wav2vec2-base-timit-demo-google-colab

This model is a fine-tuned version of facebook/wav2vec2-base on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.5328
• Wer: 0.2175

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: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 1000
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
3.4608	1.0	500	1.4610	0.9290
0.796	2.01	1000	0.5283	0.4217
0.4219	3.01	1500	0.4277	0.3353
0.2919	4.02	2000	0.4154	0.3102
0.2263	5.02	2500	0.4096	0.2954
0.1885	6.02	3000	0.4274	0.2944
0.1595	7.03	3500	0.4529	0.2681
0.1371	8.03	4000	0.4309	0.2721
0.1218	9.04	4500	0.4574	0.2629
0.1118	10.04	5000	0.5396	0.2605
0.0988	11.04	5500	0.5031	0.2683
0.0944	12.05	6000	0.5040	0.2595
0.0781	13.05	6500	0.4909	0.2611
0.0714	14.06	7000	0.4740	0.2635
0.0648	15.06	7500	0.4613	0.2509
0.0628	16.06	8000	0.4731	0.2508
0.0539	17.07	8500	0.5100	0.2448
0.0542	18.07	9000	0.5048	0.2507
0.0453	19.08	9500	0.5290	0.2466
0.0446	20.08	10000	0.5482	0.2398
0.0405	21.08	10500	0.5768	0.2422
0.0368	22.09	11000	0.5848	0.2403
0.0349	23.09	11500	0.5469	0.2321
0.0326	24.1	12000	0.5618	0.2294
0.028	25.1	12500	0.5590	0.2297
0.0254	26.1	13000	0.5531	0.2291
0.0258	27.11	13500	0.5302	0.2215
0.0244	28.11	14000	0.5388	0.2188
0.0195	29.12	14500	0.5328	0.2175

Framework versions

• Transformers 4.35.2
• Pytorch 2.1.0+cu118
• Datasets 1.18.3
• Tokenizers 0.15.0