google/fleurs
Viewer • Updated • 768k • 57.3k • 402
How to use espnet/wanchichen_fleurs_asr_conformer_hier_lid_utt with ESPnet:
from espnet2.bin.asr_inference import Speech2Text
model = Speech2Text.from_pretrained(
"espnet/wanchichen_fleurs_asr_conformer_hier_lid_utt"
)
speech, rate = soundfile.read("speech.wav")
text, *_ = model(speech)[0]espnet/wanchichen_fleurs_asr_conformer_hier_lid_utt
This model was trained by William Chen using the fleurs recipe in espnet.
cd espnet
pip install -e .
cd egs2/fleurs/asr1
./run.sh
Sat Oct 22 17:36:51 CDT 20223.8.13 (default, Mar 28 2022, 11:38:47) [GCC 7.5.0]espnet 202207pytorch 1.12.1+cu11614fcb2d42b2609f766ffaa7a79e9c921cd8398d9Tue Sep 27 20:02:22 2022 +0000| dataset | Snt | Wrd | Corr | Sub | Del | Ins | Err | S.Err |
|---|---|---|---|---|---|---|---|---|
| decode_lm_lm_train_lm_all_bpe6500_valid.loss.ave_asr_model_valid.acc.ave/dev_all | 31622 | 610500 | 72.9 | 24.4 | 2.7 | 3.1 | 30.2 | 95.5 |
| decode_lm_lm_train_lm_all_bpe6500_valid.loss.ave_asr_model_valid.acc.ave/test_all | 77809 | 1592160 | 72.2 | 25.0 | 2.9 | 3.6 | 31.5 | 96.6 |
| dataset | Snt | Wrd | Corr | Sub | Del | Ins | Err | S.Err |
|---|---|---|---|---|---|---|---|---|
| decode_lm_lm_train_lm_all_bpe6500_valid.loss.ave_asr_model_valid.acc.ave/dev_all | 31622 | 3988181 | 92.6 | 4.7 | 2.6 | 2.2 | 9.6 | 95.5 |
| decode_lm_lm_train_lm_all_bpe6500_valid.loss.ave_asr_model_valid.acc.ave/test_all | 77809 | 10235271 | 92.5 | 4.7 | 2.8 | 2.6 | 10.1 | 96.7 |
| dataset | Snt | Wrd | Corr | Sub | Del | Ins | Err | S.Err |
|---|---|---|---|---|---|---|---|---|
| decode_lm_lm_train_lm_all_bpe6500_valid.loss.ave_asr_model_valid.acc.ave/dev_all | 31622 | 3547834 | 91.4 | 5.8 | 2.8 | 2.5 | 11.0 | 95.4 |
| decode_lm_lm_train_lm_all_bpe6500_valid.loss.ave_asr_model_valid.acc.ave/test_all | 77809 | 9622352 | 91.6 | 5.6 | 2.8 | 2.8 | 11.2 | 96.6 |