Edit model card

whisper-small-multilingual-naija-10-25-2024

This model is a fine-tuned version of ccibeekeoc42/whisper-small-yoruba-07-17 on an unknown dataset. It achieves the following results on the evaluation set:

  • Loss: 0.7577

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-05
  • train_batch_size: 16
  • eval_batch_size: 8
  • seed: 42
  • optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_steps: 500
  • num_epochs: 1
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss
1.9909 0.0229 100 1.5086
1.6658 0.0457 200 1.1909
1.421 0.0686 300 1.0778
1.3603 0.0914 400 1.0138
1.3567 0.1143 500 0.9721
1.2484 0.1372 600 0.9448
1.1076 0.1600 700 0.9270
1.0729 0.1829 800 0.9088
1.042 0.2058 900 0.8900
1.0528 0.2286 1000 0.8817
1.0261 0.2515 1100 0.8674
0.9289 0.2743 1200 0.8631
0.959 0.2972 1300 0.8460
0.9355 0.3201 1400 0.8436
0.9855 0.3429 1500 0.8351
0.9426 0.3658 1600 0.8291
0.8913 0.3887 1700 0.8233
0.9202 0.4115 1800 0.8180
0.9122 0.4344 1900 0.8131
0.8454 0.4572 2000 0.8104
0.8048 0.4801 2100 0.8074
0.8824 0.5030 2200 0.8006
0.8707 0.5258 2300 0.7965
0.8955 0.5487 2400 0.7941
0.8237 0.5716 2500 0.7940
0.8774 0.5944 2600 0.7921
0.8162 0.6173 2700 0.7836
0.8308 0.6401 2800 0.7829
0.7863 0.6630 2900 0.7786
0.7536 0.6859 3000 0.7744
0.8215 0.7087 3100 0.7730
0.7852 0.7316 3200 0.7709
0.7569 0.7545 3300 0.7699
0.7298 0.7773 3400 0.7685
0.7777 0.8002 3500 0.7659
0.7358 0.8230 3600 0.7637
0.7258 0.8459 3700 0.7611
0.7674 0.8688 3800 0.7604
0.8048 0.8916 3900 0.7599
0.7694 0.9145 4000 0.7590
0.8072 0.9374 4100 0.7577
0.7765 0.9602 4200 0.7580
0.7789 0.9831 4300 0.7577

Framework versions

  • Transformers 4.46.0
  • Pytorch 2.0.1+cu118
  • Datasets 3.0.2
  • Tokenizers 0.20.1
Downloads last month
38
Safetensors
Model size
242M params
Tensor type
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.

Model tree for okezieowen/whisper-small-multilingual-naija-10-25-2024

Finetuned
(2)
this model