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.

W2V2_Bert_BIG_C_Bemba_50hr_v1

This model is a fine-tuned version of facebook/w2v-bert-2.0 on an unknown dataset. It achieves the following results on the evaluation set:

  • Loss: 0.6150
  • Wer: 0.3709
  • Cer: 0.0976

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-06
  • train_batch_size: 2
  • eval_batch_size: 2
  • seed: 42
  • gradient_accumulation_steps: 2
  • total_train_batch_size: 4
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: cosine
  • lr_scheduler_warmup_ratio: 0.05
  • num_epochs: 100

Training results

Training Loss Epoch Step Validation Loss Wer Cer
1.9873 1.0 6423 0.8901 0.6111 0.1723
0.7807 2.0 12846 0.8360 0.5188 0.1525
0.683 3.0 19269 0.6705 0.4842 0.1410
0.6282 4.0 25692 0.6473 0.4762 0.1394
0.594 5.0 32115 0.6369 0.4463 0.1314
0.5645 6.0 38538 0.6244 0.4360 0.1287
0.5322 7.0 44961 0.6186 0.4191 0.1273
0.5045 8.0 51384 0.6334 0.4127 0.1230
0.4767 9.0 57807 0.6017 0.4117 0.1227
0.4505 10.0 64230 0.6142 0.4092 0.1214
0.4247 11.0 70653 0.6155 0.4033 0.1208
0.3974 12.0 77076 0.6161 0.4013 0.1198
0.3714 13.0 83499 0.6415 0.4032 0.1211
0.3437 14.0 89922 0.6691 0.4007 0.1207
0.3175 15.0 96345 0.7251 0.4052 0.1212
0.2921 16.0 102768 0.7279 0.4003 0.1218
0.2681 17.0 109191 0.7837 0.4103 0.1216
0.2455 18.0 115614 0.8336 0.4074 0.1233
0.2242 19.0 122037 0.8544 0.4158 0.1247
0.2044 20.0 128460 0.8591 0.4243 0.1270
0.1857 21.0 134883 0.9652 0.4123 0.1245
0.1676 22.0 141306 1.0143 0.4254 0.1266

Framework versions

  • Transformers 4.44.2
  • Pytorch 2.2.0+cu121
  • Datasets 3.0.0
  • Tokenizers 0.19.1
Downloads last month
2
Safetensors
Model size
606M 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/W2V2_Bert_BIG_C_Bemba_50hr_v1

Finetuned
(238)
this model

Collection including asr-africa/W2V2_Bert_BIG_C_Bemba_50hr_v1