Kalbert – AI Model by Chakita | AlphaNeural AI

You can deploy this model and start earning money today!

Chakita/
Kalbert

like

0

generated_from_trainer

autotrain_compatible

endpoints_compatible

Views

No views yet

Kalbert

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

Loss: 1.5324

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: 2e-05
train_batch_size: 4
eval_batch_size: 4
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
num_epochs: 10
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
1.5835	1.0	3953	1.7985
1.6098	2.0	7906	1.7434
1.5266	3.0	11859	1.6934
1.5179	4.0	15812	1.6665
1.5459	5.0	19765	1.6135
1.5511	6.0	23718	1.6002
1.5209	7.0	27671	1.5657
1.5413	8.0	31624	1.5578
1.4828	9.0	35577	1.5465
1.4651	10.0	39530	1.5451

Framework versions

Transformers 4.25.1
Pytorch 1.13.0+cu116
Datasets 2.8.0
Tokenizers 0.13.2