File size: 1,778 Bytes
6d864f7 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 |
---
license: apache-2.0
tags:
- generated_from_trainer
model-index:
- name: my-wav2vec2-base-timit-demo-colab-my
results: []
---
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->
# my-wav2vec2-base-timit-demo-colab-my
This model is a fine-tuned version of [facebook/wav2vec2-base](https://huggingface.co/facebook/wav2vec2-base) on the None dataset.
It achieves the following results on the evaluation set:
- Loss: 0.5569
- Wer: 0.3481
## 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: 32
- 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 |
|:-------------:|:-----:|:----:|:---------------:|:------:|
| 3.4083 | 4.0 | 500 | 1.0932 | 0.7510 |
| 0.5536 | 8.0 | 1000 | 0.4965 | 0.4819 |
| 0.2242 | 12.0 | 1500 | 0.4779 | 0.4077 |
| 0.1249 | 16.0 | 2000 | 0.4921 | 0.4006 |
| 0.0844 | 20.0 | 2500 | 0.4809 | 0.3753 |
| 0.0613 | 24.0 | 3000 | 0.5307 | 0.3680 |
| 0.0459 | 28.0 | 3500 | 0.5569 | 0.3481 |
### Framework versions
- Transformers 4.11.3
- Pytorch 1.10.0+cu111
- Datasets 1.18.3
- Tokenizers 0.10.3
|