add tool_dir as an arg

Author: AudranBert

ssak/utils/kaldi.py

@@ -58,10 +58,10 @@ def parse_line(line):
 }
 
 
-def check_kaldi_dir(dirname, language=None, strict_sort=False):
+def check_kaldi_dir(dirname, language=None, strict_sort=False, tool_dir=None):
     strict_sort = "true" if strict_sort else "false"
-    tool_dir = os.path.join(os.path.dirname(os.path.dirname(os.path.dirname(os.path.realpath(__file__)))), "tools", "kaldi", "utils")
-
+    if not tool_dir:
+        tool_dir = os.path.join(os.path.dirname(os.path.dirname(os.path.dirname(os.path.realpath(__file__)))), "tools", "kaldi", "utils")
     if os.path.isfile(os.path.join(dirname, "text")):
         with open(os.path.join(dirname, "text")) as f:
             texts = dict(parse_line(line) for line in f)

tools/kaldi/datasets2kaldi/yodas2kaldi.py

@@ -13,7 +13,7 @@
     parser = argparse.ArgumentParser(description="Convert yodas dataset to Kaldi format")
     parser.add_argument("--force", action="store_true", default=True)
     parser.add_argument("--input", type=str, default="/data-server/datasets/audio/transcript/fr/YODAS/fr000")
-    parser.add_argument("--output", type=str, default="/data-server/datasets/audio/kaldi/fr/YODAS/fr000_2")
+    parser.add_argument("--output", type=str, default="/data-server/datasets/audio/kaldi/fr/YODAS/fr000")
     args = parser.parse_args()
 
     input_dataset = args.input
@@ -51,7 +51,7 @@
     audio_ids = Row2Info("id", ["audio_id"], 3, None, None)
     spk_ids = Row2Info("id", ["speaker"], 4, None, None)
     dev_reader = Reader2Kaldi(input_dataset, processors=[texts, durations, audios, audio_ids, spk_ids])
-    dataset = dev_reader.load(debug=False, accept_missing_speaker=True)
+    dataset = dev_reader.load(debug=False, accept_missing_speaker=False)
 
     def filter(row):
         if row.id.startswith("E--pPwqi_50-"):