rebort
commited on
Commit
•
82743c5
1
Parent(s):
98374c9
End of training
Browse files
README.md
CHANGED
|
@@ -5,9 +5,24 @@ tags:
|
|
| 5 |
- generated_from_trainer
|
| 6 |
datasets:
|
| 7 |
- common_voice_13_0
|
|
|
|
|
|
|
| 8 |
model-index:
|
| 9 |
- name: wav2vec2-large-xlsr-53-demo-colab
|
| 10 |
-
results:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 11 |
---
|
| 12 |
|
| 13 |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
|
@@ -17,13 +32,8 @@ should probably proofread and complete it, then remove this comment. -->
|
|
| 17 |
|
| 18 |
This model is a fine-tuned version of [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) on the common_voice_13_0 dataset.
|
| 19 |
It achieves the following results on the evaluation set:
|
| 20 |
-
-
|
| 21 |
-
-
|
| 22 |
-
- eval_runtime: 134.854
|
| 23 |
-
- eval_samples_per_second: 9.418
|
| 24 |
-
- eval_steps_per_second: 1.179
|
| 25 |
-
- epoch: 35.09
|
| 26 |
-
- step: 1000
|
| 27 |
|
| 28 |
## Model description
|
| 29 |
|
|
@@ -50,9 +60,22 @@ The following hyperparameters were used during training:
|
|
| 50 |
- total_train_batch_size: 32
|
| 51 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 52 |
- lr_scheduler_type: linear
|
| 53 |
-
- lr_scheduler_warmup_steps:
|
| 54 |
- num_epochs: 100
|
| 55 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 56 |
### Framework versions
|
| 57 |
|
| 58 |
- Transformers 4.34.0
|
|
|
|
| 5 |
- generated_from_trainer
|
| 6 |
datasets:
|
| 7 |
- common_voice_13_0
|
| 8 |
+
metrics:
|
| 9 |
+
- wer
|
| 10 |
model-index:
|
| 11 |
- name: wav2vec2-large-xlsr-53-demo-colab
|
| 12 |
+
results:
|
| 13 |
+
- task:
|
| 14 |
+
name: Automatic Speech Recognition
|
| 15 |
+
type: automatic-speech-recognition
|
| 16 |
+
dataset:
|
| 17 |
+
name: common_voice_13_0
|
| 18 |
+
type: common_voice_13_0
|
| 19 |
+
config: sah
|
| 20 |
+
split: test
|
| 21 |
+
args: sah
|
| 22 |
+
metrics:
|
| 23 |
+
- name: Wer
|
| 24 |
+
type: wer
|
| 25 |
+
value: 0.5698038864511508
|
| 26 |
---
|
| 27 |
|
| 28 |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
|
|
|
| 32 |
|
| 33 |
This model is a fine-tuned version of [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) on the common_voice_13_0 dataset.
|
| 34 |
It achieves the following results on the evaluation set:
|
| 35 |
+
- Loss: 0.8836
|
| 36 |
+
- Wer: 0.5698
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 37 |
|
| 38 |
## Model description
|
| 39 |
|
|
|
|
| 60 |
- total_train_batch_size: 32
|
| 61 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 62 |
- lr_scheduler_type: linear
|
| 63 |
+
- lr_scheduler_warmup_steps: 500
|
| 64 |
- num_epochs: 100
|
| 65 |
|
| 66 |
+
### Training results
|
| 67 |
+
|
| 68 |
+
| Training Loss | Epoch | Step | Validation Loss | Wer |
|
| 69 |
+
|:-------------:|:-----:|:----:|:---------------:|:------:|
|
| 70 |
+
| 0.1354 | 14.04 | 400 | 0.8703 | 0.6377 |
|
| 71 |
+
| 0.1297 | 28.07 | 800 | 0.8601 | 0.6317 |
|
| 72 |
+
| 0.0937 | 42.11 | 1200 | 0.9103 | 0.6320 |
|
| 73 |
+
| 0.0751 | 56.14 | 1600 | 0.8848 | 0.6044 |
|
| 74 |
+
| 0.0582 | 70.18 | 2000 | 0.8630 | 0.5770 |
|
| 75 |
+
| 0.0492 | 84.21 | 2400 | 0.8889 | 0.5786 |
|
| 76 |
+
| 0.0402 | 98.25 | 2800 | 0.8836 | 0.5698 |
|
| 77 |
+
|
| 78 |
+
|
| 79 |
### Framework versions
|
| 80 |
|
| 81 |
- Transformers 4.34.0
|