Commit
•
943fd9c
1
Parent(s):
9446f06
Model save
Browse files
README.md
CHANGED
|
@@ -2,8 +2,6 @@
|
|
| 2 |
license: apache-2.0
|
| 3 |
base_model: facebook/wav2vec2-large-xlsr-53
|
| 4 |
tags:
|
| 5 |
-
- automatic-speech-recognition
|
| 6 |
-
- DewiBrynJones/banc-trawsgrifiadau-bangor-clean-with-ccv
|
| 7 |
- generated_from_trainer
|
| 8 |
metrics:
|
| 9 |
- wer
|
|
@@ -17,10 +15,10 @@ should probably proofread and complete it, then remove this comment. -->
|
|
| 17 |
|
| 18 |
# wav2vec2-xlsr-53-ft-btb-ccv-cy
|
| 19 |
|
| 20 |
-
This model is a fine-tuned version of [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) on
|
| 21 |
It achieves the following results on the evaluation set:
|
| 22 |
-
- Loss: 0.
|
| 23 |
-
- Wer: 0.
|
| 24 |
|
| 25 |
## Model description
|
| 26 |
|
|
@@ -45,34 +43,26 @@ The following hyperparameters were used during training:
|
|
| 45 |
- seed: 42
|
| 46 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 47 |
- lr_scheduler_type: linear
|
| 48 |
-
- lr_scheduler_warmup_steps:
|
| 49 |
-
- training_steps:
|
| 50 |
- mixed_precision_training: Native AMP
|
| 51 |
|
| 52 |
### Training results
|
| 53 |
|
| 54 |
-
| Training Loss | Epoch | Step
|
| 55 |
-
|
| 56 |
-
| 4.
|
| 57 |
-
| 1.
|
| 58 |
-
| 0.
|
| 59 |
-
| 0.
|
| 60 |
-
| 0.
|
| 61 |
-
| 0.
|
| 62 |
-
| 0.
|
| 63 |
-
| 0.
|
| 64 |
-
| 0.
|
| 65 |
-
| 0.
|
| 66 |
-
| 0.
|
| 67 |
-
| 0.
|
| 68 |
-
| 0.5352 | 1.0923 | 6500 | 0.5567 | 0.4259 |
|
| 69 |
-
| 0.6231 | 1.1763 | 7000 | 0.8251 | 0.6300 |
|
| 70 |
-
| 0.7995 | 1.2603 | 7500 | 0.6610 | 0.4686 |
|
| 71 |
-
| 0.7696 | 1.3443 | 8000 | 0.7331 | 0.4870 |
|
| 72 |
-
| 0.7978 | 1.4283 | 8500 | 0.7790 | 0.5432 |
|
| 73 |
-
| 0.7844 | 1.5124 | 9000 | 0.8260 | 0.5482 |
|
| 74 |
-
| 0.8211 | 1.5964 | 9500 | 0.8065 | 0.5491 |
|
| 75 |
-
| 0.7744 | 1.6804 | 10000 | 0.7622 | 0.5187 |
|
| 76 |
|
| 77 |
|
| 78 |
### Framework versions
|
|
|
|
| 2 |
license: apache-2.0
|
| 3 |
base_model: facebook/wav2vec2-large-xlsr-53
|
| 4 |
tags:
|
|
|
|
|
|
|
| 5 |
- generated_from_trainer
|
| 6 |
metrics:
|
| 7 |
- wer
|
|
|
|
| 15 |
|
| 16 |
# wav2vec2-xlsr-53-ft-btb-ccv-cy
|
| 17 |
|
| 18 |
+
This model is a fine-tuned version of [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) on an unknown dataset.
|
| 19 |
It achieves the following results on the evaluation set:
|
| 20 |
+
- Loss: 0.5155
|
| 21 |
+
- Wer: 0.3983
|
| 22 |
|
| 23 |
## Model description
|
| 24 |
|
|
|
|
| 43 |
- seed: 42
|
| 44 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 45 |
- lr_scheduler_type: linear
|
| 46 |
+
- lr_scheduler_warmup_steps: 600
|
| 47 |
+
- training_steps: 6000
|
| 48 |
- mixed_precision_training: Native AMP
|
| 49 |
|
| 50 |
### Training results
|
| 51 |
|
| 52 |
+
| Training Loss | Epoch | Step | Validation Loss | Wer |
|
| 53 |
+
|:-------------:|:------:|:----:|:---------------:|:------:|
|
| 54 |
+
| 4.6277 | 0.0840 | 500 | 1.7781 | 0.9507 |
|
| 55 |
+
| 1.2514 | 0.1680 | 1000 | 1.0467 | 0.7117 |
|
| 56 |
+
| 0.986 | 0.2521 | 1500 | 0.8940 | 0.6578 |
|
| 57 |
+
| 0.8702 | 0.3361 | 2000 | 0.8158 | 0.5837 |
|
| 58 |
+
| 0.8063 | 0.4201 | 2500 | 0.7132 | 0.5281 |
|
| 59 |
+
| 0.7518 | 0.5041 | 3000 | 0.7196 | 0.5188 |
|
| 60 |
+
| 0.7045 | 0.5881 | 3500 | 0.6231 | 0.4771 |
|
| 61 |
+
| 0.6726 | 0.6722 | 4000 | 0.6075 | 0.4517 |
|
| 62 |
+
| 0.6355 | 0.7562 | 4500 | 0.5602 | 0.4329 |
|
| 63 |
+
| 0.5938 | 0.8402 | 5000 | 0.5438 | 0.4226 |
|
| 64 |
+
| 0.5842 | 0.9242 | 5500 | 0.5245 | 0.4091 |
|
| 65 |
+
| 0.557 | 1.0082 | 6000 | 0.5155 | 0.3983 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 66 |
|
| 67 |
|
| 68 |
### Framework versions
|