init

attach_speaker_embedding_s2s.py

@@ -14,6 +14,8 @@ hf_dataset = f"seamless-align-{direction}"
 dataset = load_dataset(f"{hf_org}/{hf_dataset}", f"subset_{dataset_id}", split="train")
 audio_loader = Audio()
 se_model = os.getenv("SE_MODEL", "metavoice")
+max_seq_length = 10000000
+
 if se_model == "metavoice":
     from speaker_embedding_metavoice import MetaVoiceSE
     speaker_embedder = MetaVoiceSE()
@@ -36,7 +38,11 @@ else:
 def error_file(example):
     for side in sides:
         try:
-            audio_loader.decode_example(example[f"{side}.audio"])
+            wav = audio_loader.decode_example(example[f"{side}.audio"])
+            if len(wav["array"]) == 0 or len(wav["array"]) > max_seq_length:
+                return False
+        except ValueError:
+            return False
         except LibsndfileError:
             return False
     return True

main_s2s.sh

@@ -1,69 +1,44 @@
-####################
-# enA-koA: 511_358 #
-####################
+######################
+# enA-frA: 3_054_258 #
+######################
 # test
 export DATASET_ID=test
-export DIRECTION="enA-koA"
+export DIRECTION="enA-frA"
 export LINE_NO_START=0
 export LINE_NO_END=10
 python fetch_dataset_s2s.py
 # main
-for i in $(seq 87 100);
+for i in $(seq 115 200);
 do
   export N_POOL=15
   export DATASET_ID=${i}
-  export DIRECTION="enA-koA"
+  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
-# tokenize
-for i in $(seq 37 50);
+for i in $(seq 443 500);
 do
   export N_POOL=15
   export DATASET_ID=${i}
-  export DIRECTION="enA-koA"
+  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 fetch_dataset_s2s.py
 done
-for i in $(seq 65 75);
+for i in $(seq 519 550);
 do
   export N_POOL=15
   export DATASET_ID=${i}
-  export DIRECTION="enA-koA"
+  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
-done
-# speaker embedding
-for i in $(seq 1 20);
-do
-  export DATASET_ID=${i}
-  export DIRECTION="enA-koA"
-  python attach_speaker_embedding_s2s.py
-done
-for i in $(seq 51 64);
-do
-  export DATASET_ID=${i}
-  export DIRECTION="enA-koA"
-  python attach_speaker_embedding_s2s.py
+  python fetch_dataset_s2s.py
 done
-
-######################
-# enA-frA: 3_054_258 #
-######################
-# test
-export DATASET_ID=test
-export DIRECTION="enA-frA"
-export LINE_NO_START=0
-export LINE_NO_END=10
-python fetch_dataset_s2s.py
-# main
-for i in $(seq 115 200);
+for i in $(seq 551 600);
 do
   export N_POOL=15
   export DATASET_ID=${i}
@@ -73,7 +48,8 @@ do
   echo ${LINE_NO_START}
   python fetch_dataset_s2s.py
 done
-for i in $(seq 209 300);
+# tokenize
+for i in $(seq 61 145);
 do
   export N_POOL=15
   export DATASET_ID=${i}
@@ -81,9 +57,9 @@ do
   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
-for i in $(seq 440 500);
+for i in $(seq 200 299);
 do
   export N_POOL=15
   export DATASET_ID=${i}
@@ -91,9 +67,9 @@ do
   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
-for i in $(seq 515 600);
+for i in $(seq 426 450);
 do
   export N_POOL=15
   export DATASET_ID=${i}
@@ -101,10 +77,9 @@ do
   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 61 105);
+for i in $(seq 500 530);
 do
   export N_POOL=15
   export DATASET_ID=${i}
@@ -115,14 +90,14 @@ do
   python tokenize_dataset_s2s.py
 done
 # speaker embedding
-for i in $(seq 61 105);
+for i in $(seq );
 do
   export DATASET_ID=${i}
   export DIRECTION="enA-frA"
   python attach_speaker_embedding_s2s.py
 done
 # tokenize
-for i in $(seq 361 425);
+for i in $(seq 106 425);
 do
   export N_POOL=15
   export DATASET_ID=${i}
@@ -161,7 +136,6 @@ do
   echo ${LINE_NO_START}
   python fetch_dataset_s2s.py
 done
-
 for i in $(seq 201 394);
 do
   export N_POOL=15
@@ -172,8 +146,7 @@ do
   echo ${LINE_NO_START}
   python fetch_dataset_s2s.py
 done
-# tokenize
-for i in $(seq 51 104);
+for i in $(seq 73 91);
 do
   export N_POOL=15
   export DATASET_ID=${i}
@@ -181,9 +154,20 @@ do
   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 fetch_dataset_s2s.py
 done
-for i in $(seq 370 394);
+# 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 $(seq 105 180);
 do
   export N_POOL=15
   export DATASET_ID=${i}
@@ -217,7 +201,7 @@ export LINE_NO_START=0
 export LINE_NO_END=10
 python fetch_dataset_s2s.py
 # main
-for i in $(seq 98 125);
+for i in $(seq 194 220);
 do
   export N_POOL=15
   export DATASET_ID=${i}
@@ -227,7 +211,7 @@ do
   echo ${LINE_NO_START}
   python fetch_dataset_s2s.py
 done
-for i in $(seq 165 250);
+for i in $(seq 220 250);
 do
   export N_POOL=15
   export DATASET_ID=${i}
@@ -269,25 +253,18 @@ do
   python tokenize_dataset_s2s.py
 done
 # speaker embedding
-for i in $(seq 1 50);
-do
-  export DATASET_ID=${i}
-  export DIRECTION="enA-esA"
-  python attach_speaker_embedding_s2s.py
-done
-# tokenize
 for i in $(seq 251 280);
 do
-  export N_POOL=15
   export DATASET_ID=${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 tokenize_dataset_s2s.py
+  python attach_speaker_embedding_s2s.py
 done
-# speaker embedding
-for i in $(seq 251 280);
+
+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="enA-esA"
@@ -296,100 +273,93 @@ done
 
 
 
-
 ####################
-# enA-hiA: 454_942 #
+# enA-koA: 511_358 #
 ####################
 # test
 export DATASET_ID=test
-export DIRECTION="enA-hiA"
+export DIRECTION="enA-koA"
 export LINE_NO_START=0
 export LINE_NO_END=10
 python fetch_dataset_s2s.py
 # main
-for i in $(seq 1 91);
+for i in $(seq 1 100);
 do
   export N_POOL=15
   export DATASET_ID=${i}
-  export DIRECTION="enA-hiA"
+  export DIRECTION="enA-koA"
   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 1 91);
+for i in $(seq 76 98);
 do
+  export N_POOL=15
   export DATASET_ID=${i}
-  export DIRECTION="enA-hiA"
+  export DIRECTION="enA-koA"
+  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
 export SE_MODEL="metavoice"
-export SE_MODEL="w2vbert-600m"
+export SE_MODEL="w2vbert-600m"  # running
+export SE_MODEL="xlsr-2b"  # running
 export SE_MODEL="hubert-xl"
-export SE_MODEL="xlsr-2b"
-for i in $(seq 17 91);
+for i in $(seq 1 90);
 do
   export DATASET_ID=${i}
-  export DIRECTION="enA-hiA"
+  export DIRECTION="enA-koA"
   python attach_speaker_embedding_s2s.py
 done
 
-export SE_MODEL="hubert-xl"
-export DATASET_ID=84
-export DIRECTION="enA-hiA"
-python attach_speaker_embedding_s2s.py
-####################
-# enA-viA: 740_598 #
-####################
-# test
-export DATASET_ID=test
-export DIRECTION="enA-viA"
-export LINE_NO_START=0
-export LINE_NO_END=10
-python fetch_dataset_s2s.py
-# main
-for i in $(seq 1 149);
+export SE_MODEL="metavoice"
+for i in 65 50;
 do
-  export N_POOL=15
   export DATASET_ID=${i}
-  export DIRECTION="enA-viA"
-  export LINE_NO_START=$(((DATASET_ID-1) * 2500))
-  export LINE_NO_END=$((DATASET_ID * 2500))
-  echo ${LINE_NO_START}
-  python fetch_dataset_s2s.py
+  export DIRECTION="enA-koA"
+  python attach_speaker_embedding_s2s.py
 done
-# tokenize
-for i in $(seq 1 149);
+for i in $(seq 90 98);
 do
   export DATASET_ID=${i}
-  export DIRECTION="enA-viA"
-  python tokenize_dataset_s2s.py
+  export DIRECTION="enA-koA"
+  python attach_speaker_embedding_s2s.py
 done
-# speaker embedding
-export SE_MODEL="metavoice"
-export SE_MODEL="xlsr-2b"
-export SE_MODEL="w2vbert-600m"
+
+# todo
 export SE_MODEL="hubert-xl"
-for i in $(seq 1 149);
+for i in $(seq 1 98);
 do
   export DATASET_ID=${i}
-  export DIRECTION="enA-viA"
+  export DIRECTION="enA-koA"
   python attach_speaker_embedding_s2s.py
 done
-
 export SE_MODEL="w2vbert-600m"
-export DATASET_ID=116
-python attach_speaker_embedding_s2s.py
-
-export SE_MODEL="hubert-xl"
-for i in 107 115 126;
+for i in $(seq 90 98);
 do
   export DATASET_ID=${i}
-  export DIRECTION="enA-viA"
+  export DIRECTION="enA-koA"
+  python attach_speaker_embedding_s2s.py
+done
+for i in 34 49 50 87 88;
+do
+  export DATASET_ID=${i}
+  export DIRECTION="enA-koA"
   python attach_speaker_embedding_s2s.py
 done
+export SE_MODEL="xlsr-2b"
+for i in $(seq 90 98);
+do
+  export DATASET_ID=${i}
+  export DIRECTION="enA-koA"
+  python attach_speaker_embedding_s2s.py
+done
+
+
 
 ######################
 # enA-zhA: 1_289_192 #
@@ -424,9 +394,9 @@ do
 done
 # speaker embedding
 export SE_MODEL="metavoice"
-#export SE_MODEL="xlsr-2b"
 export SE_MODEL="w2vbert-600m"
-export SE_MODEL="hubert-xl"
+export SE_MODEL="hubert-xl"  # running
+export SE_MODEL="xlsr-2b"  # running
 for i in $(seq 1 258);
 do
   export DATASET_ID=${i}
@@ -435,6 +405,28 @@ do
 done
 
 
+export SE_MODEL="hubert-xl"  # running
+for i in 165 167;
+do
+  export DATASET_ID=${i}
+  export DIRECTION="enA-zhA"
+  python attach_speaker_embedding_s2s.py
+done
+
+
+export SE_MODEL="w2vbert-600m"  # running
+for i in 142 165 167 207 253;
+do
+  export DATASET_ID=${i}
+  export DIRECTION="enA-zhA"
+  python attach_speaker_embedding_s2s.py
+done
+
+
+
+
+
+
 ####################
 # enA-jaA: 718_606 #
 ####################
@@ -466,7 +458,7 @@ done
 export SE_MODEL="metavoice"
 export SE_MODEL="w2vbert-600m"
 export SE_MODEL="hubert-xl"
-export SE_MODEL="xlsr-2b"
+export SE_MODEL="xlsr-2b"  # running
 for i in $(seq 1 144);
 do
   export DATASET_ID=${i}
@@ -474,19 +466,80 @@ do
   python attach_speaker_embedding_s2s.py
 done
 
+####################
+# enA-viA: 740_598 #
+####################
+# test
+export DATASET_ID=test
+export DIRECTION="enA-viA"
+export LINE_NO_START=0
+export LINE_NO_END=10
+python fetch_dataset_s2s.py
+# main
+for i in $(seq 1 149);
+do
+  export N_POOL=15
+  export DATASET_ID=${i}
+  export DIRECTION="enA-viA"
+  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 $(seq 1 149);
+do
+  export DATASET_ID=${i}
+  export DIRECTION="enA-viA"
+  python tokenize_dataset_s2s.py
+done
+# speaker embedding
+export SE_MODEL="metavoice"
 export SE_MODEL="w2vbert-600m"
-for i in 9 107 126 142;
+export SE_MODEL="hubert-xl"
+export SE_MODEL="xlsr-2b"
+for i in $(seq 1 149);
 do
   export DATASET_ID=${i}
-  export DIRECTION="enA-jaA"
+  export DIRECTION="enA-viA"
   python attach_speaker_embedding_s2s.py
 done
 
+####################
+# enA-hiA: 454_942 #
+####################
+# test
+export DATASET_ID=test
+export DIRECTION="enA-hiA"
+export LINE_NO_START=0
+export LINE_NO_END=10
+python fetch_dataset_s2s.py
+# main
+for i in $(seq 1 91);
+do
+  export N_POOL=15
+  export DATASET_ID=${i}
+  export DIRECTION="enA-hiA"
+  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 $(seq 1 91);
+do
+  export DATASET_ID=${i}
+  export DIRECTION="enA-hiA"
+  python tokenize_dataset_s2s.py
+done
+# speaker embedding
+export SE_MODEL="metavoice"
+export SE_MODEL="w2vbert-600m"
 export SE_MODEL="hubert-xl"
-for i in 103 114 115 121 129;
+export SE_MODEL="xlsr-2b"  # running
+for i in $(seq 1 91);
 do
   export DATASET_ID=${i}
-  export DIRECTION="enA-jaA"
+  export DIRECTION="enA-hiA"
   python attach_speaker_embedding_s2s.py
-done
-
+done