You need to agree to share your contact information to access this model

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

Log in or Sign Up to review the conditions and access this model content.

Whisper Small Bemba - Beijuka Bruno

This model is a fine-tuned version of openai/whisper-small on the BEMBA dataset. It achieves the following results on the evaluation set:

  • Loss: 1.2836
  • Model Preparation Time: 0.0075
  • Wer: 0.4233
  • Cer: 0.1171

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: 4
  • eval_batch_size: 8
  • seed: 42
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_steps: 500
  • num_epochs: 100

Training results

Training Loss Epoch Step Validation Loss Model Preparation Time Wer Cer
0.9215 1.0 2546 0.6730 0.0075 0.4818 0.1286
0.5391 2.0 5092 0.6217 0.0075 0.4434 0.1188
0.366 3.0 7638 0.6423 0.0075 0.4449 0.1222
0.2213 4.0 10184 0.6881 0.0075 0.4430 0.1281
0.1152 5.0 12730 0.7444 0.0075 0.4607 0.1329
0.059 6.0 15276 0.8259 0.0075 0.4572 0.1283
0.0336 7.0 17822 0.8798 0.0075 0.4313 0.1186
0.0233 8.0 20368 0.9262 0.0075 0.4321 0.1196
0.0182 9.0 22914 0.9683 0.0075 0.4333 0.1230
0.0159 10.0 25460 0.9992 0.0075 0.4301 0.1215
0.0123 11.0 28006 1.0515 0.0075 0.4279 0.1191
0.0114 12.0 30552 1.0733 0.0075 0.4296 0.1154
0.0105 13.0 33098 1.0854 0.0075 0.4277 0.1171
0.0096 14.0 35644 1.1244 0.0075 0.4228 0.1156
0.0084 15.0 38190 1.1250 0.0075 0.4284 0.1201
0.0082 16.0 40736 1.1691 0.0075 0.4163 0.1142
0.0075 17.0 43282 1.1766 0.0075 0.4172 0.1140
0.0073 18.0 45828 1.1734 0.0075 0.4323 0.1197
0.0067 19.0 48374 1.1985 0.0075 0.4201 0.1170
0.006 20.0 50920 1.2099 0.0075 0.4297 0.1192
0.0057 21.0 53466 1.2517 0.0075 0.4293 0.1197
0.0058 22.0 56012 1.2351 0.0075 0.4292 0.1202
0.0057 23.0 58558 1.2380 0.0075 0.4362 0.1195
0.0048 24.0 61104 1.2756 0.0075 0.4202 0.1165
0.0049 25.0 63650 1.2816 0.0075 0.4185 0.1156
0.0046 26.0 66196 1.3191 0.0075 0.4116 0.1115
0.0041 27.0 68742 1.3041 0.0075 0.4187 0.1160
0.0041 28.0 71288 1.3195 0.0075 0.4193 0.1144
0.0042 29.0 73834 1.3229 0.0075 0.4215 0.1187
0.0033 30.0 76380 1.3282 0.0075 0.4192 0.1159
0.0035 31.0 78926 1.3424 0.0075 0.4256 0.1196
0.0033 32.0 81472 1.3538 0.0075 0.4217 0.1164
0.0072 33.0 84018 1.3673 0.0075 0.4141 0.1148
0.0041 34.0 86564 1.3853 0.0075 0.4195 0.1154
0.0035 35.0 89110 1.3809 0.0075 0.4145 0.1146
0.0034 36.0 91656 1.3930 0.0075 0.4170 0.1158

Framework versions

  • Transformers 4.44.2
  • Pytorch 2.1.0+cu118
  • Datasets 3.0.0
  • Tokenizers 0.19.1
Downloads last month
0
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 asr-africa/whisper_BIG-C_BEMBA_20hr_v2

Finetuned
(2147)
this model

Evaluation results