W2V2_Bert_Afrivoice_FLEURS_Shona_100hr_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: 0.2169
• Model Preparation Time: 0.0117
• Wer: 0.2047
• Cer: 0.0377

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-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: cosine
• lr_scheduler_warmup_ratio: 0.025
• num_epochs: 100

Training results

Training Loss	Epoch	Step	Validation Loss	Model Preparation Time	Wer	Cer
0.0958	1.0	4449	0.1764	0.0117	0.2148	0.0352
0.093	2.0	8898	0.1879	0.0117	0.2282	0.0382
0.0993	3.0	13347	0.1860	0.0117	0.2285	0.0384
0.0971	4.0	17796	0.1906	0.0117	0.2379	0.0390
0.0912	5.0	22245	0.1843	0.0117	0.2268	0.0381
0.084	6.0	26694	0.1970	0.0117	0.2247	0.0376
0.0786	7.0	31143	0.2031	0.0117	0.2433	0.0419
0.0716	8.0	35592	0.2114	0.0117	0.2360	0.0394
0.0673	9.0	40041	0.2146	0.0117	0.2358	0.0388
0.0617	10.0	44490	0.2282	0.0117	0.2311	0.0384
0.0559	11.0	48939	0.2302	0.0117	0.2306	0.0386

Framework versions

• Transformers 4.45.2
• Pytorch 2.1.0+cu118
• Datasets 3.0.1
• Tokenizers 0.20.1