Edit model card

hing-mbert-ours-run-4

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: 3.0173
  • Accuracy: 0.68
  • Precision: 0.6330
  • Recall: 0.6325
  • F1: 0.6320

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
0.9781 1.0 100 0.8852 0.55 0.4044 0.5284 0.4211
0.7568 2.0 200 0.8110 0.655 0.5994 0.6013 0.5762
0.5121 3.0 300 0.9735 0.65 0.6145 0.6131 0.5965
0.314 4.0 400 1.1324 0.65 0.6305 0.6355 0.6266
0.1298 5.0 500 2.8247 0.61 0.5804 0.5087 0.5092
0.1013 6.0 600 2.8183 0.635 0.6212 0.5674 0.5667
0.0989 7.0 700 2.3235 0.635 0.5944 0.5922 0.5916
0.0481 8.0 800 2.5240 0.68 0.6334 0.6172 0.6221
0.018 9.0 900 2.6782 0.65 0.6123 0.6054 0.6062
0.0285 10.0 1000 2.3400 0.67 0.6206 0.6327 0.6189
0.014 11.0 1100 2.6558 0.66 0.6098 0.5992 0.6018
0.0085 12.0 1200 2.9366 0.66 0.6076 0.5961 0.5991
0.0106 13.0 1300 2.8567 0.665 0.6198 0.6193 0.6186
0.0097 14.0 1400 3.1526 0.64 0.6089 0.5975 0.5954
0.0022 15.0 1500 2.7305 0.69 0.6404 0.6404 0.6398
0.0016 16.0 1600 2.7670 0.69 0.6418 0.6434 0.6425
0.0017 17.0 1700 2.8193 0.7 0.6533 0.6566 0.6546
0.0009 18.0 1800 2.9959 0.685 0.6400 0.6389 0.6384
0.0006 19.0 1900 3.0153 0.68 0.6330 0.6325 0.6320
0.0005 20.0 2000 3.0173 0.68 0.6330 0.6325 0.6320

Framework versions

  • Transformers 4.25.1
  • Pytorch 1.13.0+cu116
  • Tokenizers 0.13.2
Downloads last month
7
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-4

Finetuned
(5)
this model