wav2vec2-xls-r-300m-scandinavian-E4-100h-30-epochs-20250201_v1.2

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.8034
• Wer: 87.9662
• Cer: 25.6683

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: 16
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 5000
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
3.5002	0.7788	1000	3.3325	99.9561	99.1218
2.9685	1.5576	2000	2.9292	99.9551	99.1218
0.8981	2.3364	3000	0.6540	51.2661	14.4010
0.7203	3.1153	4000	0.3972	35.4549	9.8695
0.6642	3.8941	5000	0.3259	29.5895	8.1753
0.6634	4.6729	6000	0.2929	26.4860	7.3066
0.5613	5.4517	7000	0.4702	31.8866	8.7770
1.3205	6.2305	8000	0.9818	76.5633	29.0254
1.304	7.0093	9000	0.9924	99.5143	42.4246
1.1542	7.7882	10000	0.7604	56.7837	14.0385
0.9404	8.5670	11000	0.7283	49.1434	13.0316
0.9964	9.3458	12000	0.7189	46.0305	11.7982
1.2419	10.1246	13000	0.7556	65.1854	17.5888
1.2285	10.9034	14000	0.8025	89.1539	26.5302
1.0388	11.6822	15000	0.8034	87.9672	25.6645
1.0672	12.4611	16000	0.8034	87.9662	25.6671
1.2088	13.2399	17000	0.8034	87.9651	25.6707
1.0838	14.0187	18000	0.8034	87.9672	25.6647
1.0981	14.7975	19000	0.8034	87.9672	25.6644
1.0798	15.5763	20000	0.8034	87.9662	25.6653
1.2961	16.3551	21000	0.8034	87.9703	25.6658
1.155	17.1340	22000	0.8034	87.9682	25.6669
1.151	17.9128	23000	0.8034	87.9682	25.6645
1.0612	18.6916	24000	0.8034	87.9630	25.6674
1.2208	19.4704	25000	0.8034	87.9630	25.6665
1.1297	20.2492	26000	0.8034	87.9724	25.6642
1.0827	21.0280	27000	0.8034	87.9620	25.6647
1.0807	21.8069	28000	0.8034	87.9651	25.6649
1.2411	22.5857	29000	0.8034	87.9703	25.6673
1.2018	23.3645	30000	0.8034	87.9672	25.6656
1.0622	24.1433	31000	0.8034	87.9557	25.6655
1.1159	24.9221	32000	0.8034	87.9724	25.6676
1.2522	25.7009	33000	0.8034	87.9703	25.6664
1.1681	26.4798	34000	0.8034	87.9693	25.6674
1.1144	27.2586	35000	0.8034	87.9641	25.6669
1.1385	28.0374	36000	0.8034	87.9641	25.6665
1.147	28.8162	37000	0.8034	87.9651	25.6653
1.2391	29.5950	38000	0.8034	87.9662	25.6683

Framework versions

• Transformers 4.48.2
• Pytorch 2.5.1+cu124
• Datasets 3.2.0
• Tokenizers 0.21.0