Edit model card

indic-bert-finetuned-code-mixed-DS

This model is a fine-tuned version of ai4bharat/indic-bert on an unknown dataset. It achieves the following results on the evaluation set:

  • Loss: 0.8647
  • Accuracy: 0.5795
  • Precision: 0.5485
  • Recall: 0.5287
  • F1: 0.4391

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: 1e-06
  • train_batch_size: 16
  • eval_batch_size: 16
  • seed: 43
  • 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
1.0937 2.0 497 1.0813 0.3602 0.3587 0.4257 0.2834
1.0189 3.99 994 0.9482 0.5493 0.3887 0.5246 0.4080
0.9208 5.99 1491 0.9002 0.5714 0.3813 0.5292 0.4170
0.8803 7.98 1988 0.8758 0.5654 0.3889 0.5300 0.4159
0.8482 9.98 2485 0.8657 0.5795 0.3867 0.5365 0.4228
0.8293 11.98 2982 0.8734 0.5835 0.3796 0.5298 0.4214
0.8131 13.97 3479 0.8567 0.5835 0.5018 0.5414 0.4350
0.8 15.97 3976 0.8547 0.5835 0.5610 0.5460 0.4361
0.7933 17.96 4473 0.8650 0.5775 0.5317 0.5252 0.4373
0.7835 19.96 4970 0.8647 0.5795 0.5485 0.5287 0.4391

Framework versions

  • Transformers 4.20.1
  • Pytorch 1.10.1+cu111
  • Datasets 2.3.2
  • Tokenizers 0.12.1
Downloads last month
24
Safetensors
Model size
33.4M params
Tensor type
I64
·
F32
·
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.