hing-mbert-ours-run-2

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: 3.3919
  • Accuracy: 0.62
  • Precision: 0.5759
  • Recall: 0.5631
  • F1: 0.5669

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
1.0284 1.0 100 1.2914 0.595 0.5712 0.4800 0.4642
0.8127 2.0 200 0.8552 0.59 0.5744 0.5675 0.4891
0.5499 3.0 300 1.1212 0.645 0.6544 0.5600 0.5475
0.3433 4.0 400 1.2017 0.605 0.5872 0.5866 0.5791
0.2218 5.0 500 1.8329 0.655 0.6458 0.6064 0.6055
0.1763 6.0 600 2.4194 0.655 0.6140 0.5802 0.5871
0.1022 7.0 700 2.3894 0.66 0.6171 0.6045 0.6048
0.0631 8.0 800 2.8259 0.605 0.5704 0.5255 0.5259
0.0254 9.0 900 2.9135 0.65 0.6013 0.5734 0.5784
0.0316 10.0 1000 3.0548 0.62 0.5862 0.5650 0.5670
0.026 11.0 1100 3.1020 0.62 0.5722 0.5593 0.5619
0.0152 12.0 1200 3.0692 0.62 0.5685 0.5597 0.5621
0.0156 13.0 1300 3.1068 0.615 0.5771 0.5589 0.5624
0.0237 14.0 1400 3.3487 0.62 0.5924 0.5589 0.5642
0.0094 15.0 1500 3.2007 0.615 0.5665 0.5639 0.5650
0.0054 16.0 1600 3.2838 0.62 0.5807 0.5657 0.5690
0.005 17.0 1700 3.2258 0.615 0.5846 0.5723 0.5747
0.005 18.0 1800 3.3572 0.63 0.5827 0.5698 0.5736
0.0022 19.0 1900 3.3642 0.62 0.5759 0.5631 0.5669
0.0019 20.0 2000 3.3919 0.62 0.5759 0.5631 0.5669

Framework versions

  • Transformers 4.25.1
  • Pytorch 1.13.0+cu116
  • Tokenizers 0.13.2
Downloads last month
103
Inference Providers NEW
This model is not currently available via any of the supported Inference Providers.

Model tree for SkyR/hing-mbert-ours-run-2

Finetuned
(5)
this model