update model card README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,72 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
tags:
|
| 3 |
+
- generated_from_trainer
|
| 4 |
+
model-index:
|
| 5 |
+
- name: wav2vec2-xlsr-fi-train-aug-lm-1B
|
| 6 |
+
results: []
|
| 7 |
+
---
|
| 8 |
+
|
| 9 |
+
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
| 10 |
+
should probably proofread and complete it, then remove this comment. -->
|
| 11 |
+
|
| 12 |
+
# wav2vec2-xlsr-fi-train-aug-lm-1B
|
| 13 |
+
|
| 14 |
+
This model was trained from scratch on the None dataset.
|
| 15 |
+
It achieves the following results on the evaluation set:
|
| 16 |
+
- Loss: 0.1499
|
| 17 |
+
- Wer: 0.1955
|
| 18 |
+
|
| 19 |
+
## Model description
|
| 20 |
+
|
| 21 |
+
More information needed
|
| 22 |
+
|
| 23 |
+
## Intended uses & limitations
|
| 24 |
+
|
| 25 |
+
More information needed
|
| 26 |
+
|
| 27 |
+
## Training and evaluation data
|
| 28 |
+
|
| 29 |
+
More information needed
|
| 30 |
+
|
| 31 |
+
## Training procedure
|
| 32 |
+
|
| 33 |
+
### Training hyperparameters
|
| 34 |
+
|
| 35 |
+
The following hyperparameters were used during training:
|
| 36 |
+
- learning_rate: 0.0001
|
| 37 |
+
- train_batch_size: 8
|
| 38 |
+
- eval_batch_size: 8
|
| 39 |
+
- seed: 42
|
| 40 |
+
- gradient_accumulation_steps: 2
|
| 41 |
+
- total_train_batch_size: 16
|
| 42 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 43 |
+
- lr_scheduler_type: linear
|
| 44 |
+
- lr_scheduler_warmup_steps: 100
|
| 45 |
+
- num_epochs: 4
|
| 46 |
+
- mixed_precision_training: Native AMP
|
| 47 |
+
|
| 48 |
+
### Training results
|
| 49 |
+
|
| 50 |
+
| Training Loss | Epoch | Step | Validation Loss | Wer |
|
| 51 |
+
|:-------------:|:-----:|:----:|:---------------:|:------:|
|
| 52 |
+
| 0.6473 | 0.29 | 400 | 0.2857 | 0.3825 |
|
| 53 |
+
| 0.6039 | 0.58 | 800 | 0.2459 | 0.3476 |
|
| 54 |
+
| 0.4757 | 0.87 | 1200 | 0.2338 | 0.3274 |
|
| 55 |
+
| 0.4473 | 1.15 | 1600 | 0.2246 | 0.3128 |
|
| 56 |
+
| 0.4322 | 1.44 | 2000 | 0.1962 | 0.2805 |
|
| 57 |
+
| 0.3961 | 1.73 | 2400 | 0.2070 | 0.2797 |
|
| 58 |
+
| 0.3642 | 2.02 | 2800 | 0.1790 | 0.2473 |
|
| 59 |
+
| 0.3561 | 2.31 | 3200 | 0.1769 | 0.2375 |
|
| 60 |
+
| 0.282 | 2.6 | 3600 | 0.1672 | 0.2263 |
|
| 61 |
+
| 0.2978 | 2.89 | 4000 | 0.1636 | 0.2192 |
|
| 62 |
+
| 0.2722 | 3.17 | 4400 | 0.1637 | 0.2102 |
|
| 63 |
+
| 0.2924 | 3.46 | 4800 | 0.1506 | 0.2021 |
|
| 64 |
+
| 0.2631 | 3.75 | 5200 | 0.1499 | 0.1955 |
|
| 65 |
+
|
| 66 |
+
|
| 67 |
+
### Framework versions
|
| 68 |
+
|
| 69 |
+
- Transformers 4.16.0.dev0
|
| 70 |
+
- Pytorch 1.10.1+cu102
|
| 71 |
+
- Datasets 1.17.1.dev0
|
| 72 |
+
- Tokenizers 0.11.0
|