init

fetch_dataset_s2s.py

@@ -43,6 +43,8 @@ hf_org = os.getenv("HF_ORG", "asahi417")
 hf_dataset = f"seamless-align-{direction}"
 skip_download = bool(int(os.getenv("SKIP_DOWNLOAD", 0)))
 sampling_rate = 16000  # seamless-align aligns audio in 16kHz
+max_seq_length = 1000000
+min_seq_length = 50000
 
 
 def wget(url: str, output_file: Optional[str] = None):
@@ -131,6 +133,9 @@ def get_audio(dataframe: pd.DataFrame):
                     wav = wav[floor(start / sampling_rate * sr):ceil(end / sampling_rate * sr)]
                     print(f"wav shape (after truncate):{wav.shape}")
                     wav = wav[:int(end/sampling_rate * sr) + sr]
+                    if len(wav) < min_seq_length or len(wav) > max_seq_length:
+                        print(f"invalid wave length: {len(wav)}")
+                        return None
                     print(f"SAVING: {features[f'{side}.path']}")
                     sf.write(features[f"{side}.path"], wav, sr)
                     # if sr != sampling_rate:

main_s2s.sh

@@ -18,16 +18,6 @@ do
   echo ${LINE_NO_START}
   python fetch_dataset_s2s.py
 done
-for i in $(seq 443 500);
-do
-  export N_POOL=15
-  export DATASET_ID=${i}
-  export DIRECTION="enA-frA"
-  export LINE_NO_START=$(((DATASET_ID-1) * 2500))
-  export LINE_NO_END=$((DATASET_ID * 2500))
-  echo ${LINE_NO_START}
-  python fetch_dataset_s2s.py
-done
 for i in $(seq 519 550);
 do
   export N_POOL=15
@@ -90,25 +80,35 @@ do
   python tokenize_dataset_s2s.py
 done
 # speaker embedding
-for i in $(seq );
+export SE_MODEL="metavoice"
+export SE_MODEL="hubert-xl"  # running
+export SE_MODEL="xlsr-2b"  # running
+export SE_MODEL="w2vbert-600m"  # running
+for i in $(seq 1 140);
 do
   export DATASET_ID=${i}
   export DIRECTION="enA-frA"
   python attach_speaker_embedding_s2s.py
 done
-# tokenize
-for i in $(seq 106 425);
+for i in $(seq 200 240);
 do
-  export N_POOL=15
   export DATASET_ID=${i}
   export DIRECTION="enA-frA"
-  export LINE_NO_START=$(((DATASET_ID-1) * 2500))
-  export LINE_NO_END=$((DATASET_ID * 2500))
-  echo ${LINE_NO_START}
-  python tokenize_dataset_s2s.py
+  python attach_speaker_embedding_s2s.py
 done
-# speaker embedding
-for i in $(seq 361 425);
+for i in $(seq 300 380);
+do
+  export DATASET_ID=${i}
+  export DIRECTION="enA-frA"
+  python attach_speaker_embedding_s2s.py
+done
+for i in $(seq 400 460);
+do
+  export DATASET_ID=${i}
+  export DIRECTION="enA-frA"
+  python attach_speaker_embedding_s2s.py
+done
+for i in $(seq 500 570);
 do
   export DATASET_ID=${i}
   export DIRECTION="enA-frA"
@@ -117,162 +117,167 @@ done
 
 
 ######################
-# enA-deA: 1_965_186 #
+# enA-esA: 2_658_022 #
 ######################
 # test
 export DATASET_ID=test
-export DIRECTION="deA-enA"
+export DIRECTION="enA-esA"
 export LINE_NO_START=0
 export LINE_NO_END=10
 python fetch_dataset_s2s.py
 # main
-for i in $(seq 136 200);
+export N_POOL=15
+export DATASET_ID=54
+export DIRECTION="enA-esA"
+export LINE_NO_START=$(((DATASET_ID-1) * 2500))
+export LINE_NO_END=$((DATASET_ID * 2500))
+echo ${LINE_NO_START}
+python fetch_dataset_s2s.py
+for i in $(seq 210 219);
 do
   export N_POOL=15
   export DATASET_ID=${i}
-  export DIRECTION="deA-enA"
+  export DIRECTION="enA-esA"
   export LINE_NO_START=$(((DATASET_ID-1) * 2500))
   export LINE_NO_END=$((DATASET_ID * 2500))
   echo ${LINE_NO_START}
   python fetch_dataset_s2s.py
 done
-for i in $(seq 201 394);
+for i in $(seq 245 250);
 do
   export N_POOL=15
   export DATASET_ID=${i}
-  export DIRECTION="deA-enA"
+  export DIRECTION="enA-esA"
   export LINE_NO_START=$(((DATASET_ID-1) * 2500))
   export LINE_NO_END=$((DATASET_ID * 2500))
   echo ${LINE_NO_START}
   python fetch_dataset_s2s.py
 done
-for i in $(seq 73 91);
+for i in $(seq 345 390);
 do
   export N_POOL=15
   export DATASET_ID=${i}
-  export DIRECTION="deA-enA"
+  export DIRECTION="enA-esA"
+  export LINE_NO_START=$(((DATASET_ID-1) * 2500))
+  export LINE_NO_END=$((DATASET_ID * 2500))
+  echo ${LINE_NO_START}
+  python fetch_dataset_s2s.py
+done
+for i in $(seq 390 420);
+do
+  export N_POOL=15
+  export DATASET_ID=${i}
+  export DIRECTION="enA-esA"
   export LINE_NO_START=$(((DATASET_ID-1) * 2500))
   export LINE_NO_END=$((DATASET_ID * 2500))
   echo ${LINE_NO_START}
   python fetch_dataset_s2s.py
 done
 # tokenize
-#for i in $(seq 73 91);
-#do
-#  export N_POOL=15
-#  export DATASET_ID=${i}
-#  export DIRECTION="deA-enA"
-#  export LINE_NO_START=$(((DATASET_ID-1) * 2500))
-#  export LINE_NO_END=$((DATASET_ID * 2500))
-#  echo ${LINE_NO_START}
-#  python tokenize_dataset_s2s.py
-#done
-for i in $(seq 105 180);
+for i in $(seq 51 210);
 do
   export N_POOL=15
   export DATASET_ID=${i}
-  export DIRECTION="deA-enA"
+  export DIRECTION="enA-esA"
   export LINE_NO_START=$(((DATASET_ID-1) * 2500))
   export LINE_NO_END=$((DATASET_ID * 2500))
   echo ${LINE_NO_START}
   python tokenize_dataset_s2s.py
 done
 # speaker embedding
-for i in $(seq 28 104);
+for i in $(seq 251 280);
 do
   export DATASET_ID=${i}
-  export DIRECTION="deA-enA"
+  export DIRECTION="enA-esA"
   python attach_speaker_embedding_s2s.py
 done
-# speaker embedding
-for i in $(seq 370 394);
+
+export SE_MODEL="w2vbert-600m"  # running
+export SE_MODEL="xlsr-2b"  # running
+export SE_MODEL="hubert-xl"
+export SE_MODEL="metavoice"
+for i in $(seq 26 125);
 do
   export DATASET_ID=${i}
-  export DIRECTION="deA-enA"
+  export DIRECTION="enA-esA"
   python attach_speaker_embedding_s2s.py
 done
+
+
+
+
 ######################
-# enA-esA: 2_658_022 #
+# enA-deA: 1_965_186 #
 ######################
 # test
 export DATASET_ID=test
-export DIRECTION="enA-esA"
+export DIRECTION="deA-enA"
 export LINE_NO_START=0
 export LINE_NO_END=10
 python fetch_dataset_s2s.py
 # main
-for i in $(seq 194 220);
-do
-  export N_POOL=15
-  export DATASET_ID=${i}
-  export DIRECTION="enA-esA"
-  export LINE_NO_START=$(((DATASET_ID-1) * 2500))
-  export LINE_NO_END=$((DATASET_ID * 2500))
-  echo ${LINE_NO_START}
-  python fetch_dataset_s2s.py
-done
-for i in $(seq 220 250);
+for i in $(seq 201 394);
 do
   export N_POOL=15
   export DATASET_ID=${i}
-  export DIRECTION="enA-esA"
+  export DIRECTION="deA-enA"
   export LINE_NO_START=$(((DATASET_ID-1) * 2500))
   export LINE_NO_END=$((DATASET_ID * 2500))
   echo ${LINE_NO_START}
   python fetch_dataset_s2s.py
 done
-for i in $(seq 333 420);
+for i in $(seq 321 353);
 do
   export N_POOL=15
   export DATASET_ID=${i}
-  export DIRECTION="enA-esA"
+  export DIRECTION="deA-enA"
   export LINE_NO_START=$(((DATASET_ID-1) * 2500))
   export LINE_NO_END=$((DATASET_ID * 2500))
   echo ${LINE_NO_START}
   python fetch_dataset_s2s.py
 done
-for i in $(seq 434 532);
+# tokenize
+for i in $(seq 195 200);
 do
   export N_POOL=15
   export DATASET_ID=${i}
-  export DIRECTION="enA-esA"
+  export DIRECTION="deA-enA"
   export LINE_NO_START=$(((DATASET_ID-1) * 2500))
   export LINE_NO_END=$((DATASET_ID * 2500))
   echo ${LINE_NO_START}
-  python fetch_dataset_s2s.py
+  python tokenize_dataset_s2s.py
 done
-# tokenize
-for i in $(seq 1 50);
+for i in $(seq 73 91);
 do
   export N_POOL=15
   export DATASET_ID=${i}
-  export DIRECTION="enA-esA"
+  export DIRECTION="deA-enA"
   export LINE_NO_START=$(((DATASET_ID-1) * 2500))
   export LINE_NO_END=$((DATASET_ID * 2500))
   echo ${LINE_NO_START}
   python tokenize_dataset_s2s.py
 done
 # speaker embedding
-for i in $(seq 251 280);
+for i in $(seq 28 104);
 do
   export DATASET_ID=${i}
-  export DIRECTION="enA-esA"
+  export DIRECTION="deA-enA"
   python attach_speaker_embedding_s2s.py
 done
-
-export SE_MODEL="w2vbert-600m"  # running
-export SE_MODEL="xlsr-2b"  # running
-export SE_MODEL="hubert-xl"
-export SE_MODEL="metavoice"
-for i in $(seq 26 125);
+# speaker embedding
+export SE_MODEL="w2vbert-600m"
+for i in $(seq 1 320);
 do
   export DATASET_ID=${i}
-  export DIRECTION="enA-esA"
+  export DIRECTION="deA-enA"
   python attach_speaker_embedding_s2s.py
 done
 
 
 
+
+
+
 ####################
 # enA-koA: 511_358 #
 ####################
@@ -294,7 +299,7 @@ do
   python fetch_dataset_s2s.py
 done
 # tokenize
-for i in $(seq 76 98);
+for i in $(seq 1 100);
 do
   export N_POOL=15
   export DATASET_ID=${i}
@@ -306,31 +311,8 @@ do
 done
 # speaker embedding
 export SE_MODEL="metavoice"
-export SE_MODEL="w2vbert-600m"  # running
-export SE_MODEL="xlsr-2b"  # running
-export SE_MODEL="hubert-xl"
-for i in $(seq 1 90);
-do
-  export DATASET_ID=${i}
-  export DIRECTION="enA-koA"
-  python attach_speaker_embedding_s2s.py
-done
-
-export SE_MODEL="metavoice"
-for i in 65 50;
-do
-  export DATASET_ID=${i}
-  export DIRECTION="enA-koA"
-  python attach_speaker_embedding_s2s.py
-done
-for i in $(seq 90 98);
-do
-  export DATASET_ID=${i}
-  export DIRECTION="enA-koA"
-  python attach_speaker_embedding_s2s.py
-done
-
-# todo
+export SE_MODEL="w2vbert-600m"
+export SE_MODEL="xlsr-2b"
 export SE_MODEL="hubert-xl"
 for i in $(seq 1 100);
 do
@@ -339,20 +321,6 @@ do
   python attach_speaker_embedding_s2s.py
 done
 
-export SE_MODEL="w2vbert-600m"
-export DATASET_ID=34
-export DIRECTION="enA-koA"
-python attach_speaker_embedding_s2s.py
-export DATASET_ID=49
-export DIRECTION="enA-koA"
-python attach_speaker_embedding_s2s.py
-export SE_MODEL="xlsr-2b"
-export DATASET_ID=49
-export DIRECTION="enA-koA"
-python attach_speaker_embedding_s2s.py
-
-
-
 ######################
 # enA-zhA: 1_289_192 #
 ######################
@@ -387,8 +355,8 @@ done
 # speaker embedding
 export SE_MODEL="metavoice"
 export SE_MODEL="w2vbert-600m"
-export SE_MODEL="hubert-xl"  # running
-export SE_MODEL="xlsr-2b"  # running
+export SE_MODEL="hubert-xl"
+export SE_MODEL="xlsr-2b"
 for i in $(seq 1 258);
 do
   export DATASET_ID=${i}
@@ -396,30 +364,6 @@ do
   python attach_speaker_embedding_s2s.py
 done
 
-for i in 162 163 164 165 175 83 84;
-do
-  export SE_MODEL="xlsr-2b"  # running
-  export DIRECTION="enA-zhA"
-  python attach_speaker_embedding_s2s.py
-done
-
-for i in 49 34;
-do
-  export SE_MODEL="w2vbert-600m"
-  export DATASET_ID=${i}
-  export DIRECTION="enA-koA"
-  python attach_speaker_embedding_s2s.py
-done
-
-for i in 34 81 49 50 21;
-do
-  export SE_MODEL="xlsr-2b"
-  export DATASET_ID=${i}
-  export DIRECTION="enA-koA"
-  python attach_speaker_embedding_s2s.py
-done
-
-
 ####################
 # enA-jaA: 718_606 #
 ####################
@@ -451,7 +395,7 @@ done
 export SE_MODEL="metavoice"
 export SE_MODEL="w2vbert-600m"
 export SE_MODEL="hubert-xl"
-export SE_MODEL="xlsr-2b"  # running
+export SE_MODEL="xlsr-2b"
 for i in $(seq 1 144);
 do
   export DATASET_ID=${i}
@@ -529,7 +473,7 @@ done
 export SE_MODEL="metavoice"
 export SE_MODEL="w2vbert-600m"
 export SE_MODEL="hubert-xl"
-export SE_MODEL="xlsr-2b"  # running
+export SE_MODEL="xlsr-2b"
 for i in $(seq 1 91);
 do
   export DATASET_ID=${i}