torgo_xlsr_finetune_F01

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: 1.2809
• Wer: 0.2462

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
3.7203	0.83	1000	3.5860	1.0
3.3181	1.66	2000	3.2148	1.0
1.547	2.49	3000	1.6683	0.8497
0.8041	3.32	4000	1.1466	0.6036
0.5827	4.15	5000	1.1892	0.5093
0.5022	4.98	6000	1.1251	0.4295
0.4373	5.81	7000	1.4837	0.4270
0.3729	6.64	8000	1.1526	0.3430
0.3448	7.47	9000	1.0968	0.3489
0.2964	8.3	10000	1.2292	0.3014
0.2806	9.13	11000	1.4337	0.3345
0.2684	9.96	12000	1.4606	0.3472
0.2587	10.79	13000	1.1020	0.2937
0.2187	11.62	14000	1.3086	0.3192
0.2052	12.45	15000	1.3270	0.2912
0.1757	13.28	16000	1.2323	0.2784
0.2061	14.11	17000	1.1289	0.2716
0.1749	14.94	18000	1.2851	0.2784
0.1476	15.77	19000	1.1545	0.2547
0.1446	16.6	20000	1.2718	0.2487
0.1319	17.43	21000	1.2026	0.2487
0.1533	18.26	22000	1.2299	0.2530
0.1349	19.09	23000	1.3010	0.2513
0.1185	19.92	24000	1.2809	0.2462

Framework versions

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