minor fixes

1. changed the naming scheme from `speed-perturb` to `perturb-speed` to align with the librispeech recipe

>> 00256a7669/egs/librispeech/ASR/local/compute_fbank_librispeech.py (L65)

2. changed arg type for `perturb-speed` to str2bool

egs/aidatatang_200zh/ASR/local/compute_fbank_aidatatang_200zh.py

@@ -32,7 +32,7 @@ import torch
 from lhotse import CutSet, Fbank, FbankConfig, LilcomChunkyWriter
 from lhotse.recipes.utils import read_manifests_if_cached
 
-from icefall.utils import get_executor
+from icefall.utils import get_executor, str2bool
 
 # Torch's multithreaded behavior needs to be disabled or
 # it wastes a lot of CPU and slow things down.
@@ -42,7 +42,7 @@ torch.set_num_threads(1)
 torch.set_num_interop_threads(1)
 
 
-def compute_fbank_aidatatang_200zh(num_mel_bins: int = 80, speed_perturb: bool = False):
+def compute_fbank_aidatatang_200zh(num_mel_bins: int = 80, perturb_speed: bool = False):
     src_dir = Path("data/manifests/aidatatang_200zh")
     output_dir = Path("data/fbank")
     num_jobs = min(15, os.cpu_count())
@@ -85,7 +85,8 @@ def compute_fbank_aidatatang_200zh(num_mel_bins: int = 80, speed_perturb: bool =
                 recordings=m["recordings"],
                 supervisions=m["supervisions"],
             )
-            if "train" in partition and speed_perturb:
+            if "train" in partition and perturb_speed:
+                logging.info(f"Doing speed perturb")
                 cut_set = (
                     cut_set + cut_set.perturb_speed(0.9) + cut_set.perturb_speed(1.1)
                 )
@@ -110,8 +111,8 @@ def get_args():
         help="""The number of mel bins for Fbank""",
     )
     parser.add_argument(
-        "--speed-perturb",
-        type=bool,
+        "--perturb-speed",
+        type=str2bool,
         default=False,
         help="Enable 0.9 and 1.1 speed perturbation for data augmentation. Default: False.",
     )
@@ -125,5 +126,5 @@ if __name__ == "__main__":
 
     args = get_args()
     compute_fbank_aidatatang_200zh(
-        num_mel_bins=args.num_mel_bins, speed_perturb=args.speed_perturb
+        num_mel_bins=args.num_mel_bins, perturb_speed=args.perturb_speed
     )

egs/aidatatang_200zh/ASR/prepare.sh

@@ -77,7 +77,7 @@ if [ $stage -le 4 ] && [ $stop_stage -ge 4 ]; then
   log "Stage 4: Compute fbank for aidatatang_200zh"
   if [ ! -f data/fbank/.aidatatang_200zh.done ]; then
     mkdir -p data/fbank
-    ./local/compute_fbank_aidatatang_200zh.py --speed-perturb True
+    ./local/compute_fbank_aidatatang_200zh.py --perturb-speed True
     touch data/fbank/.aidatatang_200zh.done
   fi
 fi

egs/aishell/ASR/local/compute_fbank_aidatatang_200zh.py

@@ -32,7 +32,7 @@ import torch
 from lhotse import CutSet, Fbank, FbankConfig, LilcomChunkyWriter
 from lhotse.recipes.utils import read_manifests_if_cached
 
-from icefall.utils import get_executor
+from icefall.utils import get_executor, str2bool
 
 # Torch's multithreaded behavior needs to be disabled or
 # it wastes a lot of CPU and slow things down.
@@ -42,7 +42,7 @@ torch.set_num_threads(1)
 torch.set_num_interop_threads(1)
 
 
-def compute_fbank_aidatatang_200zh(num_mel_bins: int = 80, speed_perturb: bool = False):
+def compute_fbank_aidatatang_200zh(num_mel_bins: int = 80, perturb_speed: bool = False):
     src_dir = Path("data/manifests")
     output_dir = Path("data/fbank")
     num_jobs = min(15, os.cpu_count())
@@ -85,7 +85,8 @@ def compute_fbank_aidatatang_200zh(num_mel_bins: int = 80, speed_perturb: bool =
                 recordings=m["recordings"],
                 supervisions=m["supervisions"],
             )
-            if "train" in partition and speed_perturb:
+            if "train" in partition and perturb_speed:
+                logging.info(f"Doing speed perturb")
                 cut_set = (
                     cut_set + cut_set.perturb_speed(0.9) + cut_set.perturb_speed(1.1)
                 )
@@ -110,8 +111,8 @@ def get_args():
         help="""The number of mel bins for Fbank""",
     )
     parser.add_argument(
-        "--speed-perturb",
-        type=bool,
+        "--perturb-speed",
+        type=str2bool,
         default=False,
         help="Enable 0.9 and 1.1 speed perturbation for data augmentation. Default: False.",
     )
@@ -125,5 +126,5 @@ if __name__ == "__main__":
 
     args = get_args()
     compute_fbank_aidatatang_200zh(
-        num_mel_bins=args.num_mel_bins, speed_perturb=args.speed_perturb
+        num_mel_bins=args.num_mel_bins, perturb_speed=args.perturb_speed
     )

egs/aishell/ASR/local/compute_fbank_aishell.py

@@ -32,7 +32,7 @@ import torch
 from lhotse import CutSet, Fbank, FbankConfig, LilcomChunkyWriter
 from lhotse.recipes.utils import read_manifests_if_cached
 
-from icefall.utils import get_executor
+from icefall.utils import get_executor, str2bool
 
 # Torch's multithreaded behavior needs to be disabled or
 # it wastes a lot of CPU and slow things down.
@@ -42,7 +42,7 @@ torch.set_num_threads(1)
 torch.set_num_interop_threads(1)
 
 
-def compute_fbank_aishell(num_mel_bins: int = 80, speed_perturb: bool = False):
+def compute_fbank_aishell(num_mel_bins: int = 80, perturb_speed: bool = False):
     src_dir = Path("data/manifests")
     output_dir = Path("data/fbank")
     num_jobs = min(15, os.cpu_count())
@@ -81,7 +81,8 @@ def compute_fbank_aishell(num_mel_bins: int = 80, speed_perturb: bool = False):
                 recordings=m["recordings"],
                 supervisions=m["supervisions"],
             )
-            if "train" in partition and speed_perturb:
+            if "train" in partition and perturb_speed:
+                logging.info(f"Doing speed perturb")
                 cut_set = (
                     cut_set + cut_set.perturb_speed(0.9) + cut_set.perturb_speed(1.1)
                 )
@@ -105,8 +106,8 @@ def get_args():
         help="""The number of mel bins for Fbank""",
     )
     parser.add_argument(
-        "--speed-perturb",
-        type=bool,
+        "--perturb-speed",
+        type=str2bool,
         default=False,
         help="Enable 0.9 and 1.1 speed perturbation for data augmentation. Default: False.",
     )
@@ -120,5 +121,5 @@ if __name__ == "__main__":
 
     args = get_args()
     compute_fbank_aishell(
-        num_mel_bins=args.num_mel_bins, speed_perturb=args.speed_perturb
+        num_mel_bins=args.num_mel_bins, perturb_speed=args.perturb_speed
     )

egs/aishell/ASR/prepare.sh

@@ -114,7 +114,7 @@ if [ $stage -le 3 ] && [ $stop_stage -ge 3 ]; then
   log "Stage 3: Compute fbank for aishell"
   if [ ! -f data/fbank/.aishell.done ]; then
     mkdir -p data/fbank
-    ./local/compute_fbank_aishell.py --speed-perturb True
+    ./local/compute_fbank_aishell.py --perturb-speed True
     touch data/fbank/.aishell.done
   fi
 fi

egs/aishell/ASR/prepare_aidatatang_200zh.sh

@@ -53,7 +53,7 @@ if [ $stage -le 2 ] && [ $stop_stage -ge 2 ]; then
   log "Stage 2: Process aidatatang_200zh"
   if [ ! -f data/fbank/.aidatatang_200zh_fbank.done ]; then
     mkdir -p data/fbank
-    ./local/compute_fbank_aidatatang_200zh.py --speed-perturb True
+    ./local/compute_fbank_aidatatang_200zh.py --perturb-speed True
     touch data/fbank/.aidatatang_200zh_fbank.done
   fi
 fi

egs/aishell2/ASR/local/compute_fbank_aishell2.py

@@ -32,7 +32,7 @@ import torch
 from lhotse import CutSet, Fbank, FbankConfig, LilcomChunkyWriter
 from lhotse.recipes.utils import read_manifests_if_cached
 
-from icefall.utils import get_executor
+from icefall.utils import get_executor, str2bool
 
 # Torch's multithreaded behavior needs to be disabled or
 # it wastes a lot of CPU and slow things down.
@@ -42,7 +42,7 @@ torch.set_num_threads(1)
 torch.set_num_interop_threads(1)
 
 
-def compute_fbank_aishell2(num_mel_bins: int = 80, speed_perturb: bool = False):
+def compute_fbank_aishell2(num_mel_bins: int = 80, perturb_speed: bool = False):
     src_dir = Path("data/manifests")
     output_dir = Path("data/fbank")
     num_jobs = min(15, os.cpu_count())
@@ -81,7 +81,8 @@ def compute_fbank_aishell2(num_mel_bins: int = 80, speed_perturb: bool = False):
                 recordings=m["recordings"],
                 supervisions=m["supervisions"],
             )
-            if "train" in partition and speed_perturb:
+            if "train" in partition and perturb_speed:
+                logging.info(f"Doing speed perturb")
                 cut_set = (
                     cut_set + cut_set.perturb_speed(0.9) + cut_set.perturb_speed(1.1)
                 )
@@ -105,8 +106,8 @@ def get_args():
         help="""The number of mel bins for Fbank""",
     )
     parser.add_argument(
-        "--speed-perturb",
-        type=bool,
+        "--perturb-speed",
+        type=str2bool,
         default=False,
         help="Enable 0.9 and 1.1 speed perturbation for data augmentation. Default: False.",
     )
@@ -121,5 +122,5 @@ if __name__ == "__main__":
 
     args = get_args()
     compute_fbank_aishell2(
-        num_mel_bins=args.num_mel_bins, speed_perturb=args.speed_perturb
+        num_mel_bins=args.num_mel_bins, perturb_speed=args.perturb_speed
     )

egs/aishell2/ASR/prepare.sh

@@ -101,7 +101,7 @@ if [ $stage -le 3 ] && [ $stop_stage -ge 3 ]; then
   log "Stage 3: Compute fbank for aishell2"
   if [ ! -f data/fbank/.aishell2.done ]; then
     mkdir -p data/fbank
-    ./local/compute_fbank_aishell2.py --speed-perturb True
+    ./local/compute_fbank_aishell2.py --perturb-speed True
     touch data/fbank/.aishell2.done
   fi
 fi

egs/aishell4/ASR/local/compute_fbank_aishell4.py

@@ -32,7 +32,7 @@ import torch
 from lhotse import ChunkedLilcomHdf5Writer, CutSet, Fbank, FbankConfig
 from lhotse.recipes.utils import read_manifests_if_cached
 
-from icefall.utils import get_executor
+from icefall.utils import get_executor, str2bool
 
 # Torch's multithreaded behavior needs to be disabled or
 # it wastes a lot of CPU and slow things down.
@@ -42,7 +42,7 @@ torch.set_num_threads(1)
 torch.set_num_interop_threads(1)
 
 
-def compute_fbank_aishell4(num_mel_bins: int = 80, speed_perturb: bool = False):
+def compute_fbank_aishell4(num_mel_bins: int = 80, perturb_speed: bool = False):
     src_dir = Path("data/manifests/aishell4")
     output_dir = Path("data/fbank")
     num_jobs = min(15, os.cpu_count())
@@ -83,7 +83,8 @@ def compute_fbank_aishell4(num_mel_bins: int = 80, speed_perturb: bool = False):
                 recordings=m["recordings"],
                 supervisions=m["supervisions"],
             )
-            if "train" in partition and speed_perturb:
+            if "train" in partition and perturb_speed:
+                logging.info(f"Doing speed perturb")
                 cut_set = (
                     cut_set + cut_set.perturb_speed(0.9) + cut_set.perturb_speed(1.1)
                 )
@@ -115,8 +116,8 @@ def get_args():
         help="""The number of mel bins for Fbank""",
     )
     parser.add_argument(
-        "--speed-perturb",
-        type=bool,
+        "--perturb-speed",
+        type=str2bool,
         default=False,
         help="Enable 0.9 and 1.1 speed perturbation for data augmentation. Default: False.",
     )
@@ -131,5 +132,5 @@ if __name__ == "__main__":
 
     args = get_args()
     compute_fbank_aishell4(
-        num_mel_bins=args.num_mel_bins, speed_perturb=args.speed_perturb
+        num_mel_bins=args.num_mel_bins, perturb_speed=args.perturb_speed
     )

egs/aishell4/ASR/prepare.sh

@@ -107,7 +107,7 @@ if [ $stage -le 5 ] && [ $stop_stage -ge 5 ]; then
   log "Stage 5: Compute fbank for aishell4"
   if [ ! -f data/fbank/.aishell4.done ]; then
     mkdir -p data/fbank
-    ./local/compute_fbank_aishell4.py --speed-perturb True
+    ./local/compute_fbank_aishell4.py --perturb-speed True
     touch data/fbank/.aishell4.done
   fi
 fi

egs/alimeeting/ASR/local/compute_fbank_alimeeting.py

@@ -32,7 +32,7 @@ import torch
 from lhotse import CutSet, Fbank, FbankConfig, LilcomChunkyWriter
 from lhotse.recipes.utils import read_manifests_if_cached
 
-from icefall.utils import get_executor
+from icefall.utils import get_executor, str2bool
 
 # Torch's multithreaded behavior needs to be disabled or
 # it wastes a lot of CPU and slow things down.
@@ -42,7 +42,7 @@ torch.set_num_threads(1)
 torch.set_num_interop_threads(1)
 
 
-def compute_fbank_alimeeting(num_mel_bins: int = 80, speed_perturb: bool = False):
+def compute_fbank_alimeeting(num_mel_bins: int = 80, perturb_speed: bool = False):
     src_dir = Path("data/manifests/alimeeting")
     output_dir = Path("data/fbank")
     num_jobs = min(15, os.cpu_count())
@@ -82,7 +82,8 @@ def compute_fbank_alimeeting(num_mel_bins: int = 80, speed_perturb: bool = False
                 recordings=m["recordings"],
                 supervisions=m["supervisions"],
             )
-            if "train" in partition and speed_perturb:
+            if "train" in partition and perturb_speed:
+                logging.info(f"Doing speed perturb")
                 cut_set = (
                     cut_set + cut_set.perturb_speed(0.9) + cut_set.perturb_speed(1.1)
                 )
@@ -115,8 +116,8 @@ def get_args():
         help="""The number of mel bins for Fbank""",
     )
     parser.add_argument(
-        "--speed-perturb",
-        type=bool,
+        "--perturb-speed",
+        type=str2bool,
         default=False,
         help="Enable 0.9 and 1.1 speed perturbation for data augmentation. Default: False.",
     )
@@ -131,5 +132,5 @@ if __name__ == "__main__":
 
     args = get_args()
     compute_fbank_alimeeting(
-        num_mel_bins=args.num_mel_bins, speed_perturb=args.speed_perturb
+        num_mel_bins=args.num_mel_bins, perturb_speed=args.perturb_speed
     )

egs/alimeeting/ASR/prepare.sh

@@ -97,7 +97,7 @@ if [ $stage -le 5 ] && [ $stop_stage -ge 5 ]; then
   log "Stage 5: Compute fbank for alimeeting"
   if [ ! -f data/fbank/.alimeeting.done ]; then
     mkdir -p data/fbank
-    ./local/compute_fbank_alimeeting.py --speed-perturb True
+    ./local/compute_fbank_alimeeting.py --perturb-speed True
     touch data/fbank/.alimeeting.done
   fi
 fi

egs/alimeeting/ASR_v2/local/compute_fbank_alimeeting.py

@@ -40,6 +40,8 @@ from lhotse.features.kaldifeat import (
 )
 from lhotse.recipes.utils import read_manifests_if_cached
 
+from icefall.utils import str2bool
+
 # Torch's multithreaded behavior needs to be disabled or
 # it wastes a lot of CPU and slow things down.
 # Do this outside of main() in case it needs to take effect
@@ -49,7 +51,7 @@ torch.set_num_interop_threads(1)
 torch.multiprocessing.set_sharing_strategy("file_system")
 
 
-def compute_fbank_ami(speed_perturb: bool = False):
+def compute_fbank_ami(perturb_speed: bool = False):
     src_dir = Path("data/manifests")
     output_dir = Path("data/fbank")
 
@@ -89,6 +91,7 @@ def compute_fbank_ami(speed_perturb: bool = False):
         cuts: CutSet, storage_path: Path, manifest_path: Path, speed_perturb: bool
     ) -> None:
         if speed_perturb:
+            logging.info(f"Doing speed perturb")
             cuts = cuts + cuts.perturb_speed(0.9) + cuts.perturb_speed(1.1)
         _ = cuts.compute_and_store_features_batch(
             extractor=extractor,
@@ -113,7 +116,7 @@ def compute_fbank_ami(speed_perturb: bool = False):
         cuts_ihm,
         output_dir / "feats_train_ihm",
         src_dir / "cuts_train_ihm.jsonl.gz",
-        speed_perturb,
+        perturb_speed,
     )
 
     logging.info("Processing train split IHM + reverberated IHM")
@@ -122,7 +125,7 @@ def compute_fbank_ami(speed_perturb: bool = False):
         cuts_ihm_rvb,
         output_dir / "feats_train_ihm_rvb",
         src_dir / "cuts_train_ihm_rvb.jsonl.gz",
-        speed_perturb,
+        perturb_speed,
     )
 
     logging.info("Processing train split SDM")
@@ -135,7 +138,7 @@ def compute_fbank_ami(speed_perturb: bool = False):
         cuts_sdm,
         output_dir / "feats_train_sdm",
         src_dir / "cuts_train_sdm.jsonl.gz",
-        speed_perturb,
+        perturb_speed,
     )
 
     logging.info("Processing train split GSS")
@@ -148,7 +151,7 @@ def compute_fbank_ami(speed_perturb: bool = False):
         cuts_gss,
         output_dir / "feats_train_gss",
         src_dir / "cuts_train_gss.jsonl.gz",
-        speed_perturb,
+        perturb_speed,
     )
 
     logging.info("Preparing test cuts: IHM, SDM, GSS (optional)")
@@ -197,8 +200,8 @@ def compute_fbank_ami(speed_perturb: bool = False):
 def get_args():
     parser = argparse.ArgumentParser()
     parser.add_argument(
-        "--speed-perturb",
-        type=bool,
+        "--perturb-speed",
+        type=str2bool,
         default=False,
         help="Enable 0.9 and 1.1 speed perturbation for data augmentation. Default: False.",
     )
@@ -211,4 +214,4 @@ if __name__ == "__main__":
 
     args = get_args()
 
-    compute_fbank_ami(speed_perturb=args.speed_perturb)
+    compute_fbank_ami(perturb_speed=args.perturb_speed)

egs/alimeeting/ASR_v2/prepare.sh

@@ -85,7 +85,7 @@ fi
 if [ $stage -le 5 ] && [ $stop_stage -ge 5 ]; then
   log "Stage 5: Compute fbank for alimeeting"
   mkdir -p data/fbank
-  python local/compute_fbank_alimeeting.py --speed-perturb True
+  python local/compute_fbank_alimeeting.py --perturb-speed True
   log "Combine features from train splits"
   lhotse combine data/manifests/cuts_train_{ihm,ihm_rvb,sdm,gss}.jsonl.gz - | shuf |\
     gzip -c > data/manifests/cuts_train_all.jsonl.gz

egs/wenetspeech/ASR/local/preprocess_wenetspeech.py

@@ -25,6 +25,7 @@ from lhotse import CutSet, SupervisionSegment
 from lhotse.recipes.utils import read_manifests_if_cached
 
 from icefall import setup_logger
+from icefall.utils import str2bool
 
 # Similar text filtering and normalization procedure as in:
 # https://github.com/SpeechColab/WenetSpeech/blob/main/toolkits/kaldi/wenetspeech_data_prep.sh
@@ -46,7 +47,7 @@ def has_no_oov(
     return oov_pattern.search(sup.text) is None
 
 
-def preprocess_wenet_speech(speed_perturb: bool = False):
+def preprocess_wenet_speech(perturb_speed: bool = False):
     src_dir = Path("data/manifests")
     output_dir = Path("data/fbank")
     output_dir.mkdir(exist_ok=True)
@@ -111,7 +112,7 @@ def preprocess_wenet_speech(speed_perturb: bool = False):
         )
         # Run data augmentation that needs to be done in the
         # time domain.
-        if partition not in ["DEV", "TEST_NET", "TEST_MEETING"] and speed_perturb:
+        if partition not in ["DEV", "TEST_NET", "TEST_MEETING"] and perturb_speed:
             logging.info(
                 f"Speed perturb for {partition} with factors 0.9 and 1.1 "
                 "(Perturbing may take 8 minutes and saving may take 20 minutes)"
@@ -124,8 +125,8 @@ def preprocess_wenet_speech(speed_perturb: bool = False):
 def get_args():
     parser = argparse.ArgumentParser()
     parser.add_argument(
-        "--speed-perturb",
-        type=bool,
+        "--perturb-speed",
+        type=str2bool,
         default=False,
         help="Enable 0.9 and 1.1 speed perturbation for data augmentation. Default: False.",
     )
@@ -136,7 +137,7 @@ def main():
     setup_logger(log_filename="./log-preprocess-wenetspeech")
 
     args = get_args()
-    preprocess_wenet_speech(speed_perturb=args.speed_perturb)
+    preprocess_wenet_speech(perturb_speed=args.perturb_speed)
     logging.info("Done")
 
 

egs/wenetspeech/ASR/prepare.sh

@@ -91,7 +91,7 @@ fi
 if [ $stage -le 3 ] && [ $stop_stage -ge 3 ]; then
   log "Stage 3: Preprocess WenetSpeech manifest"
   if [ ! -f data/fbank/.preprocess_complete ]; then
-    python3 ./local/preprocess_wenetspeech.py --speed-perturb True
+    python3 ./local/preprocess_wenetspeech.py --perturb-speed True
     touch data/fbank/.preprocess_complete
   fi
 fi