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: 3000
num_epochs: 30
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
2.7619	0.3502	1000	1.0505	79.7500	27.7494
1.5883	0.7004	2000	0.5542	52.5223	16.3287
1.1356	1.0504	3000	0.5714	52.2068	16.0785
1.0134	1.4006	4000	0.4909	48.9568	14.9144
1.049	1.7508	5000	0.4658	46.3140	14.0887
0.8085	2.1009	6000	0.4274	44.0995	12.9620
0.725	2.4511	7000	0.3970	41.9410	12.1666
0.805	2.8013	8000	0.3763	40.2846	11.4667
0.6876	3.1513	9000	0.3804	40.3847	11.7141
0.6424	3.5015	10000	0.3761	39.6587	11.3674
0.6857	3.8517	11000	0.3511	37.7114	11.1336
0.6235	4.2017	12000	0.3565	37.6741	10.7539
0.6165	4.5519	13000	0.3212	36.0533	10.1479
0.5964	4.9021	14000	0.3031	34.7786	9.6009
0.6145	5.2521	15000	0.3198	35.3324	9.8529
0.6662	5.6023	16000	0.3006	34.1068	9.5917
0.663	5.9525	17000	0.3007	33.6539	9.4145
0.6603	6.3026	18000	0.3044	33.9652	9.3871
0.6754	6.6528	19000	0.2873	33.1790	9.0868
0.6519	7.0028	20000	0.2672	31.9034	8.8680
0.6412	7.3530	21000	0.2759	32.1841	8.8061
0.6028	7.7032	22000	0.2853	31.8364	8.7611
0.5709	8.0532	23000	0.2838	32.5454	9.1492
0.571	8.4034	24000	0.2512	31.2495	8.6213
0.6812	8.7536	25000	0.2698	32.6989	9.2092
0.5007	9.1037	26000	0.2726	31.5692	8.6188
0.4518	9.4539	27000	0.2600	30.0146	8.3456
0.5155	9.8041	28000	0.2494	30.0994	8.2447
0.4055	10.1541	29000	0.2502	29.6261	8.0922
0.3963	10.5043	30000	0.2512	30.0646	8.2523
0.4581	10.8545	31000	0.2405	28.9417	7.8634
0.3554	11.2045	32000	0.2547	29.5303	8.0054
0.3613	11.5547	33000	0.2452	28.4074	7.7358
0.3798	11.9049	34000	0.2373	29.2818	8.0558
0.347	12.2549	35000	0.2354	28.1614	7.6554
0.3401	12.6051	36000	0.2216	27.2794	7.2458
0.3242	12.9553	37000	0.2315	26.9537	7.1938
0.3393	13.3054	38000	0.2313	27.3888	7.3823
0.3717	13.6556	39000	0.2218	26.5771	7.0449
0.3662	14.0056	40000	0.2357	27.6390	7.5118
0.3541	14.3558	41000	0.2265	26.1751	6.8972
0.3686	14.7060	42000	0.2215	26.1937	6.9842
0.3832	15.0560	43000	0.2242	26.0046	6.9059
0.4	15.4062	44000	0.2239	26.0326	6.9814
0.4254	15.7564	45000	0.2192	25.6408	6.8145
0.3963	16.1065	46000	0.2154	25.6933	6.8087
0.3457	16.4567	47000	0.2177	26.4965	7.1219
0.3609	16.8069	48000	0.2196	25.6645	7.0066
0.3055	17.1569	49000	0.2217	26.0267	6.8638
0.296	17.5071	50000	0.2094	25.0878	6.5902
0.2876	17.8573	51000	0.2244	25.5763	6.7381
0.2356	18.2073	52000	0.2205	24.9292	6.5937
0.2801	18.5575	53000	0.2203	24.8028	6.5381
0.2626	18.9077	54000	0.2076	24.4457	6.3823
0.224	19.2577	55000	0.2080	24.4262	6.4048
0.2196	19.6079	56000	0.2169	25.1039	6.6133
0.2341	19.9582	57000	0.2072	24.2269	6.3431
0.1899	20.3082	58000	0.2080	24.1506	6.3084
0.1848	20.6584	59000	0.2076	24.3211	6.3481
0.1816	21.0084	60000	0.2068	23.8003	6.1728
0.207	21.3586	61000	0.2126	23.6527	6.1666
0.1935	21.7088	62000	0.2104	23.3550	6.0682
0.2268	22.0588	63000	0.2122	23.5009	6.0946
0.2236	22.4090	64000	0.2082	23.2601	6.0101
0.2118	22.7592	65000	0.2071	23.4195	6.0859
0.2608	23.1093	66000	0.2023	23.3975	6.0759
0.2122	23.4595	67000	0.2102	23.3958	6.0669
0.2192	23.8097	68000	0.2046	22.9106	5.9151
0.2492	24.1597	69000	0.2106	22.9445	5.9293
0.229	24.5099	70000	0.2077	22.8877	5.9126
0.2188	24.8601	71000	0.2127	22.9759	5.9271
0.2203	25.2101	72000	0.2080	22.7563	5.8728
0.2732	25.5603	73000	0.2083	22.8555	5.9057
0.2355	25.9105	74000	0.2065	22.6748	5.8377
0.1742	26.2605	75000	0.2068	22.6723	5.8282
0.1717	26.6108	76000	0.2081	22.6307	5.8337
0.1944	26.9610	77000	0.2069	22.6341	5.8331
0.1526	27.3110	78000	0.2067	22.6460	5.8402
0.1894	27.6612	79000	0.2067	22.6367	5.8324
0.1358	28.0112	80000	0.2080	22.6638	5.8319
0.1346	28.3614	81000	0.2077	22.6486	5.8285
0.1318	28.7116	82000	0.2082	22.6647	5.8406
0.1584	29.0616	83000	0.2085	22.6876	5.8403
0.1491	29.4118	84000	0.2085	22.6808	5.8377
0.1347	29.7620	85000	0.2085	22.6884	5.8380

Framework versions

Transformers 4.47.1
Pytorch 2.5.1+cu124
Datasets 3.2.0
Tokenizers 0.21.0

davidilag/wav2vec2-xls-r-1b-scandinavian-251h-30-epochs-20250111_v9

wav2vec2-xls-r-1b-scandinavian-251h-30-epochs-20250111_v9

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

davidilag/
wav2vec2-xls-r-1b-scandinavian-251h-30-epochs-20250111_v9