torgo_xlsr_finetune_M02_keep_all

This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.6539
• Wer: 0.2436

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
3.5043	0.56	1000	3.3139	1.0
2.1248	1.12	2000	1.9926	0.8898
1.0178	1.67	3000	1.5324	0.6683
0.7315	2.23	4000	1.7989	0.5959
0.6289	2.79	5000	1.3984	0.4987
0.5123	3.35	6000	1.2977	0.4228
0.4751	3.91	7000	1.3967	0.3988
0.4354	4.47	8000	1.5080	0.4274
0.3817	5.03	9000	1.7897	0.4014
0.3758	5.58	10000	1.3421	0.3385
0.358	6.14	11000	1.6429	0.3427
0.3083	6.7	12000	1.2683	0.3084
0.2805	7.26	13000	1.7095	0.3122
0.2856	7.82	14000	1.7918	0.3317
0.2574	8.38	15000	1.5411	0.2947
0.2495	8.93	16000	1.4551	0.2997
0.2651	9.49	17000	1.5073	0.2825
0.2517	10.05	18000	1.6405	0.2920
0.2274	10.61	19000	1.4440	0.2604
0.2278	11.17	20000	1.4020	0.2875
0.2472	11.73	21000	1.6264	0.2897
0.1875	12.28	22000	1.5901	0.2783
0.175	12.84	23000	1.4056	0.2501
0.1751	13.4	24000	1.4809	0.2631
0.1607	13.96	25000	1.4363	0.2551
0.1712	14.52	26000	1.6480	0.2524
0.1581	15.08	27000	1.5084	0.2615
0.1623	15.63	28000	1.4066	0.2482
0.1397	16.19	29000	1.7111	0.2619
0.1536	16.75	30000	1.4691	0.2402
0.1343	17.31	31000	1.5406	0.2329
0.1428	17.87	32000	1.5261	0.2413
0.1125	18.43	33000	1.6416	0.2337
0.1214	18.98	34000	1.6803	0.2425
0.124	19.54	35000	1.6539	0.2436

Framework versions

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