Edit model card

hing-mbert-ours-run-1

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: 2.9965
  • Accuracy: 0.665
  • Precision: 0.6151
  • Recall: 0.6082
  • F1: 0.6090

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
0.9757 1.0 100 0.7526 0.665 0.6358 0.6149 0.5854
0.7227 2.0 200 1.1062 0.69 0.6679 0.6031 0.6025
0.4345 3.0 300 1.2601 0.67 0.6512 0.6031 0.6001
0.2738 4.0 400 1.4485 0.67 0.6333 0.5968 0.6050
0.1171 5.0 500 1.9132 0.65 0.6216 0.6170 0.5944
0.0941 6.0 600 1.8293 0.685 0.6420 0.6439 0.6409
0.0348 7.0 700 2.3249 0.675 0.6424 0.6386 0.6384
0.0317 8.0 800 2.4134 0.67 0.6171 0.6120 0.6128
0.0056 9.0 900 2.6733 0.68 0.6343 0.6313 0.6300
0.0095 10.0 1000 2.5950 0.685 0.6318 0.6289 0.6295
0.0081 11.0 1100 2.3885 0.69 0.6407 0.6434 0.6410
0.023 12.0 1200 2.4087 0.67 0.6206 0.6231 0.6212
0.0054 13.0 1300 2.4516 0.675 0.6229 0.6227 0.6128
0.0047 14.0 1400 2.6152 0.68 0.6285 0.6256 0.6263
0.0063 15.0 1500 2.8077 0.69 0.6498 0.6281 0.6309
0.0028 16.0 1600 2.7084 0.675 0.6254 0.6214 0.6207
0.0025 17.0 1700 2.8360 0.67 0.6175 0.6128 0.6145
0.0011 18.0 1800 2.8591 0.655 0.6001 0.5958 0.5971
0.0005 19.0 1900 2.9419 0.665 0.6151 0.6082 0.6090
0.0002 20.0 2000 2.9965 0.665 0.6151 0.6082 0.6090

Framework versions

  • Transformers 4.25.1
  • Pytorch 1.13.0+cu116
  • Tokenizers 0.13.2
Downloads last month
8
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-1

Finetuned
(5)
this model