metadata

license: apache-2.0
base_model: facebook/wav2vec2-xls-r-300m
tags:
  - generated_from_trainer
metrics:
  - wer
model-index:
  - name: 1-char-based-freeze_cnn-dropout0.1
    results: []

1-char-based-freeze_cnn-dropout0.1

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: 0.2454
Wer: 0.1804

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: 2e-05
train_batch_size: 12
eval_batch_size: 4
seed: 42
distributed_type: multi-GPU
num_devices: 4
total_train_batch_size: 48
total_eval_batch_size: 16
optimizer: Adam with betas=(0.9,0.98) and epsilon=1e-08
lr_scheduler_type: linear
lr_scheduler_warmup_ratio: 0.1
training_steps: 200000

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
5.3032	0.09	2500	7.0530	1.0
3.4521	0.18	5000	3.6019	1.0
3.3037	0.26	7500	3.4931	1.0
3.2012	0.35	10000	3.4193	1.0
2.3026	0.44	12500	1.9423	0.9873
1.4238	0.53	15000	0.8772	0.6695
1.1592	0.62	17500	0.6630	0.5011
0.861	0.7	20000	0.5460	0.4239
0.8123	0.79	22500	0.4794	0.3830
0.7568	0.88	25000	0.4369	0.3463
0.7182	0.97	27500	0.4111	0.3289
0.6896	1.06	30000	0.4041	0.3102
0.6655	1.14	32500	0.3933	0.2986
0.5738	1.23	35000	0.3676	0.2829
0.6361	1.32	37500	0.3533	0.2727
0.6142	1.41	40000	0.3545	0.2716
0.6346	1.5	42500	0.3428	0.2615
0.5739	1.58	45000	0.3470	0.2578
0.544	1.67	47500	0.3207	0.2490
0.5283	1.76	50000	0.3202	0.2424
0.5552	1.85	52500	0.3187	0.2379
0.5218	1.94	55000	0.3242	0.2383
0.4939	2.02	57500	0.3277	0.2418
0.5141	2.11	60000	0.3058	0.2329
0.5189	2.2	62500	0.3086	0.2273
0.4993	2.29	65000	0.3005	0.2245
0.5156	2.38	67500	0.2998	0.2223
0.4787	2.46	70000	0.2940	0.2173
0.5296	2.55	72500	0.3003	0.2225
0.4759	2.64	75000	0.2995	0.2144
0.485	2.73	77500	0.2882	0.2126
0.4888	2.82	80000	0.2893	0.2189
0.51	2.9	82500	0.2767	0.2046
0.4703	2.99	85000	0.2899	0.2124
0.4406	3.08	87500	0.2787	0.2068
0.4328	3.17	90000	0.2823	0.2070
0.4399	3.26	92500	0.2802	0.2058
0.4788	3.34	95000	0.2741	0.2084
0.4621	3.43	97500	0.2817	0.2038
0.523	3.52	100000	0.2735	0.2015
0.4689	3.61	102500	0.2631	0.1995
0.4502	3.7	105000	0.2689	0.1986
0.4402	3.78	107500	0.2726	0.1987
0.4189	3.87	110000	0.2724	0.1994
0.4526	3.96	112500	0.2596	0.1918
0.4755	4.05	115000	0.2583	0.1900
0.4374	4.14	117500	0.2590	0.1944
0.4155	4.23	120000	0.2695	0.1961
0.4463	4.31	122500	0.2605	0.1909
0.4007	4.4	125000	0.2529	0.1891
0.4156	4.49	127500	0.2568	0.1913
0.4124	4.58	130000	0.2559	0.1889
0.4085	4.67	132500	0.2610	0.1922
0.4474	4.75	135000	0.2588	0.1961
0.4098	4.84	137500	0.2512	0.1877
0.3941	4.93	140000	0.2549	0.1891
0.3917	5.02	142500	0.2544	0.1863
0.4324	5.11	145000	0.2564	0.1882
0.4255	5.19	147500	0.2536	0.1885
0.3894	5.28	150000	0.2538	0.1860
0.4108	5.37	152500	0.2539	0.1860
0.4312	5.46	155000	0.2526	0.1849
0.3786	5.55	157500	0.2504	0.1837
0.4033	5.63	160000	0.2516	0.1852
0.3973	5.72	162500	0.2570	0.1870
0.3994	5.81	165000	0.2499	0.1846
0.4183	5.9	167500	0.2489	0.1835
0.3826	5.99	170000	0.2468	0.1847
0.4103	6.07	172500	0.2477	0.1806
0.4291	6.16	175000	0.2492	0.1835
0.4417	6.25	177500	0.2464	0.1824
0.3962	6.34	180000	0.2476	0.1815
0.4633	6.43	182500	0.2447	0.1818
0.422	6.51	185000	0.2455	0.1802
0.4098	6.6	187500	0.2488	0.1814
0.4018	6.69	190000	0.2453	0.1804
0.4559	6.78	192500	0.2458	0.1823
0.4033	6.87	195000	0.2451	0.1794
0.3829	6.95	197500	0.2453	0.1804
0.3676	7.04	200000	0.2454	0.1804

Framework versions

Transformers 4.34.0
Pytorch 2.0.1
Datasets 2.14.5
Tokenizers 0.14.1