wav2vec2-large-xlsr-grosman-53-texts-exp-1

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

• Loss: 6.9233
• Wer: 0.9898
• Cer: 0.8761
• Per: 0.9896

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: 3e-05
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 200
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer	Per
103.0605	1.0	14	22.9419	1.0	4.5340	1.0
103.0605	2.0	28	11.7027	1.0	0.9520	1.0
103.0605	3.0	42	10.4194	1.0	0.9536	1.0
103.0605	4.0	56	10.1827	1.0094	0.9144	1.0094
103.0605	5.0	70	9.3103	1.0	0.9537	1.0
103.0605	6.0	84	10.7017	0.9990	0.8930	0.9990
103.0605	7.0	98	9.0149	0.9914	0.8884	0.9914
20.3215	8.0	112	8.8777	1.0079	0.8647	1.0079
20.3215	9.0	126	8.3998	0.9980	0.8853	0.9980
20.3215	10.0	140	8.5023	0.9873	0.8483	0.9873
20.3215	11.0	154	10.3925	0.9919	0.8857	0.9919
20.3215	12.0	168	8.0281	0.9868	0.8428	0.9868
20.3215	13.0	182	7.4940	0.9837	0.8383	0.9837
20.3215	14.0	196	8.3053	0.9985	0.8501	0.9985
8.131	15.0	210	9.6699	0.9873	0.8626	0.9873
8.131	16.0	224	9.2462	1.0018	0.8801	1.0018
8.131	17.0	238	7.6267	1.0152	0.8154	1.0152
8.131	18.0	252	7.4720	0.9904	0.8806	0.9904
8.131	19.0	266	7.2126	1.0015	0.8390	1.0015
8.131	20.0	280	8.0197	0.9947	0.8710	0.9947
8.131	21.0	294	7.7732	0.9924	0.8575	0.9924
5.9777	22.0	308	7.2334	0.9947	0.8547	0.9947
5.9777	23.0	322	8.4644	0.9924	0.8911	0.9924
5.9777	24.0	336	6.9233	0.9898	0.8761	0.9896
5.9777	25.0	350	9.3194	1.0	0.9350	1.0
5.9777	26.0	364	8.0294	0.9987	0.9365	0.9987
5.9777	27.0	378	9.5686	0.9992	0.9281	0.9992
5.9777	28.0	392	9.3024	1.0	0.9418	1.0
4.8478	29.0	406	9.2086	0.9972	0.9184	0.9972
4.8478	30.0	420	9.6811	0.9995	0.9388	0.9995
4.8478	31.0	434	10.2141	0.9992	0.9401	0.9992
4.8478	32.0	448	10.5870	0.9990	0.9359	0.9990
4.8478	33.0	462	9.7545	0.9977	0.9197	0.9977
4.8478	34.0	476	9.9550	0.9987	0.9412	0.9987
4.8478	35.0	490	10.3337	0.9975	0.9280	0.9975
4.0401	36.0	504	10.0079	0.9997	0.9336	0.9997
4.0401	37.0	518	11.0687	0.9982	0.9373	0.9982
4.0401	38.0	532	9.6581	1.0	0.9502	1.0
4.0401	39.0	546	9.3628	0.9987	0.9310	0.9987
4.0401	40.0	560	9.5577	0.9985	0.9305	0.9985
4.0401	41.0	574	10.5525	0.9997	0.9465	0.9997
4.0401	42.0	588	9.8161	0.9995	0.9411	0.9995
3.8322	43.0	602	10.0675	0.9997	0.9461	0.9997
3.8322	44.0	616	9.8196	0.9997	0.9358	0.9997

Framework versions

• Transformers 4.28.0
• Pytorch 2.4.1+cu121
• Datasets 3.2.0
• Tokenizers 0.13.3