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 Loss	Epoch	Step	Validation Loss	Wer	Cer
3.3611	0.9906	1000	3.2332	100.0	100.0
2.9003	1.9807	2000	2.8895	100.0	100.0
0.8292	2.9708	3000	0.5835	50.6505	14.4139
0.6	3.9609	4000	0.3336	33.3089	9.2976
0.5112	4.9510	5000	0.2612	27.2455	7.3862
0.4183	5.9411	6000	0.2219	23.5382	6.3992
0.3555	6.9312	7000	0.2016	21.7709	5.9146
0.2589	7.9212	8000	0.1936	20.7790	5.6182
0.295	8.9113	9000	0.1828	20.5748	5.5029
0.3082	9.9014	10000	0.1747	19.4431	5.2727
0.2971	10.8915	11000	0.1862	19.2443	5.2850
0.2523	11.8816	12000	0.1730	18.5114	5.0436
0.2625	12.8717	13000	0.1719	18.2143	4.9644
0.3564	13.8618	14000	0.2576	19.4126	5.1626
0.3835	14.8519	15000	0.2446	19.5185	5.2376
0.9611	15.8420	16000	0.8293	31.5864	8.4708
1.2053	16.8321	17000	1.0922	85.9190	39.8617
1.6349	17.8222	18000	1.4312	95.7050	62.3241
1.7414	18.8123	19000	1.5612	99.9257	90.1000
1.5747	19.8024	20000	1.4791	98.9634	73.0416
1.4862	20.7925	21000	1.3880	97.5215	64.0388
1.494	21.7826	22000	1.3281	95.2266	55.3230
1.4748	22.7727	23000	1.3006	91.1545	45.9123
1.4855	23.7628	24000	1.2962	84.4739	37.1604
1.4329	24.7528	25000	1.3060	78.2040	30.8656
1.4158	25.7429	26000	1.3269	74.1013	26.9627
1.3977	26.7330	27000	1.3381	73.0199	25.7280
1.5212	27.7231	28000	1.3618	71.0243	23.8946
1.544	28.7132	29000	1.3651	70.7676	23.6830
1.5178	29.7033	30000	1.3654	70.7916	23.6644

davidilag/wav2vec2-xls-r-300m-scandinavian-E4-100h-30-epochs-20250201_v2.3

wav2vec2-xls-r-300m-scandinavian-E4-100h-30-epochs-20250201_v2.3