metadata

license: cc-by-nc-4.0
base_model: facebook/wav2vec2-large-uralic-voxpopuli-v2
tags:
  - generated_from_trainer
metrics:
  - wer
model-index:
  - name: w2v2-large-uralic-voxpopuli-v2-karelian-cs-w-rus
    results: []

w2v2-large-uralic-voxpopuli-v2-karelian-cs-w-rus

This model is a fine-tuned version of facebook/wav2vec2-large-uralic-voxpopuli-v2 on the None dataset. It achieves the following results on the evaluation set:

Loss: 0.5997
Wer: 0.4401
Cer: 0.1228

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: 0.0001
train_batch_size: 8
eval_batch_size: 8
seed: 43
gradient_accumulation_steps: 4
total_train_batch_size: 32
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 500
training_steps: 10000

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
6.2662	0.48	500	1.8324	0.9998	0.6007
1.032	0.97	1000	0.7219	0.6129	0.1779
0.6635	1.45	1500	0.6393	0.5535	0.1572
0.5568	1.94	2000	0.5816	0.5192	0.1486
0.4728	2.42	2500	0.5838	0.5076	0.1444
0.427	2.9	3000	0.5746	0.4915	0.1377
0.3789	3.39	3500	0.5658	0.4781	0.1342
0.3543	3.87	4000	0.5859	0.4737	0.1349
0.3223	4.35	4500	0.5678	0.4708	0.1327
0.3053	4.84	5000	0.5572	0.4652	0.1311
0.2852	5.32	5500	0.5768	0.4594	0.1285
0.2663	5.81	6000	0.5687	0.4594	0.1282
0.2585	6.29	6500	0.5891	0.4572	0.1280
0.2477	6.77	7000	0.5930	0.4524	0.1257
0.2341	7.26	7500	0.5841	0.4456	0.1244
0.2325	7.74	8000	0.5973	0.4460	0.1242
0.2195	8.22	8500	0.6069	0.4403	0.1234
0.2187	8.71	9000	0.5899	0.4390	0.1229
0.2134	9.19	9500	0.5944	0.4398	0.1231
0.2111	9.68	10000	0.5997	0.4401	0.1228

Framework versions

Transformers 4.33.3
Pytorch 2.5.1
Datasets 2.15.0
Tokenizers 0.13.3