SR Steinkamp commited on
Commit
4bd6647
·
1 Parent(s): 6f1a318

updated WER score

Browse files
Files changed (1) hide show
  1. README.md +6 -6
README.md CHANGED
@@ -1,3 +1,4 @@
 
1
  language: nl
2
  datasets:
3
  - common_voicemetrics:
@@ -11,7 +12,7 @@ license: apache-2.0
11
  model-index:
12
  - name: `simonsr XLSR Wav2Vec2 Large 53`
13
  results:
14
- - task:
15
  name: Speech Recognition
16
  type: automatic-speech-recognition
17
  dataset:
@@ -21,7 +22,7 @@ model-index:
21
  metrics:
22
  - name: Test WER
23
  type: wer
24
- value: {wer_result_on_test} #TODO (IMPORTANT): replace {wer_result_on_test} with the WER error rate you achieved on the common_voice test set. It should be in the format XX.XX (don't add the % sign here). **Please** remember to fill out this value after you evaluated your model, so that your model appears on the leaderboard. If you fill out this model card before evaluating your model, please remember to edit the model card afterward to fill in your value
25
  ---
26
 
27
  # Wav2Vec2-Large-XLSR-53-Dutch
@@ -79,7 +80,7 @@ from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
79
  import unidecode
80
  import re
81
 
82
- test_dataset = load_dataset("common_voice", "nl", split="test")
83
  wer = load_metric("wer")
84
 
85
  processor = Wav2Vec2Processor.from_pretrained("{model_id}") #TODO: replace {model_id} with your model id. The model id consists of {your_username}/{your_modelname}, *e.g.* `elgeish/wav2vec2-large-xlsr-53-arabic`
@@ -91,7 +92,7 @@ resampler = torchaudio.transforms.Resample(48_000, 16_000)
91
 
92
  # Preprocessing the datasets.
93
  # We need to read the aduio files as arrays
94
- def speech_file_to_array_fn(batch):
95
  batch["sentence"] = unidecode.unidecode(batch["sentence"])
96
  batch["sentence"] = re.sub(chars_to_ignore_regex, '', batch["sentence"]).lower()
97
  speech_array, sampling_rate = torchaudio.load(batch["path"])
@@ -117,8 +118,7 @@ result = test_dataset.map(evaluate, batched=True, batch_size=8)
117
  print("WER: {:2f}".format(100 * wer.compute(predictions=result["pred_strings"], references=result["sentence"])))
118
  ```
119
 
120
- **Test Result**: XX.XX % # TODO: write output of print here. IMPORTANT: Please remember to also replace {wer_result_on_test} at the top of with this value here. tags.
121
-
122
 
123
  ## Training
124
 
 
1
+ ---
2
  language: nl
3
  datasets:
4
  - common_voicemetrics:
 
12
  model-index:
13
  - name: `simonsr XLSR Wav2Vec2 Large 53`
14
  results:
15
+ - task:
16
  name: Speech Recognition
17
  type: automatic-speech-recognition
18
  dataset:
 
22
  metrics:
23
  - name: Test WER
24
  type: wer
25
+ value: 38.74
26
  ---
27
 
28
  # Wav2Vec2-Large-XLSR-53-Dutch
 
80
  import unidecode
81
  import re
82
 
83
+ test_dataset = load_dataset("common_voice", "nl", split="test")
84
  wer = load_metric("wer")
85
 
86
  processor = Wav2Vec2Processor.from_pretrained("{model_id}") #TODO: replace {model_id} with your model id. The model id consists of {your_username}/{your_modelname}, *e.g.* `elgeish/wav2vec2-large-xlsr-53-arabic`
 
92
 
93
  # Preprocessing the datasets.
94
  # We need to read the aduio files as arrays
95
+ def speech_file_to_array_fn(batch):
96
  batch["sentence"] = unidecode.unidecode(batch["sentence"])
97
  batch["sentence"] = re.sub(chars_to_ignore_regex, '', batch["sentence"]).lower()
98
  speech_array, sampling_rate = torchaudio.load(batch["path"])
 
118
  print("WER: {:2f}".format(100 * wer.compute(predictions=result["pred_strings"], references=result["sentence"])))
119
  ```
120
 
121
+ **Test Result**: 38.74 %
 
122
 
123
  ## Training
124