File size: 1,465 Bytes
b95136a 001f7fa b95136a adaa99e cddd879 adaa99e b95136a adaa99e b95136a |
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 |
---
language:
- en
license: mit
base_model: microsoft/speecht5_tts
tags:
- scottish
- tts
- glaswegian
- generated_from_trainer
datasets:
- divakaivan/glaswegian_audio
model-index:
- name: GlaswegianTTS v0.1.0
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. -->
# GlaswegianTTS v0.1.0
This model is a fine-tuned version of [microsoft/speecht5_tts](https://huggingface.co/microsoft/speecht5_tts) on the glaswegian_tts_v0.1.0 dataset.
## Training procedure
### Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 1e-05
- train_batch_size: 16
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 32
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 1000
- training_steps: 4000
- mixed_precision_training: Native AMP
### Training results
| Training Loss | Epoch | Step | Validation Loss |
|:-------------:|:-----:|:----:|:---------------:|
| 0.4348 | 125.0 | 1000 | 0.4666 |
| 0.3783 | 250.0 | 2000 | 0.4955 |
| 0.3715 | 375.0 | 3000 | 0.5152 |
| 0.3564 | 500.0 | 4000 | 0.5275 |
### Framework versions
- Transformers 4.42.0.dev0
- Pytorch 2.3.0+cu121
- Datasets 2.19.1
- Tokenizers 0.19.1
|