Downloads · 30 days
159
7% of all-time downloads
AndrewMcDowell/wav2vec2-xls-r-300m-japanese
wav2vec2-xls-r-300m-japanese is a automatic speech recognition model from AndrewMcDowell. Use it when you need speech turned into text. It is set up for transformers. The card lists the license as apache-2.0.
should probably proofread and complete it, then remove this comment. --
Downloads · 30 days
159
7% of all-time downloads
All-time downloads
2.2K
Public
Repo size
34.1 GB
Likes
0
Public
Click a slice to open those files.
.bin1.3 GB · 100%
From the Hugging Face model README
This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the MOZILLA-FOUNDATION/COMMON_VOICE_8_0 - JA dataset.
Kanji are converted into Hiragana using the pykakasi library during training and evaluation. The model can output both Hiragana and Katakana characters. Since there is no spacing, WER is not a suitable metric for evaluating performance and CER is more suitable.
On mozilla-foundation/common_voice_8_0 it achieved:
On speech-recognition-community-v2/dev_data it achieved:
It achieves the following results on the evaluation set:
More information needed
More information needed
More information needed
The following hyperparameters were used during training:
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 4.0974 | 4.72 | 1000 | 4.0178 | 1.9535 |
| 2.1276 | 9.43 | 2000 | 0.9301 | 1.2128 |
| 1.7622 | 14.15 | 3000 | 0.7103 | 1.5527 |
| 1.6397 | 18.87 | 4000 | 0.6729 | 1.4269 |
| 1.5468 | 23.58 | 5000 | 0.6087 | 1.2497 |
| 1.4885 | 28.3 | 6000 | 0.5786 | 1.3222 |
| 1.451 | 33.02 | 7000 | 0.5726 | 1.3768 |
| 1.3912 | 37.74 | 8000 | 0.5518 | 1.2497 |
| 1.3617 | 42.45 | 9000 | 0.5352 | 1.2694 |
| 1.3113 | 47.17 | 10000 | 0.5228 | 1.2781 |
mozilla-foundation/common_voice_8_0 with split testpython ./eval.py --model_id AndrewMcDowell/wav2vec2-xls-r-300m-japanese --dataset mozilla-foundation/common_voice_8_0 --config ja --split test --log_outputs
mozilla-foundation/common_voice_8_0 with split testpython ./eval.py --model_id AndrewMcDowell/wav2vec2-xls-r-300m-japanese --dataset speech-recognition-community-v2/dev_data --config de --split validation --chunk_length_s 5.0 --stride_length_s 1.0