macarious/torgo_xlsr_finetune_M04_keep_all

torgo_xlsr_finetune_M04__keep_all

This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on the torgo dataset. It achieves the following results on the evaluation set:

• Loss: 1.4844
• Wer: 0.2303

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
3.5764	0.55	1000	3.3931	1.0
2.2058	1.1	2000	1.8051	0.8589
1.0531	1.66	3000	1.5038	0.6508
0.7783	2.21	4000	1.2594	0.5208
0.6084	2.76	5000	1.3131	0.4586
0.5241	3.31	6000	1.3666	0.4339
0.4586	3.87	7000	1.3358	0.3866
0.4149	4.42	8000	1.2625	0.3332
0.3796	4.97	9000	1.5808	0.3629
0.3685	5.52	10000	1.2197	0.3298
0.3322	6.07	11000	1.6204	0.3473
0.3133	6.63	12000	1.6558	0.3446
0.2833	7.18	13000	1.5270	0.3100
0.2941	7.73	14000	1.4321	0.3134
0.2709	8.28	15000	1.3682	0.3092
0.2362	8.83	16000	1.2184	0.2787
0.2205	9.39	17000	1.4273	0.2863
0.2515	9.94	18000	1.3085	0.2665
0.2185	10.49	19000	1.5292	0.2852
0.2197	11.04	20000	1.4625	0.2817
0.2122	11.6	21000	1.4086	0.2634
0.1869	12.15	22000	1.6290	0.2791
0.1839	12.7	23000	1.4520	0.2722
0.1946	13.25	24000	1.5211	0.2653
0.1871	13.8	25000	1.3136	0.2390
0.1831	14.36	26000	1.4022	0.2581
0.1644	14.91	27000	1.5609	0.2673
0.1499	15.46	28000	1.3431	0.2429
0.1566	16.01	29000	1.5110	0.2566
0.1533	16.57	30000	1.4567	0.2345
0.1446	17.12	31000	1.5160	0.2478
0.1451	17.67	32000	1.4081	0.2379
0.1269	18.22	33000	1.5296	0.2379
0.1438	18.77	34000	1.5765	0.2406
0.1112	19.33	35000	1.5061	0.2337
0.1215	19.88	36000	1.4844	0.2303

Framework versions

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