macarious/torgo_xlsr_finetune_M02_old

torgo_xlsr_finetune_M02

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: 1.7002
• Wer: 0.3119

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
3.5308	0.92	1000	3.3287	1.0
1.6778	1.83	2000	1.8864	0.8387
0.8622	2.75	3000	1.4902	0.6310
0.6098	3.66	4000	1.3727	0.5758
0.4854	4.58	5000	1.5900	0.5258
0.4259	5.49	6000	1.4559	0.4403
0.3824	6.41	7000	1.4472	0.4332
0.3162	7.33	8000	1.4480	0.3913
0.3334	8.24	9000	1.5251	0.3663
0.2884	9.16	10000	1.2532	0.3779
0.2745	10.07	11000	1.4908	0.4029
0.2252	10.99	12000	1.7431	0.4055
0.2363	11.9	13000	1.6840	0.3877
0.2135	12.82	14000	1.7977	0.4029
0.2157	13.74	15000	1.6831	0.3743
0.1835	14.65	16000	1.9256	0.3556
0.1718	15.57	17000	1.8000	0.3449
0.1466	16.48	18000	1.8610	0.3414
0.1708	17.4	19000	1.5912	0.3191
0.1516	18.31	20000	1.8241	0.3164
0.1494	19.23	21000	1.7002	0.3119

Framework versions

• Transformers 4.26.1
• Pytorch 2.1.2
• Datasets 2.16.1
• Tokenizers 0.13.3