From daff070d6849ceab9cd8637566614cd5f1e84a10 Mon Sep 17 00:00:00 2001
From: Kinan Martin <kinanmartin0@gmail.com>
Date: Thu, 15 May 2025 07:24:26 +0900
Subject: [PATCH] restore version of mls_english compute_fbank_mls_english.py
 and prepare.sh from commit 547f5c5

---
 .../ASR/local/compute_fbank_mls_english.py          |  6 +++++-
 egs/mls_english/ASR/prepare.sh                      | 13 ++++++++-----
 2 files changed, 13 insertions(+), 6 deletions(-)

diff --git a/egs/mls_english/ASR/local/compute_fbank_mls_english.py b/egs/mls_english/ASR/local/compute_fbank_mls_english.py
index 8c5cae842..e9bd81551 100644
--- a/egs/mls_english/ASR/local/compute_fbank_mls_english.py
+++ b/egs/mls_english/ASR/local/compute_fbank_mls_english.py
@@ -120,6 +120,11 @@ def main():
         cut_sets = make_cutset_blueprints(mls_eng_hf_dataset_path)
         for part, cut_set in cut_sets:
             logging.info(f"Processing {part}")
+            cut_set = cut_set.save_audios(
+                num_jobs=num_jobs,
+                storage_path=(args.audio_dir / part).as_posix(),
+            ) # makes new cutset that loads audio from paths to actual audio files
+            
             cut_set = cut_set.compute_and_store_features(
                 extractor=extractor,
                 num_jobs=num_jobs,
@@ -127,7 +132,6 @@ def main():
                 storage_type=LilcomChunkyWriter,
             )
 
-            cut_set = cut_set.save_audios(args.audio_dir / part) # makes new cutset that uses paths to actual audio files
             cut_set.to_file(args.manifest_dir / f"mls_eng_cuts_{part}.jsonl.gz")
 
         logging.info("All fbank computed for MLS English.")
diff --git a/egs/mls_english/ASR/prepare.sh b/egs/mls_english/ASR/prepare.sh
index ad1c10080..14ca69dae 100755
--- a/egs/mls_english/ASR/prepare.sh
+++ b/egs/mls_english/ASR/prepare.sh
@@ -60,9 +60,9 @@ if [ $stage -le 1 ] && [ $stop_stage -ge 1 ]; then
           --audio-dir data/audio \
           --dl-dir $dl_dir/mls_english
           # --dl-dir /root/datasets/parler-tts--mls_eng
-        python local/validate_manifest.py --manifest data/manifests/mls_english_cuts_train.jsonl.gz
-        python local/validate_manifest.py --manifest data/manifests/mls_english_cuts_dev.jsonl.gz
-        python local/validate_manifest.py --manifest data/manifests/mls_english_cuts_test.jsonl.gz
+        python local/validate_manifest.py --manifest data/manifests/mls_eng_cuts_train.jsonl.gz
+        python local/validate_manifest.py --manifest data/manifests/mls_eng_cuts_dev.jsonl.gz
+        python local/validate_manifest.py --manifest data/manifests/mls_eng_cuts_test.jsonl.gz
         touch data/manifests/.mls_english-validated.done
     fi
 fi
@@ -71,7 +71,10 @@ if [ $stage -le 2 ] && [ $stop_stage -ge 2 ]; then
   log "Stage 2: Prepare transcript for BPE training"
   if [ ! -f data/lang/transcript.txt ]; then
     log "Generating transcripts for BPE training"
-    ./local/utils/generate_transcript.py --lang-dir data/lang
+    python local/utils/generate_transcript.py \
+      --dataset-path $dl_dir/mls_english \
+      --lang-dir data/lang \
+      --split train 
   fi
 fi
 
@@ -83,7 +86,7 @@ if [ $stage -le 3 ] && [ $stop_stage -ge 3 ]; then
     mkdir -p $bpe_dir
 
     if [ ! -f $bpe_dir/bpe.model ]; then
-      ./local/train_bpe_model.py \
+      python local/train_bpe_model.py \
         --lang-dir $bpe_dir \
         --vocab-size $vocab_size \
         --transcript data/lang/transcript.txt