Commit
•
504d95e
1
Parent(s):
bf0f232
Update README.md
Browse filesAdd eval results.
README.md
CHANGED
@@ -9,8 +9,23 @@ tags:
|
|
9 |
datasets:
|
10 |
- common_voice
|
11 |
model-index:
|
12 |
-
- name: ''
|
13 |
-
results:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
14 |
---
|
15 |
|
16 |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
@@ -19,6 +34,9 @@ should probably proofread and complete it, then remove this comment. -->
|
|
19 |
#
|
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_8_0 - JA dataset.
|
|
|
|
|
|
|
22 |
It achieves the following results on the evaluation set:
|
23 |
- Loss: 0.5212
|
24 |
- Wer: 1.3068
|
@@ -72,3 +90,10 @@ The following hyperparameters were used during training:
|
|
72 |
- Pytorch 1.10.2+cu102
|
73 |
- Datasets 1.18.2.dev0
|
74 |
- Tokenizers 0.11.0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
9 |
datasets:
|
10 |
- common_voice
|
11 |
model-index:
|
12 |
+
- name: 'XLS-R-300-m'
|
13 |
+
results:
|
14 |
+
- task:
|
15 |
+
name: Automatic Speech Recognition
|
16 |
+
type: automatic-speech-recognition
|
17 |
+
dataset:
|
18 |
+
name: Common Voice 8
|
19 |
+
type: mozilla-foundation/common_voice_8_0
|
20 |
+
args: ja
|
21 |
+
metrics:
|
22 |
+
- name: Test WER
|
23 |
+
type: wer
|
24 |
+
value: 95.82
|
25 |
+
- name: Test CER
|
26 |
+
type: cer
|
27 |
+
value: 23.64
|
28 |
+
|
29 |
---
|
30 |
|
31 |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
|
|
34 |
#
|
35 |
|
36 |
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_8_0 - JA dataset.
|
37 |
+
|
38 |
+
Kanji are converted into Hiragana using the [pykakasi](https://pykakasi.readthedocs.io/en/latest/index.html) library during training and evaluation. The model can output both Hiragana and Katakana characters.
|
39 |
+
|
40 |
It achieves the following results on the evaluation set:
|
41 |
- Loss: 0.5212
|
42 |
- Wer: 1.3068
|
|
|
90 |
- Pytorch 1.10.2+cu102
|
91 |
- Datasets 1.18.2.dev0
|
92 |
- Tokenizers 0.11.0
|
93 |
+
|
94 |
+
#### Evaluation Commands
|
95 |
+
1. To evaluate on `mozilla-foundation/common_voice_8_0` with split `test`
|
96 |
+
|
97 |
+
```bash
|
98 |
+
python ./eval.py --model_id AndrewMcDowell/wav2vec2-xls-r-300m-japanese --dataset mozilla-foundation/common_voice_8_0 --config ja --split test --log_outputs
|
99 |
+
```
|