patrickvonplaten
commited on
Commit
·
dda267d
1
Parent(s):
78168e1
Update README.md
Browse files
README.md
CHANGED
@@ -21,7 +21,7 @@ model-index:
|
|
21 |
metrics:
|
22 |
- name: Test WER
|
23 |
type: wer
|
24 |
-
value: 41.
|
25 |
---
|
26 |
|
27 |
# Wav2Vec2-Large-XLSR-53-Sorbian
|
@@ -84,7 +84,7 @@ processor = Wav2Vec2Processor.from_pretrained("iarfmoose/wav2vec2-large-xlsr-sor
|
|
84 |
model = Wav2Vec2ForCTC.from_pretrained("iarfmoose/wav2vec2-large-xlsr-sorbian")
|
85 |
model.to("cuda")
|
86 |
|
87 |
-
chars_to_ignore_regex = '[
|
88 |
resampler = torchaudio.transforms.Resample(48_000, 16_000)
|
89 |
|
90 |
def speech_file_to_array_fn(batch):
|
@@ -110,7 +110,7 @@ result = test_dataset.map(evaluate, batched=True, batch_size=8)
|
|
110 |
print("WER: {:2f}".format(100 * wer.compute(predictions=result["pred_strings"], references=result["sentence"])))
|
111 |
```
|
112 |
|
113 |
-
**Test Result**: 41.
|
114 |
|
115 |
|
116 |
## Training
|
|
|
21 |
metrics:
|
22 |
- name: Test WER
|
23 |
type: wer
|
24 |
+
value: 41.74
|
25 |
---
|
26 |
|
27 |
# Wav2Vec2-Large-XLSR-53-Sorbian
|
|
|
84 |
model = Wav2Vec2ForCTC.from_pretrained("iarfmoose/wav2vec2-large-xlsr-sorbian")
|
85 |
model.to("cuda")
|
86 |
|
87 |
+
chars_to_ignore_regex = '[\\\\\\\\\\\\\\\\,\\\\\\\\\\\\\\\\?\\\\\\\\\\\\\\\\.\\\\\\\\\\\\\\\\!\\\\\\\\\\\\\\\\-\\\\\\\\\\\\\\\\;\\\\\\\\\\\\\\\\:\\\\\\\\\\\\\\\\"\\\\\\\\\\\\\\\\“\\\\\\\\\\\\\\\\%\\\\\\\\\\\\\\\\‘\\\\\\\\\\\\\\\\”\\\\\\\\\\\\\\\\�\\\\\\\\\\\\\\\\–\\\\\\\\\\\\\\\\—\\\\\\\\\\\\\\\\¬\\\\\\\\\\\\\\\\⅛]'
|
88 |
resampler = torchaudio.transforms.Resample(48_000, 16_000)
|
89 |
|
90 |
def speech_file_to_array_fn(batch):
|
|
|
110 |
print("WER: {:2f}".format(100 * wer.compute(predictions=result["pred_strings"], references=result["sentence"])))
|
111 |
```
|
112 |
|
113 |
+
**Test Result**: 41.74 %
|
114 |
|
115 |
|
116 |
## Training
|