Aku Rouhe
commited on
Commit
•
8859d4f
1
Parent(s):
2bdbcec
Add tags, audionormalizer
Browse files- Full-A.yaml +6 -0
- README.md +18 -0
Full-A.yaml
CHANGED
|
@@ -1,6 +1,12 @@
|
|
| 1 |
num_units: 1750
|
| 2 |
tokenizer: !new:sentencepiece.SentencePieceProcessor
|
| 3 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 4 |
# Feature parameters
|
| 5 |
sample_rate: 16000
|
| 6 |
n_fft: 400
|
|
|
|
| 1 |
num_units: 1750
|
| 2 |
tokenizer: !new:sentencepiece.SentencePieceProcessor
|
| 3 |
|
| 4 |
+
# Audio input normalization:
|
| 5 |
+
# Yes, the input should be 22050 and then features computed with the
|
| 6 |
+
# wrong sample rate...
|
| 7 |
+
audio_normalizer: !new:speechbrain.dataio.preprocess.AudioNormalizer
|
| 8 |
+
sample_rate: 22050
|
| 9 |
+
|
| 10 |
# Feature parameters
|
| 11 |
sample_rate: 16000
|
| 12 |
n_fft: 400
|
README.md
ADDED
|
@@ -0,0 +1,18 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
language: "fi"
|
| 3 |
+
thumbnail:
|
| 4 |
+
tags:
|
| 5 |
+
- automatic-speech-recognition
|
| 6 |
+
- Attention
|
| 7 |
+
- pytorch
|
| 8 |
+
- speechbrain
|
| 9 |
+
|
| 10 |
+
metrics:
|
| 11 |
+
- wer
|
| 12 |
+
- cer
|
| 13 |
+
---
|
| 14 |
+
|
| 15 |
+
|
| 16 |
+
# CRDNN with Attention trained on LP
|
| 17 |
+
|
| 18 |
+
This is a an initial model, partly wrong configuration, just to show an initial example.
|