Set overwrite=True when extracting features in batches. (#487)

egs/gigaspeech/ASR/local/compute_fbank_gigaspeech_dev_test.py

@@ -20,11 +20,7 @@ import logging
 from pathlib import Path
 
 import torch
-from lhotse import (
+from lhotse import CutSet, KaldifeatFbank, KaldifeatFbankConfig
-    CutSet,
-    KaldifeatFbank,
-    KaldifeatFbankConfig,
-)
 
 # Torch's multithreaded behavior needs to be disabled or
 # it wastes a lot of CPU and slow things down.
@@ -69,6 +65,7 @@ def compute_fbank_gigaspeech_dev_test():
             storage_path=f"{in_out_dir}/feats_{partition}",
             num_workers=num_workers,
             batch_duration=batch_duration,
+            overwrite=True,
         )
         cut_set = cut_set.trim_to_supervisions(
             keep_overlapping=False, min_duration=None

egs/gigaspeech/ASR/local/compute_fbank_gigaspeech_splits.py

@@ -22,11 +22,7 @@ from datetime import datetime
 from pathlib import Path
 
 import torch
-from lhotse import (
+from lhotse import CutSet, KaldifeatFbank, KaldifeatFbankConfig
-    CutSet,
-    KaldifeatFbank,
-    KaldifeatFbankConfig,
-)
 
 # Torch's multithreaded behavior needs to be disabled or
 # it wastes a lot of CPU and slow things down.
@@ -120,6 +116,7 @@ def compute_fbank_gigaspeech_splits(args):
             storage_path=f"{output_dir}/feats_XL_{idx}",
             num_workers=args.num_workers,
             batch_duration=args.batch_duration,
+            overwrite=True,
         )
 
         logging.info("About to split cuts into smaller chunks.")

egs/librispeech/ASR/local/compute_fbank_gigaspeech_dev_test.py

@@ -68,6 +68,7 @@ def compute_fbank_gigaspeech_dev_test():
             storage_path=f"{in_out_dir}/{prefix}_feats_{partition}",
             num_workers=num_workers,
             batch_duration=batch_duration,
+            overwrite=True,
         )
         cut_set = cut_set.trim_to_supervisions(
             keep_overlapping=False, min_duration=None

egs/librispeech/ASR/local/compute_fbank_gigaspeech_splits.py

@@ -126,6 +126,7 @@ def compute_fbank_gigaspeech_splits(args):
             storage_path=f"{output_dir}/{prefix}_feats_XL_{idx}",
             num_workers=args.num_workers,
             batch_duration=args.batch_duration,
+            overwrite=True,
         )
 
         logging.info("About to split cuts into smaller chunks.")

egs/spgispeech/ASR/local/compute_fbank_musan.py

@@ -92,6 +92,7 @@ def compute_fbank_musan():
             batch_duration=500,
             num_workers=4,
             storage_type=LilcomChunkyWriter,
+            overwrite=True,
         )
     )
 

egs/spgispeech/ASR/local/compute_fbank_spgispeech.py

@@ -119,6 +119,7 @@ def compute_fbank_spgispeech(args):
                 batch_duration=500,
                 num_workers=4,
                 storage_type=LilcomChunkyWriter,
+                overwrite=True,
             )
             cs.to_file(cuts_train_idx_path)
 
@@ -138,6 +139,7 @@ def compute_fbank_spgispeech(args):
                 batch_duration=500,
                 num_workers=4,
                 storage_type=LilcomChunkyWriter,
+                overwrite=True,
             )
 
 

egs/wenetspeech/ASR/local/compute_fbank_wenetspeech_dev_test.py

@@ -75,6 +75,7 @@ def compute_fbank_wenetspeech_dev_test():
             num_workers=num_workers,
             batch_duration=batch_duration,
             storage_type=LilcomHdf5Writer,
+            overwrite=True,
         )
 
         logging.info(f"Saving to {cuts_path}")

egs/wenetspeech/ASR/local/compute_fbank_wenetspeech_splits.py

@@ -140,6 +140,7 @@ def compute_fbank_wenetspeech_splits(args):
             num_workers=args.num_workers,
             batch_duration=args.batch_duration,
             storage_type=LilcomChunkyWriter,
+            overwrite=True,
         )
 
         logging.info(f"Saving to {cuts_path}")