9615 0.9484 0.9614 0.9548 0.9902
0.0011 25.6684 24000 0.0744 0.9698 0.9862 0.0 0.9783 0.9785 0.9353 0.9666 0.9832 0.9837 0.8775 0.8007 0.9454 0.9878 0.8417 0.7705 0.7680 0.9592 0.9629 0.9821 0.9785 0.9785 0.9625 0.9858 0.9831 0.9779 0.9756 0.9688 0.9615 0.9502 0.9626 0.9564 0.9904
0.001 26.7380 25000 0.0750 0.9702 0.9869 1.0 0.9803 0.9752 0.9335 0.9666 0.9831 0.9791 0.8836 0.8048 0.9451 0.9871 0.8435 0.7724 0.7708 0.9589 0.9624 0.9821 0.9774 0.9782 0.9605 0.9871 0.9828 0.9760 0.9762 0.9653 0.9611 0.9499 0.9630 0.9564 0.9904
0.0009 27.8075 26000 0.0764 0.9695 0.9877 1.0 0.9798 0.9767 0.9379 0.9647 0.9825 0.9746 0.8781 0.7989 0.9463 0.9878 0.8417 0.7605 0.7708 0.9595 0.9626 0.9826 0.9780 0.9782 0.9609 0.9881 0.9810 0.974 0.9761 0.9663 0.9606 0.9493 0.9627 0.9560 0.9903
0.0008 28.8770 27000 0.0767 0.9699 0.9867 1.0 0.9788 0.9773 0.9356 0.9654 0.9844 0.9746 0.8798 0.7989 0.9441 0.9880 0.8411 0.7651 0.7676 0.9603 0.9627 0.9815 0.9797 0.9782 0.9592 0.9881 0.9815 0.9765 0.9767 0.9666 0.9604 0.9496 0.9626 0.9561 0.9904
0.0008 29.9465 28000 0.0765 0.9707 0.9869 1.0 0.9785 0.9773 0.9381 0.9671 0.9835 0.9746 0.8823 0.7963 0.9426 0.9857 0.8426 0.7621 0.7708 0.9610 0.9621 0.9815 0.9791 0.9782 0.9590 0.9878 0.9815 0.9760 0.9770 0.9668 0.9606 0.9501 0.9623 0.9562 0.9904
0.0009 31.0160 29000 0.0768 0.9705 0.9867 1.0 0.9781 0.9773 0.9376 0.9641 0.9852 0.9769 0.8812 0.7989 0.9441 0.9875 0.8435 0.7673 0.7676 0.9610 0.9623 0.9820 0.9791 0.9782 0.9617 0.9871 0.9818 0.9767 0.9767 0.9668 0.9608 0.9504 0.9626 0.9565 0.9904
0.0007 32.0856 30000 0.0767 0.9705 0.9869 1.0 0.9781 0.9773 0.9374 0.9645 0.9850 0.9769 0.8810 0.7996 0.9443 0.9873 0.8433 0.7641 0.7696 0.9603 0.9619 0.9820 0.9791 0.9782 0.9615 0.9878 0.9815 0.9767 0.9762 0.9668 0.9606 0.9504 0.9625 0.9564 0.9904

Framework versions

  • Transformers 4.41.2
  • Pytorch 2.3.1+cu121
  • Datasets 2.20.0
  • Tokenizers 0.19.1