Commit
•
31dbddd
1
Parent(s):
87623e8
update model card README.md
Browse files
README.md
CHANGED
|
@@ -1,8 +1,6 @@
|
|
| 1 |
---
|
| 2 |
tags:
|
| 3 |
- generated_from_trainer
|
| 4 |
-
metrics:
|
| 5 |
-
- wer
|
| 6 |
model-index:
|
| 7 |
- name: wav2vec2-large-asr-th-2
|
| 8 |
results: []
|
|
@@ -14,10 +12,6 @@ should probably proofread and complete it, then remove this comment. -->
|
|
| 14 |
# wav2vec2-large-asr-th-2
|
| 15 |
|
| 16 |
This model was trained from scratch on the None dataset.
|
| 17 |
-
It achieves the following results on the evaluation set:
|
| 18 |
-
- Loss: 0.2336
|
| 19 |
-
- Wer: 0.3870
|
| 20 |
-
- Cer: 0.1091
|
| 21 |
|
| 22 |
## Model description
|
| 23 |
|
|
@@ -37,27 +31,17 @@ More information needed
|
|
| 37 |
|
| 38 |
The following hyperparameters were used during training:
|
| 39 |
- learning_rate: 0.0002
|
| 40 |
-
- train_batch_size:
|
| 41 |
- eval_batch_size: 8
|
| 42 |
- seed: 42
|
| 43 |
- gradient_accumulation_steps: 3
|
| 44 |
-
- total_train_batch_size:
|
| 45 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 46 |
- lr_scheduler_type: linear
|
| 47 |
- lr_scheduler_warmup_steps: 600
|
| 48 |
-
- training_steps:
|
| 49 |
- mixed_precision_training: Native AMP
|
| 50 |
|
| 51 |
-
### Training results
|
| 52 |
-
|
| 53 |
-
| Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
|
| 54 |
-
|:-------------:|:-----:|:----:|:---------------:|:------:|:------:|
|
| 55 |
-
| 0.3757 | 0.24 | 1000 | 0.3304 | 0.4416 | 0.1280 |
|
| 56 |
-
| 0.3588 | 0.47 | 2000 | 0.3101 | 0.4273 | 0.1200 |
|
| 57 |
-
| 0.3936 | 0.71 | 3000 | 0.2693 | 0.4036 | 0.1149 |
|
| 58 |
-
| 0.6387 | 0.95 | 4000 | 0.2336 | 0.3870 | 0.1091 |
|
| 59 |
-
|
| 60 |
-
|
| 61 |
### Framework versions
|
| 62 |
|
| 63 |
- Transformers 4.27.3
|
|
|
|
| 1 |
---
|
| 2 |
tags:
|
| 3 |
- generated_from_trainer
|
|
|
|
|
|
|
| 4 |
model-index:
|
| 5 |
- name: wav2vec2-large-asr-th-2
|
| 6 |
results: []
|
|
|
|
| 12 |
# wav2vec2-large-asr-th-2
|
| 13 |
|
| 14 |
This model was trained from scratch on the None dataset.
|
|
|
|
|
|
|
|
|
|
|
|
|
| 15 |
|
| 16 |
## Model description
|
| 17 |
|
|
|
|
| 31 |
|
| 32 |
The following hyperparameters were used during training:
|
| 33 |
- learning_rate: 0.0002
|
| 34 |
+
- train_batch_size: 32
|
| 35 |
- eval_batch_size: 8
|
| 36 |
- seed: 42
|
| 37 |
- gradient_accumulation_steps: 3
|
| 38 |
+
- total_train_batch_size: 96
|
| 39 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 40 |
- lr_scheduler_type: linear
|
| 41 |
- lr_scheduler_warmup_steps: 600
|
| 42 |
+
- training_steps: 5000
|
| 43 |
- mixed_precision_training: Native AMP
|
| 44 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 45 |
### Framework versions
|
| 46 |
|
| 47 |
- Transformers 4.27.3
|