Merge branch 'master' into pt1.13.1

egs/an4/asr1/local/data_prep.py

@@ -35,7 +35,6 @@
     ) as wav_scp_f, open(
         os.path.join("data", x, "utt2spk"), "w"
     ) as utt2spk_f:
-
         text_f.truncate()
         wav_scp_f.truncate()
         utt2spk_f.truncate()

egs/csj/align1/local/gather_transcript.py

@@ -3,7 +3,6 @@
 import sys
 
 if __name__ == "__main__":
-
     texts = {}
 
     with open(sys.argv[1], "r", encoding="utf-8") as f:

egs/iwslt19/st1/local/filter_offlimit.py

@@ -14,7 +14,6 @@
 
 
 def main():
-
     offlimit_spk_list = []
     with codecs.open(args.offlimit_list, "r", encoding="utf-8") as f:
         for line in f:

egs/iwslt21/asr1/local/filter_parentheses.py

@@ -16,7 +16,6 @@
 
 
 def main():
-
     p_kanji = regex.compile(r".*\p{Script=Han}+.*")
     p_hiragana = regex.compile(r".*\p{Block=Hiragana}+.*")
     p_katakana = regex.compile(r".*\p{Block=Katakana}+.*")

egs/iwslt21/asr1/local/merge_short_segments.py

@@ -29,7 +29,6 @@
 
 
 def merge(segments, segments_dict):
-
     while True:
         num_merge = 0
         new_segments = deque([])

egs/lrs/avsr1/local/se_batch.py

@@ -56,7 +56,7 @@ def Batch(fdir, snr_l=[]):
     wav_np = np.zeros(
         [len(wav_l), maxlen], np.int16
     )  # numpy array for waveform matrix.
-    for (i, wav) in zip(range(len(wav_l)), wav_l):
+    for i, wav in zip(range(len(wav_l)), wav_l):
         wav_np[i, : len(wav)] = wav  # add waveform to numpy array.
         len_l.append(len(wav))  # append length of waveform to list.
     return wav_np, np.array(len_l, np.int32), np.array(snr_test_l, np.int32), fname_l

egs/mgb2/asr1/local/add_to_datadir.py

@@ -20,12 +20,10 @@
 text_file = open(outdir + "/text", "a")
 
 for line in sys.stdin:
-
     m = re.match(r"\w+speaker(\d+)\w+\s+(.*)", line)
     # print line
 
     if m:
-
         spk = int(m.group(1))
 
         t = m.group(2).split()

egs/mini_an4/asr1/local/data_prep.py

@@ -35,7 +35,6 @@
     ) as wav_scp_f, open(
         os.path.join("data", x, "utt2spk"), "w"
     ) as utt2spk_f:
-
         text_f.truncate()
         wav_scp_f.truncate()
         utt2spk_f.truncate()

egs/mini_an4/asr_mix1/local/data_prep.py

@@ -35,7 +35,6 @@
     ) as wav_scp_f, open(
         os.path.join("data", x, "utt2spk"), "w"
     ) as utt2spk_f:
-
         text_f.truncate()
         wav_scp_f.truncate()
         utt2spk_f.truncate()

egs/puebla_nahuatl/asr1/local/data_prep.py

@@ -38,7 +38,6 @@ def ExtractAudioID(audioname, wav_spk_info=None):
 
 
 def XMLRefine(input_xml, output_xml, readable=False):
-
     """refine trs file into
 
     :param input_xml: original transcriber xml

egs/ru_open_stt/asr1/local/ru_open_stt_prepare_data.py

@@ -101,7 +101,6 @@ def get_uttid(wav):
     with open(odir + "/text", "w", encoding="utf-8") as text, open(
         odir + "/wav.scp", "w"
     ) as wavscp, open(odir + "/utt2spk", "w") as utt2spk:
-
         for utt in subsets[subset]["all"]:
             [uttid, words, wav] = utt
             text.write("{} {}\n".format(uttid, words))

egs/wsj_mix/asr1/local/merge_scp2json.py

@@ -221,7 +221,6 @@ def shape(x):
             ("output", output_lines, output_infos),
             ("other", lines, infos),
         ]:
-
             lis = []
             for idx, (line_list, info_list) in enumerate(zip(_lines, _infos), 1):
                 if inout == "input":

egs/yoloxochitl_mixtec/asr1/local/data_prep.py

@@ -40,7 +40,6 @@ def ExtractAudioID(audioname, wav_spk_info=None):
 
 
 def XMLRefine(input_xml, output_xml, readable=False):
-
     if readable:
         append = "\n"
     else:

egs2/TEMPLATE/asr1/pyscripts/feats/mean_pool_scp.py

@@ -74,7 +74,6 @@ def main():
         with open(save_path + ".lengths", "w") as lengths_out, open(
             scp_file, "w"
         ) as sf:
-
             for length, utt_id in tqdm.tqdm(zip(lengths, utt_ids)):
                 utt_id = utt_id.rstrip()
                 length = int(length)

egs2/TEMPLATE/asr1/pyscripts/utils/extract_xvectors.py

@@ -135,7 +135,6 @@ def main(argv):
         device = "cpu"
 
     if args.toolkit in ("speechbrain", "rawnet"):
-
         # Prepare spk2utt for mean x-vector
         spk2utt = dict()
         with open(os.path.join(args.in_folder, "spk2utt"), "r") as reader:

egs2/TEMPLATE/asr1/pyscripts/utils/score_summarization.py

@@ -34,7 +34,7 @@
 
 nlg = NLGEval()  # loads the models
 print("Key", "\t", "METEOR", "\t", "ROUGE-L")
-for (key, ref, hyp) in zip(keys, labels, decoded_preds):
+for key, ref, hyp in zip(keys, labels, decoded_preds):
     metrics_dict = nlg.compute_individual_metrics([ref], hyp)
     print(key, "\t", metrics_dict["METEOR"], "\t", metrics_dict["ROUGE_L"])
 refs = [[x] for x in labels]

egs2/TEMPLATE/ssl1/pyscripts/sklearn_km.py

@@ -99,7 +99,6 @@ def get_mfcc_feature(feats_dir, fs, nj, portion):
 
 
 def get_hubert_feature(feats_dir, fs, portion, url, dir, layer):
-
     reader = HubertFeatureReader(fs, url, dir, layer)
     generator, num = get_path_iterator(f"{feats_dir}/wav.scp", portion)
     iterator = generator()

egs2/clarity21/enh1/local/prep_data.py

@@ -18,7 +18,6 @@
 
 
 def prepare_data(clarity_root, samplerate):
-
     output_folder = "./data"
     ids = {"train": set(), "dev": set()}
 

egs2/cmu_arctic/tts1/local/get_utt_list.py

@@ -14,9 +14,7 @@
 fdev = open(utt_dev_list, "w")
 feval = open(utt_eval_list, "w")
 for spk in ["slt", "clb", "bdl", "rms", "jmk", "awb", "ksp"]:
-
     for w in glob.glob("downloads/cmu_us_all_arctic/wav/{}_*.wav".format(spk, spk)):
-
         # check if in train list
         file_suffix = os.path.basename(w).split("_")[2]
 

egs2/dcase22_task1/asr1/local/data_prep.py

@@ -25,7 +25,6 @@
     with open(os.path.join("data", x, "text"), "w") as text_f, open(
         os.path.join("data", x, "wav.scp"), "w"
     ) as wav_scp_f, open(os.path.join("data", x, "utt2spk"), "w") as utt2spk_f:
-
         text_f.truncate()
         wav_scp_f.truncate()
         utt2spk_f.truncate()

egs2/dns_icassp22/enh1/local/noisyspeech_synthesizer.py

@@ -38,7 +38,6 @@
 
 
 def add_pyreverb(clean_speech, rir):
-
     reverb_speech = signal.fftconvolve(clean_speech, rir, mode="full")
 
     # make reverb_speech same length as clean_speech
@@ -86,7 +85,6 @@ def build_audio(is_clean, params, index, audio_samples_length=-1):
     # iterate through multiple clips until we have a long enough signal
     tries_left = MAXTRIES
     while remaining_length > 0 and tries_left > 0:
-
         # read next audio file and resample if necessary
 
         idx = (idx + 1) % np.size(source_files)
@@ -583,5 +581,4 @@ def main_body():
 
 
 if __name__ == "__main__":
-
     main_body()

egs2/dsing/asr1/local/data_prep.py

@@ -19,7 +19,6 @@ def __init__(self, name, workspace, db_path):
         self.db_path = db_path
 
     def add_utterance(self, utt, recording):
-
         text = utt["text"]
         arrangement, performance, country, gender, user = recording[:-4].split("-")
 

egs2/fsc/asr1/local/data_prep.py

@@ -30,7 +30,6 @@
     ) as transcript_f, open(
         os.path.join("data", x, "utt2spk"), "w"
     ) as utt2spk_f:
-
         text_f.truncate()
         wav_scp_f.truncate()
         utt2spk_f.truncate()

egs2/fsc_challenge/asr1/local/data_prep.py

@@ -30,7 +30,6 @@
     ) as transcript_f, open(
         os.path.join("data", x, "utt2spk"), "w"
     ) as utt2spk_f:
-
         text_f.truncate()
         wav_scp_f.truncate()
         utt2spk_f.truncate()

egs2/fsc_challenge/slu1/local/data_prep.py

@@ -30,7 +30,6 @@
     ) as transcript_f, open(
         os.path.join("data", x, "utt2spk"), "w"
     ) as utt2spk_f:
-
         text_f.truncate()
         wav_scp_f.truncate()
         utt2spk_f.truncate()

egs2/fsc_challenge/slu1/local/data_prep_gigaspeech.py

@@ -49,7 +49,6 @@
     ) as transcript_f, open(
         os.path.join("data", x, "utt2spk"), "w"
     ) as utt2spk_f:
-
         text_f.truncate()
         wav_scp_f.truncate()
         utt2spk_f.truncate()

egs2/fsc_challenge/slu1/local/data_prep_gt.py

@@ -31,7 +31,6 @@
     ) as transcript_f, open(
         os.path.join("data", x, "utt2spk"), "w"
     ) as utt2spk_f:
-
         text_f.truncate()
         wav_scp_f.truncate()
         utt2spk_f.truncate()

egs2/fsc_unseen/asr1/local/data_prep.py

@@ -30,7 +30,6 @@
     ) as transcript_f, open(
         os.path.join("data", x, "utt2spk"), "w"
     ) as utt2spk_f:
-
         text_f.truncate()
         wav_scp_f.truncate()
         utt2spk_f.truncate()

egs2/harpervalley/asr1/local/data_prep.py

@@ -33,7 +33,6 @@ def process_data(target_dir, source_dir, audio_dir, filename, min_length):
     ) as segments, open(
         os.path.join(target_dir, "text"), "a", encoding="utf-8"
     ) as text:
-
         metadata_f = load_json(os.path.join(source_dir, "metadata", filename + ".json"))
         transcript_f = load_json(
             os.path.join(source_dir, "transcript", filename + ".json")

egs2/iam/ocr1/local/data_prep.py

@@ -48,7 +48,6 @@ def prepare_text(lines_file_path, output_dir, split_ids):
     skipped_ids = []
     with open(lines_file_path) as lines_file:
         for line in lines_file.readlines():
-
             # skip comment lines
             if line[0] == "#":
                 continue
@@ -59,7 +58,6 @@ def prepare_text(lines_file_path, output_dir, split_ids):
             line_id = line_split[0]
 
             if line_id in split_ids:
-
                 # extract and format transcription into Kaldi style
                 transcription = " ".join(line_split[8:])
                 transcription = transcription.replace("|", " ")
@@ -175,7 +173,6 @@ def prepare_feats(
 
 
 if __name__ == "__main__":
-
     downloads_dir = "downloads/"
     data_dir = "data/"
 

egs2/l3das22/enh1/local/metric.py

@@ -33,7 +33,6 @@ def wer(clean_speech, denoised_speech):
     """
 
     def _transcription(clean_speech, denoised_speech):
-
         # transcribe clean audio
         input_values = wer_tokenizer(clean_speech, return_tensors="pt").input_values
         logits = wer_model(input_values).logits

egs2/librispeech/ssl1/local/hubert_feature_loader.py

@@ -139,9 +139,12 @@ def get_feats(self, data, ref_len=None):
             x = torch.from_numpy(x).float().to(self.device)
             x = x.view(1, -1)
 
-            feat = self.model.wav2vec2.extract_features(x, num_layers=self.layer,)[0][
-                -1
-            ][
+            feat = self.model.wav2vec2.extract_features(
+                x,
+                num_layers=self.layer,
+            )[
+                0
+            ][-1][
                 0
             ]  # (time, feat_dim)
         return feat.cpu()

egs2/librispeech/ssl1/local/learn_kmeans.py

@@ -86,7 +86,6 @@ def get_km_model(
 
 
 def load_feature_shard(rspecifier, in_filetype, percent):
-
     feats = []
     for utt, feat in file_reader_helper(rspecifier, in_filetype):
         feats.append(feat)

egs2/lrs2/lipreading1/local/feature_extract/models/pretrained.py

@@ -88,7 +88,6 @@ def _make_layer(self, block, planes, blocks, stride=1):
         return nn.Sequential(*layers)
 
     def forward(self, x):
-
         x = self.layer1(x)
         x = self.layer2(x)
         x = self.layer3(x)

egs2/lrs3/asr1/local/data_prep.py

@@ -128,11 +128,9 @@ def generate_espnet_data(
         wav = []
 
         for speaker_folder in speaker_folders:
-
             spk_id = os.path.basename(speaker_folder)
 
             for wav_file in os.listdir(speaker_folder):
-
                 if not wav_file.endswith(".wav"):
                     continue
                 text_file = wav_file.replace("wav", "txt")

egs2/microsoft_speech/asr1/local/process.py

@@ -128,7 +128,7 @@ def prepare_files(dest_folder, audio_folder, transcriptions, fnames, utt_ids):
             f.write(line)
             f.write("\n")
     with open(os.path.join(dest_folder, "wav.scp"), "w", encoding="utf-8") as f:
-        for (idx, fname) in zip(utt_ids, fnames):
+        for idx, fname in zip(utt_ids, fnames):
             fpath = os.path.join(audio_folder, fname + ".wav")
             line = idx + " " + fpath
             f.write(line)

egs2/ofuton_p_utagoe_db/svs1/local/prep_segments.py

@@ -78,7 +78,6 @@ def make_segment(file_id, labels, threshold=30, sil=["pau", "br", "sil"]):
     segments = []
     segment = SegInfo()
     for label in labels:
-
         if label.label_id in sil:
             if len(segment.segs) > 0:
                 segments.extend(segment.split(threshold=threshold))

egs2/reazonspeech/asr1/local/data.py

@@ -12,7 +12,6 @@ def save_kaldi_format(outdir, ds):
     ) as fp_wav, open(os.path.join(outdir, "utt2spk"), "w") as fp_utt2spk, open(
         os.path.join(outdir, "spk2utt"), "w"
     ) as fp_spk2utt:
-
         for item in ds.sort("name"):
             path = item["audio"]["path"]
 

egs2/sinhala/asr1/local/data_prep.py

@@ -92,7 +92,6 @@ def read_sinhala_data(audio_csv, sentences_csv, export_csv):
     ) as transcript_f, open(
         os.path.join("data", x, "utt2spk"), "w"
     ) as utt2spk_f:
-
         text_f.truncate()
         wav_scp_f.truncate()
         utt2spk_f.truncate()

egs2/slue-voxceleb/asr1/local/data_prep_slue.py

@@ -30,7 +30,6 @@
     ) as transcript_f, open(
         os.path.join("data", x, "utt2spk"), "w"
     ) as utt2spk_f:
-
         text_f.truncate()
         wav_scp_f.truncate()
         utt2spk_f.truncate()

egs2/slue-voxpopuli/asr1/local/data_prep_original_slue_format.py

@@ -41,7 +41,6 @@
     with open(os.path.join("data", x, "text"), "w") as text_f, open(
         os.path.join("data", x, "wav.scp"), "w"
     ) as wav_scp_f, open(os.path.join("data", x, "utt2spk"), "w") as utt2spk_f:
-
         text_f.truncate()
         wav_scp_f.truncate()
         utt2spk_f.truncate()

egs2/slue-voxpopuli/asr1/local/data_prep_original_slue_format_transcript.py

@@ -43,7 +43,6 @@
     ) as transcript_f, open(os.path.join("data", x, "wav.scp"), "w") as wav_scp_f, open(
         os.path.join("data", x, "utt2spk"), "w"
     ) as utt2spk_f:
-
         text_f.truncate()
         wav_scp_f.truncate()
         utt2spk_f.truncate()

egs2/slurp/asr1/local/prepare_slurp_data.py

@@ -28,7 +28,6 @@
     ) as text, open(os.path.join(odir, "wav.scp"), "w") as wavscp, open(
         os.path.join(odir, "utt2spk"), "w"
     ) as utt2spk:
-
         for line in meta:
             prompt = json.loads(line.strip())
             transcript = prompt["sentence"]