Update README.md

README.md

@@ -86,7 +86,7 @@ processor = Wav2Vec2Processor.from_pretrained("Galuh/wav2vec2-large-xlsr-indones
 model = Wav2Vec2ForCTC.from_pretrained("Galuh/wav2vec2-large-xlsr-indonesian") 
 model.to("cuda")
 
-chars_to_ignore_regex = '[\\\\\\\\,\\\\\\\\?\\\\\\\\.\\\\\\\\!\\\\\\\\-\\\\\\\\;\\\\\\\\:\\\\\\\\"\\\\\\\\“]'
+chars_to_ignore_regex = '[\,\?\.\!\-\;\:\"\“\%\‘\'\”\�]'
 
 resampler = torchaudio.transforms.Resample(48_000, 16_000)
 
@@ -95,6 +95,7 @@ resampler = torchaudio.transforms.Resample(48_000, 16_000)
 def speech_file_to_array_fn(batch):
   batch["sentence"] = re.sub(chars_to_ignore_regex, '', batch["sentence"]).lower()
   speech_array, sampling_rate = torchaudio.load(batch["path"])
+  resampler = torchaudio.transforms.Resample(sampling_rate, 16_000)
   batch["speech"] = resampler(speech_array).squeeze().numpy()
   return batch
 
@@ -108,7 +109,7 @@ def evaluate(batch):
   with torch.no_grad():
     logits = model(inputs.input_values.to("cuda"), attention_mask=inputs.attention_mask.to("cuda")).logits
 
-    pred_ids = torch.argmax(logits, dim=-1)
+  pred_ids = torch.argmax(logits, dim=-1)
   batch["pred_strings"] = processor.batch_decode(pred_ids)
   return batch
 
@@ -117,7 +118,7 @@ result = test_dataset.map(evaluate, batched=True, batch_size=8)
 print("WER: {:2f}".format(100 * wer.compute(predictions=result["pred_strings"], references=result["sentence"])))
 ```
 
-**Test Result**: 21.07 %
+**Test Result**: 18.32 %
 
 ## Training