torgo_xlsr_finetune_M03_keep_all

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: 1.6155
• Wer: 0.2360

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: 0.0001
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 8
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 1000
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
3.5946	0.56	1000	3.3418	1.0
2.3765	1.12	2000	1.8751	0.9367
1.0589	1.68	3000	1.4354	0.6588
0.7686	2.24	4000	1.3288	0.5193
0.7029	2.8	5000	1.2625	0.5071
0.5645	3.37	6000	1.3686	0.4331
0.5149	3.93	7000	1.2946	0.4392
0.4504	4.49	8000	1.4451	0.3793
0.4012	5.05	9000	1.3974	0.3324
0.3683	5.61	10000	1.6211	0.3553
0.3661	6.17	11000	1.4331	0.3488
0.3337	6.73	12000	1.6473	0.3454
0.3087	7.29	13000	1.4651	0.3096
0.2908	7.85	14000	1.3439	0.2844
0.2692	8.41	15000	1.2399	0.2871
0.262	8.97	16000	1.4219	0.3111
0.244	9.53	17000	1.5202	0.3065
0.2672	10.1	18000	1.3916	0.2840
0.2346	10.66	19000	1.6752	0.3077
0.2089	11.22	20000	1.4122	0.2734
0.2262	11.78	21000	1.4316	0.2795
0.2043	12.34	22000	1.6063	0.2943
0.1836	12.9	23000	1.5199	0.2726
0.1701	13.46	24000	1.6889	0.2722
0.1938	14.02	25000	1.5244	0.2619
0.1734	14.58	26000	1.8305	0.2692
0.1714	15.14	27000	1.6078	0.2539
0.1521	15.7	28000	1.8210	0.2665
0.1346	16.26	29000	1.7116	0.2653
0.1498	16.83	30000	1.4663	0.2432
0.1594	17.39	31000	1.5994	0.2402
0.1647	17.95	32000	1.5112	0.2356
0.1238	18.51	33000	1.6993	0.2429
0.1554	19.07	34000	1.5374	0.2379
0.1238	19.63	35000	1.6155	0.2360

Framework versions

• Transformers 4.26.1
• Pytorch 2.2.0
• Datasets 2.16.1
• Tokenizers 0.13.3