init

main_s2s.sh

@@ -1,3 +1,42 @@
+export SE_MODEL="xlsr-2b"
+for i in 517 518 519 520;
+do
+  export DATASET_ID=${i}
+  export DIRECTION="enA-esA"
+  python attach_speaker_embedding_s2s.py
+done
+export SE_MODEL="xlsr-2b"
+for i in 507 508 515 516;
+do
+  export DATASET_ID=${i}
+  export DIRECTION="enA-esA"
+  python attach_speaker_embedding_s2s.py
+done
+export SE_MODEL="xlsr-2b"
+for i in 505 506 507;
+do
+  export DATASET_ID=${i}
+  export DIRECTION="enA-esA"
+  python attach_speaker_embedding_s2s.py
+done
+export SE_MODEL="xlsr-2b"
+for i in 500 502 504;
+do
+  export DATASET_ID=${i}
+  export DIRECTION="enA-esA"
+  python attach_speaker_embedding_s2s.py
+done
+
+
+export N_POOL=15
+export DATASET_ID=${i}
+export DIRECTION="enA-jaA"
+export LINE_NO_START=$(((DATASET_ID-1) * 2500))
+export LINE_NO_END=$((DATASET_ID * 2500))
+echo ${LINE_NO_START}
+python tokenize_dataset_s2s.py
+
+
 ######################
 # enA-frA: 3_054_258 #
 ######################
@@ -8,36 +47,6 @@ export LINE_NO_START=0
 export LINE_NO_END=10
 python fetch_dataset_s2s.py
 # main
-for i in $(seq 181 200);
-do
-  export N_POOL=15
-  export DATASET_ID=${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 $(seq 256 280);
-do
-  export N_POOL=15
-  export DATASET_ID=${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 $(seq 281 299);
-do
-  export N_POOL=15
-  export DATASET_ID=${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 1 600);
 do
   export N_POOL=15
@@ -82,48 +91,7 @@ export LINE_NO_START=0
 export LINE_NO_END=10
 python fetch_dataset_s2s.py
 # main
-for i in $(seq 345 420);
-do
-  export N_POOL=15
-  export DATASET_ID=${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
-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
-for i in $(seq 451 500);
-do
-  export N_POOL=15
-  export DATASET_ID=${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 $(seq 500 514);
-do
-  export N_POOL=15
-  export DATASET_ID=${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 515 532);
+for i in $(seq 1 532);
 do
   export N_POOL=15
   export DATASET_ID=${i}
@@ -134,7 +102,7 @@ do
   python fetch_dataset_s2s.py
 done
 # tokenize
-for i in $(seq 51 210);
+for i in $(seq 1 532);
 do
   export N_POOL=15
   export DATASET_ID=${i}
@@ -145,27 +113,17 @@ do
   python tokenize_dataset_s2s.py
 done
 # speaker embedding
-for i in $(seq 251 280);
-do
-  export DATASET_ID=${i}
-  export DIRECTION="enA-esA"
-  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);
+export SE_MODEL="hubert-xl"
+export SE_MODEL="xlsr-2b"
+export SE_MODEL="w2vbert-600m"
+for i in $(seq 1 532);
 do
   export DATASET_ID=${i}
   export DIRECTION="enA-esA"
   python attach_speaker_embedding_s2s.py
 done
 
-
-
-
 ######################
 # enA-deA: 1_965_186 #
 ######################
@@ -187,7 +145,7 @@ do
   python fetch_dataset_s2s.py
 done
 # tokenize
-for i in $(seq 321 353);
+for i in $(seq 1 394);
 do
   export N_POOL=15
   export DATASET_ID=${i}
@@ -197,122 +155,12 @@ do
   echo ${LINE_NO_START}
   python tokenize_dataset_s2s.py
 done
-
 # speaker embedding
-for i in $(seq 28 104);
-do
-  export DATASET_ID=${i}
-  export DIRECTION="deA-enA"
-  python attach_speaker_embedding_s2s.py
-done
-# speaker embedding
-
-
-# running on kotoba_2
-#TODO1
-export SE_MODEL="xlsr-2b"
-for i in 1 2 4 5 6 10 41 50 55 340 377 379;
-do
-  export DATASET_ID=${i}
-  export DIRECTION="enA-frA"
-  python attach_speaker_embedding_s2s.py
-done
-for i in $(seq 73 299);
-do
-  export DATASET_ID=${i}
-  export DIRECTION="enA-frA"
-  python attach_speaker_embedding_s2s.py
-done
-for i in $(seq 381 600);
-do
-  export DATASET_ID=${i}
-  export DIRECTION="enA-frA"
-  python attach_speaker_embedding_s2s.py
-done
-#TODO2
-export SE_MODEL="hubert-xl"
-for i in 318 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361;
-do
-  export DATASET_ID=${i}
-  export DIRECTION="enA-esA"
-  python attach_speaker_embedding_s2s.py
-done
-#TODO3
-export SE_MODEL="w2vbert-600m"
-for i in $(seq 40 600);
-do
-  export DATASET_ID=${i}
-  export DIRECTION="enA-frA"
-  python attach_speaker_embedding_s2s.py
-done
-#TODO4
-export SE_MODEL="w2vbert-600m"
-for i in $(seq 467 470);
-do
-  export DATASET_ID=${i}
-  export DIRECTION="enA-esA"
-  python attach_speaker_embedding_s2s.py
-done
-export SE_MODEL="xlsr-2b"
-for i in $(seq 453 470);
-do
-  export DATASET_ID=${i}
-  export DIRECTION="enA-esA"
-  python attach_speaker_embedding_s2s.py
-done
 export SE_MODEL="metavoice"
-for i in $(seq 453 459);
-do
-  export DATASET_ID=${i}
-  export DIRECTION="enA-esA"
-  python attach_speaker_embedding_s2s.py
-done
-
-
-# running on kotoba_1
-#TODO1
-export SE_MODEL="hubert-xl"
-for i in $(seq 1 394);
-do
-  export DATASET_ID=${i}
-  export DIRECTION="deA-enA"
-  python attach_speaker_embedding_s2s.py
-done
-#TODO2
 export SE_MODEL="hubert-xl"
-for i in $(seq 38 600);
-do
-  export DATASET_ID=${i}
-  export DIRECTION="enA-frA"
-  python attach_speaker_embedding_s2s.py
-done
-#TODO3
-export SE_MODEL="metavoice"
-for i in $(seq 50 299);
-do
-  export DATASET_ID=${i}
-  export DIRECTION="enA-frA"
-  python attach_speaker_embedding_s2s.py
-done
-for i in $(seq 331 398);
-do
-  export DATASET_ID=${i}
-  export DIRECTION="enA-frA"
-  python attach_speaker_embedding_s2s.py
-done
-for i in $(seq 426 600);
-do
-  export DATASET_ID=${i}
-  export DIRECTION="enA-frA"
-  python attach_speaker_embedding_s2s.py
-done
-#TODO4
 export SE_MODEL="xlsr-2b"
-export DATASET_ID=213
-export DIRECTION="deA-enA"
-python attach_speaker_embedding_s2s.py
-export SE_MODEL="hubert-xl"
-for i in 117 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284;
+export SE_MODEL="w2vbert-600m"
+for i in $(seq 1 394);
 do
   export DATASET_ID=${i}
   export DIRECTION="deA-enA"
@@ -320,16 +168,6 @@ do
 done
 
 
-
-
-
-
-
-
-
-
-
-
 ####################
 # enA-koA: 511_358 #
 ####################
@@ -363,9 +201,9 @@ do
 done
 # speaker embedding
 export SE_MODEL="metavoice"
-export SE_MODEL="w2vbert-600m"
 export SE_MODEL="hubert-xl"
 export SE_MODEL="xlsr-2b"
+export SE_MODEL="w2vbert-600m"
 for i in $(seq 1 100);
 do
   export DATASET_ID=${i}
@@ -406,9 +244,9 @@ do
 done
 # speaker embedding
 export SE_MODEL="metavoice"
-export SE_MODEL="w2vbert-600m"
 export SE_MODEL="hubert-xl"
 export SE_MODEL="xlsr-2b"
+export SE_MODEL="w2vbert-600m"
 for i in $(seq 1 258);
 do
   export DATASET_ID=${i}
@@ -445,9 +283,9 @@ do
 done
 # speaker embedding
 export SE_MODEL="metavoice"
-export SE_MODEL="w2vbert-600m"
 export SE_MODEL="hubert-xl"
 export SE_MODEL="xlsr-2b"
+export SE_MODEL="w2vbert-600m"
 for i in $(seq 1 144);
 do
   export DATASET_ID=${i}
@@ -484,9 +322,9 @@ do
 done
 # speaker embedding
 export SE_MODEL="metavoice"
-export SE_MODEL="w2vbert-600m"
 export SE_MODEL="hubert-xl"
 export SE_MODEL="xlsr-2b"
+export SE_MODEL="w2vbert-600m"
 for i in $(seq 1 149);
 do
   export DATASET_ID=${i}
@@ -523,9 +361,9 @@ do
 done
 # speaker embedding
 export SE_MODEL="metavoice"
-export SE_MODEL="w2vbert-600m"
 export SE_MODEL="hubert-xl"
 export SE_MODEL="xlsr-2b"
+export SE_MODEL="w2vbert-600m"
 for i in $(seq 1 91);
 do
   export DATASET_ID=${i}

tokenize_dataset_s2s.py

@@ -60,8 +60,7 @@ dataset = dataset.map(
     num_proc=num_proc,
     desc="tokenize dataset"
 )
-DatasetDict({"train": dataset}).push_to_hub(f"{hf_org}/{hf_dataset}.tokenized", config_name=f"subset_{dataset_id}")
-# DatasetDict({"train": dataset}).push_to_hub(f"{hf_org}/{hf_dataset}.tokenized.encodec", config_name=f"subset_{dataset_id}")
+DatasetDict({"train": dataset}).push_to_hub(f"{hf_org}/{hf_dataset}.tokenized.encodec", config_name=f"subset_{dataset_id}")
 cache_dir = f"{expanduser('~')}/.cache/huggingface/datasets/{hf_org}___{hf_dataset}/subset_{dataset_id}"
 if os.path.exists(cache_dir):
     shutil.rmtree(cache_dir)