Edit model card

hing-mbert-ours-run-3

This model is a fine-tuned version of l3cube-pune/hing-mbert on an unknown dataset. It achieves the following results on the evaluation set:

  • Loss: 2.9769
  • Accuracy: 0.675
  • Precision: 0.6433
  • Recall: 0.6344
  • F1: 0.6344

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: 5e-05
  • train_batch_size: 16
  • eval_batch_size: 16
  • 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 Accuracy Precision Recall F1
0.9089 1.0 100 1.0993 0.635 0.6487 0.5304 0.5060
0.6657 2.0 200 0.8138 0.645 0.6550 0.6482 0.6234
0.3858 3.0 300 1.1334 0.665 0.6162 0.6061 0.5995
0.208 4.0 400 1.9041 0.685 0.6488 0.6169 0.6087
0.0996 5.0 500 2.3735 0.645 0.5867 0.5781 0.5794
0.0296 6.0 600 2.5772 0.665 0.6284 0.6208 0.6198
0.0623 7.0 700 2.8906 0.655 0.6040 0.5916 0.5926
0.0395 8.0 800 2.6567 0.675 0.6279 0.6254 0.6219
0.029 9.0 900 2.9277 0.655 0.6208 0.5950 0.5991
0.0194 10.0 1000 2.7362 0.665 0.6241 0.6208 0.6190
0.0092 11.0 1100 2.5561 0.68 0.6396 0.6401 0.6385
0.0059 12.0 1200 3.1112 0.675 0.6350 0.5967 0.6042
0.0133 13.0 1300 2.5269 0.685 0.6520 0.6607 0.6519
0.0051 14.0 1400 2.8736 0.68 0.6381 0.6158 0.6134
0.0044 15.0 1500 2.9132 0.675 0.6336 0.6180 0.6200
0.0029 16.0 1600 2.8701 0.675 0.6337 0.6214 0.6233
0.0015 17.0 1700 2.8115 0.69 0.6475 0.6388 0.6420
0.0019 18.0 1800 2.9517 0.67 0.6373 0.6276 0.6273
0.0013 19.0 1900 2.9633 0.67 0.6373 0.6276 0.6273
0.0007 20.0 2000 2.9769 0.675 0.6433 0.6344 0.6344

Framework versions

  • Transformers 4.25.1
  • Pytorch 1.13.0+cu116
  • Tokenizers 0.13.2
Downloads last month
14
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.

Model tree for SkyR/hing-mbert-ours-run-3

Finetuned
(5)
this model