xlsr-wav2vec2-3

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: 0.4201
• Wer: 0.3998

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: 0.0003
• 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: 800
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
5.0117	0.68	400	3.0284	0.9999
2.6502	1.35	800	1.0868	0.9374
0.9362	2.03	1200	0.5216	0.6491
0.6675	2.7	1600	0.4744	0.5837
0.5799	3.38	2000	0.4400	0.5802
0.5196	4.05	2400	0.4266	0.5314
0.4591	4.73	2800	0.3808	0.5190
0.4277	5.41	3200	0.3987	0.5036
0.4125	6.08	3600	0.3902	0.5040
0.3797	6.76	4000	0.4105	0.5025
0.3606	7.43	4400	0.3975	0.4823
0.3554	8.11	4800	0.3733	0.4747
0.3373	8.78	5200	0.3737	0.4726
0.3252	9.46	5600	0.3795	0.4736
0.3192	10.14	6000	0.3935	0.4736
0.3012	10.81	6400	0.3974	0.4648
0.2972	11.49	6800	0.4497	0.4724
0.2873	12.16	7200	0.4645	0.4843
0.2849	12.84	7600	0.4461	0.4709
0.274	13.51	8000	0.4002	0.4695
0.2709	14.19	8400	0.4188	0.4627
0.2619	14.86	8800	0.3987	0.4646
0.2545	15.54	9200	0.4083	0.4668
0.2477	16.22	9600	0.4525	0.4728
0.2455	16.89	10000	0.4148	0.4515
0.2281	17.57	10400	0.4304	0.4514
0.2267	18.24	10800	0.4077	0.4446
0.2136	18.92	11200	0.4209	0.4445
0.2032	19.59	11600	0.4543	0.4534
0.1999	20.27	12000	0.4184	0.4373
0.1898	20.95	12400	0.4044	0.4424
0.1846	21.62	12800	0.4098	0.4288
0.1796	22.3	13200	0.4047	0.4262
0.1715	22.97	13600	0.4077	0.4189
0.1641	23.65	14000	0.4162	0.4248
0.1615	24.32	14400	0.4392	0.4222
0.1575	25.0	14800	0.4296	0.4185
0.1456	25.68	15200	0.4363	0.4129
0.1461	26.35	15600	0.4305	0.4124
0.1422	27.03	16000	0.4237	0.4086
0.1378	27.7	16400	0.4294	0.4051
0.1326	28.38	16800	0.4311	0.4051
0.1286	29.05	17200	0.4153	0.3992
0.1283	29.73	17600	0.4201	0.3998

Framework versions

• Transformers 4.19.2
• Pytorch 1.11.0+cu113
• Datasets 2.2.2
• Tokenizers 0.12.1