xlsr-nomimo-aiish

This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.0000
• Wer: 0.3068

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: 0.0004
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 132
• num_epochs: 100
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
4.6528	1.3652	200	2.5205	1.0
1.5409	2.7304	400	0.2012	0.5428
0.3781	4.0956	600	0.0349	0.3839
0.1391	5.4608	800	0.0140	0.3264
0.1157	6.8259	1000	0.0059	0.3093
0.0752	8.1911	1200	0.0081	0.3117
0.0653	9.5563	1400	0.0072	0.3068
0.0527	10.9215	1600	0.0033	0.3117
0.0542	12.2867	1800	0.0019	0.3093
0.0489	13.6519	2000	0.0148	0.3337
0.0374	15.0171	2200	0.0024	0.3093
0.0412	16.3823	2400	0.0038	0.3154
0.0302	17.7474	2600	0.0005	0.3068
0.0313	19.1126	2800	0.0014	0.3105
0.0274	20.4778	3000	0.0003	0.3068
0.0403	21.8430	3200	0.0010	0.3068
0.0274	23.2082	3400	0.0008	0.3068
0.0264	24.5734	3600	0.0005	0.3068
0.0302	25.9386	3800	0.0022	0.3068
0.0197	27.3038	4000	0.0006	0.3068
0.0187	28.6689	4200	0.0008	0.3081
0.0242	30.0341	4400	0.0003	0.3068
0.0155	31.3993	4600	0.0086	0.3068
0.0217	32.7645	4800	0.0006	0.3068
0.0189	34.1297	5000	0.0004	0.3068
0.0197	35.4949	5200	0.0001	0.3068
0.0211	36.8601	5400	0.0002	0.3068
0.0122	38.2253	5600	0.0003	0.3068
0.017	39.5904	5800	0.0002	0.3068
0.0195	40.9556	6000	0.0005	0.3081
0.0154	42.3208	6200	0.0028	0.3093
0.0081	43.6860	6400	0.0046	0.3117
0.0155	45.0512	6600	0.0005	0.3068
0.014	46.4164	6800	0.0004	0.3081
0.0113	47.7816	7000	0.0002	0.3068
0.0122	49.1468	7200	0.0001	0.3068
0.0105	50.5119	7400	0.0000	0.3068
0.0075	51.8771	7600	0.0000	0.3068
0.01	53.2423	7800	0.0001	0.3068
0.0077	54.6075	8000	0.0002	0.3068
0.0076	55.9727	8200	0.0001	0.3068
0.0097	57.3379	8400	0.0010	0.3081
0.0083	58.7031	8600	0.0000	0.3068
0.0097	60.0683	8800	0.0000	0.3068
0.0091	61.4334	9000	0.0001	0.3068
0.0059	62.7986	9200	0.0003	0.3081
0.0055	64.1638	9400	0.0006	0.3081
0.0073	65.5290	9600	0.0001	0.3081
0.0068	66.8942	9800	0.0005	0.3081
0.0057	68.2594	10000	0.0001	0.3068
0.0054	69.6246	10200	0.0006	0.3081
0.0049	70.9898	10400	0.0000	0.3068
0.0044	72.3549	10600	0.0004	0.3081
0.0065	73.7201	10800	0.0001	0.3068
0.0026	75.0853	11000	0.0000	0.3068
0.003	76.4505	11200	0.0000	0.3068
0.0038	77.8157	11400	0.0000	0.3068
0.0023	79.1809	11600	0.0000	0.3068
0.0034	80.5461	11800	0.0000	0.3068
0.0022	81.9113	12000	0.0000	0.3068
0.0024	83.2765	12200	0.0000	0.3068
0.0027	84.6416	12400	0.0000	0.3068
0.0015	86.0068	12600	0.0000	0.3068
0.0021	87.3720	12800	0.0000	0.3068
0.0037	88.7372	13000	0.0000	0.3068
0.0019	90.1024	13200	0.0000	0.3068
0.0016	91.4676	13400	0.0000	0.3068
0.0025	92.8328	13600	0.0000	0.3081
0.0029	94.1980	13800	0.0000	0.3068
0.0023	95.5631	14000	0.0000	0.3068
0.0017	96.9283	14200	0.0000	0.3068
0.0018	98.2935	14400	0.0000	0.3068
0.0016	99.6587	14600	0.0000	0.3068

Framework versions

• Transformers 4.45.0.dev0
• Pytorch 2.4.0
• Datasets 2.21.0
• Tokenizers 0.19.1