Update README.md

README.md

@@ -40,8 +40,37 @@ import torchaudio
 from datasets import load_dataset
 from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
 
-## TODO: load from OpenSLR https://openslr.org/44/
-##
+
+def load_dataset_sundanese():
+    root_dir = Path("/dataset/ASR/sundanese")
+    url_sundanese_female = "https://www.openslr.org/resources/44/su_id_female.zip"
+    url_sundanese_male = "https://www.openslr.org/resources/44/su_id_male.zip"
+    data_dirs = [ root_dir/"su_id_female/wavs", root_dir/"su_id_male/wavs" ]
+    filenames = [ root_dir/"su_id_female/line_index.tsv", root_dir/"su_id_male/line_index.tsv" ]
+    
+    if not (root_dir/"su_id_female").exists():
+        !wget -P {root_dir} {url_sundanese_female}
+        !unzip {root_dir}/su_id_female.zip -d {root_dir}
+    if not (root_dir/"su_id_male").exists():    
+        !wget -P {root_dir} {url_sundanese_male} 
+        !unzip {root_dir}/su_id_male.zip -d {root_dir}
+    
+    dfs = []
+    
+    dfs.append(pd.read_csv(filenames[0], sep='\t\t', names=["path", "sentence"]))
+    dfs.append(pd.read_csv(filenames[1], sep='\t\t', names=["path", "sentence"]))
+    
+    for i, dir in enumerate(data_dirs):
+        dfs[i]["path"] = dfs[i].apply(lambda row: str(data_dirs[i]) + "/" + row + ".wav", axis=1)
+    df = pd.concat(dfs)
+    # df = df.sample(frac=1, random_state=1).reset_index(drop=True)
+    dataset = Dataset.from_pandas(df)
+    dataset = dataset.remove_columns('__index_level_0__')
+    
+    return dataset.train_test_split(test_size=0.1, seed=1)
+    
+dataset = load_dataset_sundanese()
+test_dataset = dataset['test']
 
 processor = Wav2Vec2Processor.from_pretrained("cahya/wav2vec2-large-xlsr-sundanese")
 model = Wav2Vec2ForCTC.from_pretrained("cahya/wav2vec2-large-xlsr-sundanese")
@@ -51,15 +80,15 @@ resampler = torchaudio.transforms.Resample(48_000, 16_000)
 # Preprocessing the datasets.
 # We need to read the aduio files as arrays
 def speech_file_to_array_fn(batch):
-\tspeech_array, sampling_rate = torchaudio.load(batch["path"])
-\tbatch["speech"] = resampler(speech_array).squeeze().numpy()
-\treturn batch
+\\tspeech_array, sampling_rate = torchaudio.load(batch["path"])
+\\tbatch["speech"] = resampler(speech_array).squeeze().numpy()
+\\treturn batch
 
 test_dataset = test_dataset.map(speech_file_to_array_fn)
 inputs = processor(test_dataset["speech"][:2], sampling_rate=16_000, return_tensors="pt", padding=True)
 
 with torch.no_grad():
-\tlogits = model(inputs.input_values, attention_mask=inputs.attention_mask).logits
+\\tlogits = model(inputs.input_values, attention_mask=inputs.attention_mask).logits
 
 predicted_ids = torch.argmax(logits, dim=-1)
 
@@ -86,31 +115,31 @@ processor = Wav2Vec2Processor.from_pretrained("cahya/wav2vec2-large-xlsr-sundane
 model = Wav2Vec2ForCTC.from_pretrained("cahya/wav2vec2-large-xlsr-sundanese") 
 model.to("cuda")
 
-chars_to_ignore_regex = '[\\,\\?\\.\\!\\-\\;\\:\\"\\“\\%\\‘\\'\\”]'
+chars_to_ignore_regex = '[\\\\,\\\\?\\\\.\\\\!\\\\-\\\\;\\\\:\\\\"\\\\“\\\\%\\\\‘\\\\'\\\\”]'
 
 resampler = torchaudio.transforms.Resample(48_000, 16_000)
 
 # Preprocessing the datasets.
 # We need to read the aduio files as arrays
 def speech_file_to_array_fn(batch):
-\tbatch["sentence"] = re.sub(chars_to_ignore_regex, '', batch["sentence"]).lower()
-\tspeech_array, sampling_rate = torchaudio.load(batch["path"])
-\tbatch["speech"] = resampler(speech_array).squeeze().numpy()
-\treturn batch
+\\tbatch["sentence"] = re.sub(chars_to_ignore_regex, '', batch["sentence"]).lower()
+\\tspeech_array, sampling_rate = torchaudio.load(batch["path"])
+\\tbatch["speech"] = resampler(speech_array).squeeze().numpy()
+\\treturn batch
 
 test_dataset = test_dataset.map(speech_file_to_array_fn)
 
 # Preprocessing the datasets.
 # We need to read the aduio files as arrays
 def evaluate(batch):
-\tinputs = processor(batch["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
+\\tinputs = processor(batch["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
 
-\twith torch.no_grad():
-\t\tlogits = model(inputs.input_values.to("cuda"), attention_mask=inputs.attention_mask.to("cuda")).logits
+\\twith torch.no_grad():
+\\t\\tlogits = model(inputs.input_values.to("cuda"), attention_mask=inputs.attention_mask.to("cuda")).logits
 
     pred_ids = torch.argmax(logits, dim=-1)
-\tbatch["pred_strings"] = processor.batch_decode(pred_ids)
-\treturn batch
+\\tbatch["pred_strings"] = processor.batch_decode(pred_ids)
+\\treturn batch
 
 result = test_dataset.map(evaluate, batched=True, batch_size=8)