torgo_xlsr_finetune_M04

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

• Loss: 1.6215
• Wer: 0.2742

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
3.4822	0.84	1000	3.2714	1.0
1.8117	1.69	2000	1.6545	0.7980
0.8407	2.53	3000	1.4796	0.5925
0.6096	3.38	4000	1.3550	0.5331
0.4956	4.22	5000	1.4899	0.4576
0.4388	5.07	6000	1.1873	0.3778
0.374	5.91	7000	1.2894	0.3591
0.3569	6.76	8000	1.2815	0.3336
0.2865	7.6	9000	1.3162	0.3073
0.2809	8.45	10000	1.4108	0.3234
0.2455	9.29	11000	1.3710	0.2920
0.2266	10.14	12000	1.4937	0.2946
0.2368	10.98	13000	1.5300	0.3056
0.2229	11.82	14000	1.3822	0.2683
0.1929	12.67	15000	1.4279	0.2683
0.1823	13.51	16000	1.5015	0.2869
0.1678	14.36	17000	1.7070	0.3048
0.1751	15.2	18000	1.4764	0.2844
0.1682	16.05	19000	1.2581	0.2742
0.1607	16.89	20000	1.5641	0.2852
0.1431	17.74	21000	1.6488	0.2742
0.1442	18.58	22000	1.5386	0.2725
0.1147	19.43	23000	1.6215	0.2742

Framework versions

• Transformers 4.26.1
• Pytorch 2.2.1
• Datasets 2.18.0
• Tokenizers 0.13.3