patrickvonplaten/hello_2b

1---2language:3- tr4tags:5- automatic-speech-recognition6- common_voice7- generated_from_trainer8datasets:9- common_voice10model-index:11- name: hello_2b12  results: []13---14 15<!-- This model card has been generated automatically according to the information the Trainer had access to. You16should probably proofread and complete it, then remove this comment. -->17 18# hello_2b19 20This model is a fine-tuned version of [facebook/wav2vec2-xls-r-2b](https://huggingface.co/facebook/wav2vec2-xls-r-2b) on the COMMON_VOICE - TR dataset.21It achieves the following results on the evaluation set:22- Loss: 1.272523- Wer: 0.953124 25## Model description26 27More information needed28 29## Intended uses & limitations30 31More information needed32 33## Training and evaluation data34 35More information needed36 37## Training procedure38 39### Training hyperparameters40 41The following hyperparameters were used during training:42- learning_rate: 1e-0543- train_batch_size: 244- eval_batch_size: 845- seed: 4246- distributed_type: multi-GPU47- num_devices: 248- gradient_accumulation_steps: 849- total_train_batch_size: 3250- total_eval_batch_size: 1651- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-0852- lr_scheduler_type: linear53- lr_scheduler_warmup_steps: 50054- num_epochs: 30.055- mixed_precision_training: Native AMP56 57### Training results58 59| Training Loss | Epoch | Step | Validation Loss | Wer    |60|:-------------:|:-----:|:----:|:---------------:|:------:|61| 3.1646        | 0.92  | 100  | 3.2106          | 1.0    |62| 0.368         | 1.85  | 200  | 2.9963          | 1.0    |63| 0.2252        | 2.77  | 300  | 2.8078          | 0.9999 |64| 0.1546        | 3.7   | 400  | 2.3458          | 0.9996 |65| 0.1468        | 4.63  | 500  | 2.0086          | 0.9986 |66| 0.1261        | 5.55  | 600  | 1.8269          | 0.9985 |67| 0.1206        | 6.48  | 700  | 1.7347          | 0.9956 |68| 0.1959        | 7.4   | 800  | 1.6819          | 0.9955 |69| 0.0502        | 8.33  | 900  | 1.6809          | 0.9965 |70| 0.0811        | 9.26  | 1000 | 1.6674          | 0.9916 |71| 0.0534        | 10.18 | 1100 | 1.5719          | 0.9898 |72| 0.0402        | 11.11 | 1200 | 1.4620          | 0.9821 |73| 0.057         | 12.04 | 1300 | 1.3015          | 0.9554 |74| 0.0385        | 12.96 | 1400 | 1.3798          | 0.9600 |75| 0.0422        | 13.88 | 1500 | 1.3538          | 0.9699 |76| 0.014         | 14.81 | 1600 | 1.2507          | 0.9443 |77| 0.0232        | 15.74 | 1700 | 1.3318          | 0.9465 |78| 0.0554        | 16.66 | 1800 | 1.2784          | 0.9462 |79| 0.0316        | 17.59 | 1900 | 1.2503          | 0.9481 |80| 0.0524        | 18.51 | 2000 | 1.3920          | 0.9604 |81| 0.0142        | 19.44 | 2100 | 1.4224          | 0.9698 |82| 0.0288        | 20.37 | 2200 | 1.3475          | 0.9635 |83| 0.0106        | 21.29 | 2300 | 1.2232          | 0.9264 |84| 0.0396        | 22.22 | 2400 | 1.3323          | 0.9615 |85| 0.0349        | 23.15 | 2500 | 1.2741          | 0.9587 |86| 0.0121        | 24.07 | 2600 | 1.2671          | 0.9586 |87| 0.0224        | 24.99 | 2700 | 1.3001          | 0.9611 |88| 0.0449        | 25.92 | 2800 | 1.2777          | 0.9572 |89| 0.0186        | 26.85 | 2900 | 1.2766          | 0.9607 |90| 0.0365        | 27.77 | 3000 | 1.2935          | 0.9598 |91| 0.0105        | 28.7  | 3100 | 1.2761          | 0.9588 |92| 0.021         | 29.63 | 3200 | 1.2686          | 0.9528 |93 94 95### Framework versions96 97- Transformers 4.13.0.dev098- Pytorch 1.10.099- Datasets 1.15.2.dev0100- Tokenizers 0.10.3101