File size: 3,185 Bytes
67b7c36 7bcef33 67b7c36 |
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 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 |
---
language: en
tags:
- automatic-speech-recognition
- timit_asr
- generated_from_trainer
datasets:
- timit_asr
model-index:
- name: unispeech-sat-base-plus-timit-ft
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. -->
# unispeech-sat-base-plus-timit-ft
This model is a fine-tuned version of [microsoft/unispeech-sat-base-plus](https://huggingface.co/microsoft/unispeech-sat-base-plus) on the TIMIT_ASR - NA dataset.
It achieves the following results on the evaluation set:
- Loss: 0.6549
- Wer: 0.4051
## 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: 1
- 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: 20.0
- mixed_precision_training: Native AMP
### Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|:-------------:|:-----:|:----:|:---------------:|:------:|
| 3.3838 | 0.69 | 100 | 3.2528 | 1.0 |
| 2.9608 | 1.38 | 200 | 2.9682 | 1.0 |
| 2.9574 | 2.07 | 300 | 2.9346 | 1.0 |
| 2.8555 | 2.76 | 400 | 2.7612 | 1.0 |
| 1.7418 | 3.45 | 500 | 1.5732 | 0.9857 |
| 0.9606 | 4.14 | 600 | 1.0014 | 0.7052 |
| 0.8334 | 4.83 | 700 | 0.7691 | 0.6161 |
| 0.852 | 5.52 | 800 | 0.7169 | 0.5997 |
| 0.5707 | 6.21 | 900 | 0.6821 | 0.5527 |
| 0.4235 | 6.9 | 1000 | 0.6078 | 0.5140 |
| 0.4357 | 7.59 | 1100 | 0.5927 | 0.4982 |
| 0.5004 | 8.28 | 1200 | 0.5814 | 0.4826 |
| 0.3757 | 8.97 | 1300 | 0.5951 | 0.4643 |
| 0.2579 | 9.66 | 1400 | 0.5990 | 0.4581 |
| 0.2087 | 10.34 | 1500 | 0.5864 | 0.4488 |
| 0.3155 | 11.03 | 1600 | 0.5836 | 0.4464 |
| 0.2701 | 11.72 | 1700 | 0.6045 | 0.4348 |
| 0.172 | 12.41 | 1800 | 0.6494 | 0.4344 |
| 0.1529 | 13.1 | 1900 | 0.5915 | 0.4241 |
| 0.2411 | 13.79 | 2000 | 0.6156 | 0.4246 |
| 0.2348 | 14.48 | 2100 | 0.6363 | 0.4206 |
| 0.1429 | 15.17 | 2200 | 0.6394 | 0.4161 |
| 0.1151 | 15.86 | 2300 | 0.6186 | 0.4167 |
| 0.1723 | 16.55 | 2400 | 0.6498 | 0.4124 |
| 0.1997 | 17.24 | 2500 | 0.6541 | 0.4076 |
| 0.1297 | 17.93 | 2600 | 0.6546 | 0.4117 |
| 0.101 | 18.62 | 2700 | 0.6471 | 0.4075 |
| 0.1272 | 19.31 | 2800 | 0.6586 | 0.4065 |
| 0.1901 | 20.0 | 2900 | 0.6549 | 0.4051 |
### Framework versions
- Transformers 4.12.0.dev0
- Pytorch 1.8.1
- Datasets 1.14.1.dev0
- Tokenizers 0.10.3
|