infinitejoy
commited on
Commit
•
e274b6c
1
Parent(s):
e8f5ea6
update model card README.md
Browse files
README.md
ADDED
@@ -0,0 +1,79 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
language:
|
3 |
+
- br
|
4 |
+
license: apache-2.0
|
5 |
+
tags:
|
6 |
+
- automatic-speech-recognition
|
7 |
+
- mozilla-foundation/common_voice_7_0
|
8 |
+
- generated_from_trainer
|
9 |
+
datasets:
|
10 |
+
- common_voice
|
11 |
+
model-index:
|
12 |
+
- name: wav2vec2-large-xls-r-300m-breton
|
13 |
+
results: []
|
14 |
+
---
|
15 |
+
|
16 |
+
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
17 |
+
should probably proofread and complete it, then remove this comment. -->
|
18 |
+
|
19 |
+
# wav2vec2-large-xls-r-300m-breton
|
20 |
+
|
21 |
+
This model is a fine-tuned version of [facebook/wav2vec2-xls-r-300m](https://huggingface.co/facebook/wav2vec2-xls-r-300m) on the MOZILLA-FOUNDATION/COMMON_VOICE_7_0 - BR dataset.
|
22 |
+
It achieves the following results on the evaluation set:
|
23 |
+
- Loss: 0.6102
|
24 |
+
- Wer: 0.4455
|
25 |
+
|
26 |
+
## Model description
|
27 |
+
|
28 |
+
More information needed
|
29 |
+
|
30 |
+
## Intended uses & limitations
|
31 |
+
|
32 |
+
More information needed
|
33 |
+
|
34 |
+
## Training and evaluation data
|
35 |
+
|
36 |
+
More information needed
|
37 |
+
|
38 |
+
## Training procedure
|
39 |
+
|
40 |
+
### Training hyperparameters
|
41 |
+
|
42 |
+
The following hyperparameters were used during training:
|
43 |
+
- learning_rate: 7e-05
|
44 |
+
- train_batch_size: 32
|
45 |
+
- eval_batch_size: 32
|
46 |
+
- seed: 42
|
47 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
48 |
+
- lr_scheduler_type: linear
|
49 |
+
- lr_scheduler_warmup_steps: 500
|
50 |
+
- num_epochs: 50.0
|
51 |
+
- mixed_precision_training: Native AMP
|
52 |
+
|
53 |
+
### Training results
|
54 |
+
|
55 |
+
| Training Loss | Epoch | Step | Validation Loss | Wer |
|
56 |
+
|:-------------:|:-----:|:----:|:---------------:|:------:|
|
57 |
+
| 2.9205 | 3.33 | 500 | 2.8659 | 1.0 |
|
58 |
+
| 1.6403 | 6.67 | 1000 | 0.9440 | 0.7593 |
|
59 |
+
| 1.3483 | 10.0 | 1500 | 0.7580 | 0.6215 |
|
60 |
+
| 1.2255 | 13.33 | 2000 | 0.6851 | 0.5722 |
|
61 |
+
| 1.1139 | 16.67 | 2500 | 0.6409 | 0.5220 |
|
62 |
+
| 1.0688 | 20.0 | 3000 | 0.6245 | 0.5055 |
|
63 |
+
| 0.99 | 23.33 | 3500 | 0.6142 | 0.4874 |
|
64 |
+
| 0.9345 | 26.67 | 4000 | 0.5946 | 0.4829 |
|
65 |
+
| 0.9058 | 30.0 | 4500 | 0.6229 | 0.4704 |
|
66 |
+
| 0.8683 | 33.33 | 5000 | 0.6153 | 0.4666 |
|
67 |
+
| 0.8367 | 36.67 | 5500 | 0.5952 | 0.4542 |
|
68 |
+
| 0.8162 | 40.0 | 6000 | 0.6030 | 0.4541 |
|
69 |
+
| 0.8042 | 43.33 | 6500 | 0.5972 | 0.4485 |
|
70 |
+
| 0.7836 | 46.67 | 7000 | 0.6070 | 0.4497 |
|
71 |
+
| 0.7556 | 50.0 | 7500 | 0.6102 | 0.4455 |
|
72 |
+
|
73 |
+
|
74 |
+
### Framework versions
|
75 |
+
|
76 |
+
- Transformers 4.16.0.dev0
|
77 |
+
- Pytorch 1.10.1+cu102
|
78 |
+
- Datasets 1.17.1.dev0
|
79 |
+
- Tokenizers 0.11.0
|