Edit model card

wav2vec2-large-sw-cv-100hr-v5

This model is a fine-tuned version of facebook/wav2vec2-large on an unknown dataset. It achieves the following results on the evaluation set:

  • Loss: 0.8148
  • Model Preparation Time: 0.0038
  • Wer: 0.5557
  • Cer: 0.2034

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: 0.0007
  • train_batch_size: 32
  • eval_batch_size: 8
  • seed: 42
  • gradient_accumulation_steps: 2
  • total_train_batch_size: 64
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_ratio: 0.033
  • num_epochs: 100
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Model Preparation Time Wer Cer
1.5047 1.0 1040 0.5420 0.0038 0.5182 0.1481
0.4317 2.0 2080 0.6919 0.0038 0.6485 0.2046
1.7978 3.0 3120 2.9072 0.0038 1.0 1.0
2.8626 4.0 4160 2.8802 0.0038 1.0 1.0
2.8609 5.0 5200 2.8577 0.0038 1.0 1.0
2.861 6.0 6240 2.8602 0.0038 1.0 1.0
2.8602 7.0 7280 2.8747 0.0038 1.0 1.0
2.8603 8.0 8320 2.8618 0.0038 1.0 1.0
2.8597 9.0 9360 2.8668 0.0038 1.0 1.0
2.8598 10.0 10400 2.8657 0.0038 1.0 1.0
2.8598 11.0 11440 2.8603 0.0038 1.0 1.0

Framework versions

  • Transformers 4.44.0
  • Pytorch 2.1.0+cu118
  • Datasets 2.20.0
  • Tokenizers 0.19.1
Downloads last month
0
Safetensors
Model size
315M params
Tensor type
F32
·
Inference API
Unable to determine this model's library. Check the docs .

Model tree for KasuleTrevor/wav2vec2-large-sw-cv-100hr-v5

Finetuned
(16)
this model