init

filter_audio.py

@@ -4,20 +4,22 @@ from os.path import join as p_join
 from tqdm import tqdm
 from typing import Dict
 from glob import glob
+import soundfile as sf
 
-from soundfile import LibsndfileError
 from datasets import Audio
 
 # dataset config
-direction = os.getenv("DIRECTION", "enA-jaA")
-sides = {i: n for n, i in enumerate(sorted(direction.split("-")), 1)}
-sides_rev = {v: k for k, v in sides.items()}
+direction_speech = os.getenv("DIRECTION_SPEECH", "enA")
+direction_text = os.getenv("DIRECTION_TEXT", "jpn")
+direction = f"{direction_speech}-{direction_text}"
 cache_dir_audio = p_join("download", "audio", direction)
 cache_dir_feature = p_join("download", "feature", direction)
-os.makedirs(cache_dir_audio, exist_ok=True)
+cache_dir_audio_fixed = p_join("download", "audio_fixed", direction, direction_speech)
 os.makedirs(cache_dir_feature, exist_ok=True)
+os.makedirs(cache_dir_audio, exist_ok=True)
+os.makedirs(cache_dir_audio_fixed, exist_ok=True)
 line_no_start = int(os.getenv("LINE_NO_START", 0))
-line_no_end = int(os.getenv("LINE_NO_END", 100000))
+line_no_end = int(os.getenv("LINE_NO_END", 100))
 
 
 def loader(feature: str) -> Dict:
@@ -25,46 +27,43 @@ def loader(feature: str) -> Dict:
         return json.load(f)
 
 
+# feature dictionary
 files = {
     int(os.path.basename(i).replace(".json", "")): i for i in glob(p_join(cache_dir_feature, "*.json"))
 }
+# feature files
+files_list = [k for k in files.keys() if line_no_start <= k <= line_no_end]
+with open(f"text.{direction}.json") as f:
+    line2text = {int(k): v for k, v in json.load(f).items()}
+text_list = [k for k in line2text.keys() if line_no_start <= k <= line_no_end]
+fixed_audio_list = [int(os.path.basename(i).replace(".mp3", "")) for i in glob(p_join(cache_dir_audio_fixed, "*"))]
+# remove broken audio files
 audio_loader = Audio()
+index_list = [i for i in list(range(line_no_start, line_no_end)) if i in files_list and i in text_list and i not in fixed_audio_list]
+print(f"filtering {len(index_list)} files....")
+for i in tqdm(index_list):
+    features = loader(files[i])
+    audio_file = features[f"{direction_speech}.path"]
+    start, end = features[f"{direction_speech}.duration_start"], features[f"{direction_speech}.duration_end"]
+    if os.path.exists(audio_file):
+        try:
+            wav = audio_loader.decode_example({"path": audio_file, "bytes": None})
+            if start < end < len(wav["array"]):
+                sf.write(p_join(cache_dir_audio_fixed, f"{i}.mp3"), wav["array"][start:end], wav["sampling_rate"])
+        except Exception as e:
+            print(e)
+        os.remove(audio_file)
 
-
-def delete_audio(target_audio_file):
-    if os.path.exists(target_audio_file):
-        os.remove(target_audio_file)
-    line_no = os.path.basename(target_audio_file).split(".")[0]
-    try:
-        feature_file = files[int(line_no)]
-        if os.path.exists(feature_file):
-            os.remove(feature_file)
-    except Exception as e:
-        print(e)
-
-
-# remove broken audio files
-n_broken_files = 0
-for i in tqdm(list(range(line_no_start, line_no_end))):
-    if i not in files:
-        continue
-    i = loader(files[i])
-    for lang_side in [sides_rev[1], sides_rev[2]]:
-        if f"{lang_side}.path" not in i:
-            continue
-        audio_file = i[f"{lang_side}.path"]
-        start, end = i[f"{lang_side}.duration_start"], i[f"{lang_side}.duration_end"]
-        if os.path.exists(audio_file):
-            try:
-                wav = audio_loader.decode_example({"path": audio_file, "bytes": None})
-                if start < end < len(wav["array"]):
-                    pass
-                else:
-                    delete_audio(audio_file)
-                    n_broken_files += 1
-                    print(n_broken_files)
-            except Exception as e:
-                print(e)
-                delete_audio(audio_file)
-                n_broken_files += 1
-                print(n_broken_files)
+#
+# print(f"features (filtered): {len(features)}")
+# dataset_to_push = DatasetDict({"train": audio_dataset})
+# repo_name = f"{hf_org}/{hf_dataset}"
+# while True:
+#     try:
+#         dataset_to_push.push_to_hub(repo_name, config_name=f"subset_{dataset_id}")
+#         break
+#     except Exception:
+#         print(f"FAILED: push_to_hub on {repo_name} failed. wait 60 sec and retry soon...")
+#         time.sleep(60)
+#     json.dump(data_dict["line_no"], f)
+# f_writer.close()

main.sh

@@ -137,11 +137,71 @@ cd ../
 
 export DIRECTION="enA-jpn"
 
-export LINE_NO_START=0
+export LINE_NO_START=50
 export LINE_NO_END=100
-export DATASET_ID="dummy"
+
+
+export LINE_NO_START=0
+export LINE_NO_END=25000
+export DATASET_ID="1"
+python push_s2t_translation.py
+
+export LINE_NO_START=25000
+export LINE_NO_END=50000
+export DATASET_ID="2"
+python push_s2t_translation.py
+
+export LINE_NO_START=50000
+export LINE_NO_END=75000
+export DATASET_ID="3"
 python push_s2t_translation.py
 
+export LINE_NO_START=75000
+export LINE_NO_END=100000
+export DATASET_ID="4"
+python push_s2t_translation.py
+
+export LINE_NO_START=100000
+export LINE_NO_END=125000
+export DATASET_ID="5"
+python push_s2t_translation.py
+
+export LINE_NO_START=125000
+export LINE_NO_END=150000
+export DATASET_ID="6"
+python push_s2t_translation.py
+
+export LINE_NO_START=150000
+export LINE_NO_END=175000
+export DATASET_ID="7"
+python push_s2t_translation.py
+
+export LINE_NO_START=175000
+export LINE_NO_END=200000
+export DATASET_ID="8"
+python push_s2t_translation.py
+
+
+
+
+export LINE_NO_START=100000
+export LINE_NO_END=150000
+export DATASET_ID="0"
+python push_s2t_translation.py
+
+
+export LINE_NO_START=150000
+export LINE_NO_END=300000
+export DATASET_ID="0"
+python push_s2t_translation.py
+
+
+export LINE_NO_START=300000
+export LINE_NO_END=360000
+export DATASET_ID="0"
+python push_s2t_translation.py
+
+
 
 ########
 # NLLB #