Commit
•
c786d8e
1
Parent(s):
44d0103
Update README.md
Browse files
README.md
CHANGED
|
@@ -1,4 +1,27 @@
|
|
|
|
|
| 1 |
tags:
|
| 2 |
- automatic-speech-recognition
|
| 3 |
- robust-speech-event
|
| 4 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
|
| 2 |
tags:
|
| 3 |
- automatic-speech-recognition
|
| 4 |
- robust-speech-event
|
| 5 |
+
|
| 6 |
+
|
| 7 |
+
---
|
| 8 |
+
|
| 9 |
+
|
| 10 |
+
|
| 11 |
+
|
| 12 |
+
This model is a fine-tuned version of [facebook/wav2vec2-xls-r-300m](https://huggingface.co/facebook/wav2vec2-xls-r-300m) on a private dataset.
|
| 13 |
+
It achieves the following results on the evaluation set:
|
| 14 |
+
|
| 15 |
+
|
| 16 |
+
The following hyper-parameters were used during training:
|
| 17 |
+
- learning_rate: 3e-4
|
| 18 |
+
- train_batch_size: 32
|
| 19 |
+
- eval_batch_size: 16
|
| 20 |
+
- seed: 42
|
| 21 |
+
- gradient_accumulation_steps: 2
|
| 22 |
+
- total_train_batch_size: 128
|
| 23 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 24 |
+
- lr_scheduler_type: linear
|
| 25 |
+
- lr_scheduler_warmup_steps: 800
|
| 26 |
+
- num_epochs: 30
|
| 27 |
+
- mixed_precision_training: Native AMP
|