patrickvonplaten
commited on
Commit
•
e479cfb
1
Parent(s):
2502577
Update README.md
Browse files
README.md
CHANGED
@@ -50,10 +50,25 @@ logits = model(input_values).logits
|
|
50 |
print(transcription)
|
51 |
```
|
52 |
|
|
|
|
|
|
|
|
|
|
|
|
|
53 |
|
54 |
| Model | WER | CER |
|
55 |
| ------------- | ------------- | ------------- |
|
56 |
-
|
|
57 |
-
|
|
58 |
-
|
59 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
50 |
print(transcription)
|
51 |
```
|
52 |
|
53 |
+
**Improvement**
|
54 |
+
|
55 |
+
This model has been compared on 512 speech samples from the Spanish Common Voice Test set and
|
56 |
+
gives a nice *20 %* performance boost:
|
57 |
+
|
58 |
+
The results can be reproduced by running *from this model repository*:
|
59 |
|
60 |
| Model | WER | CER |
|
61 |
| ------------- | ------------- | ------------- |
|
62 |
+
| patrickvonplaten/wav2vec2-large-xlsr-53-spanish-with-lm | **8.44%** | **2.93%** |
|
63 |
+
| jonatasgrosman/wav2vec2-large-xlsr-53-spanish | **10.20%** | **3.24%** |
|
64 |
+
|
65 |
+
```
|
66 |
+
bash run_ngram_wav2vec2.py 1 512
|
67 |
+
```
|
68 |
+
|
69 |
+
```
|
70 |
+
bash run_ngram_wav2vec2.py 0 512
|
71 |
+
```
|
72 |
+
|
73 |
+
with `run_ngram_wav2vec2.py` being
|
74 |
+
https://huggingface.co/patrickvonplaten/wav2vec2-large-xlsr-53-spanish-with-lm/blob/main/run_ngram_wav2vec2.py
|