fix removed items

k2-fsa · Apr 30, 2024 · c292dc6 · c292dc6
1 parent 35c2622
commit c292dc6
Show file tree

Hide file tree

Showing 3 changed files with 13 additions and 4 deletions.
diff --git a/egs/multi_zh-hans/ASR/prepare.sh b/egs/multi_zh-hans/ASR/prepare.sh
@@ -299,6 +299,15 @@ if [ $stage -le 12 ] && [ $stop_stage -ge 12 ]; then
     log "Compute KeSpeech fbank for test/dev"
     ./local/compute_fbank_kespeech_dev_test.py
 
+    if [ ! -f data/fbank/kespeech/kespeech-asr_cuts_train_phase1.jsonl.gz ]; then
+      pieces=$(find data/fbank/kespeech/train_phase1_split_${num_splits} -name "kespeech-asr_cuts_train_phase1.*.jsonl.gz")
+      lhotse combine $pieces data/fbank/kespeech/kespeech-asr_cuts_train_phase1.jsonl.gz
+    fi
+    if [ ! -f data/fbank/kespeech/kespeech-asr_cuts_train_phase2.jsonl.gz ]; then
+      pieces=$(find data/fbank/kespeech/train_phase2_split_${num_splits} -name "kespeech-asr_cuts_train_phase2.*.jsonl.gz")
+      lhotse combine $pieces data/fbank/kespeech/kespeech-asr_cuts_train_phase2.jsonl.gz
+    fi
+
     touch data/fbank/.kespeech.done
   fi
 fi

diff --git a/egs/multi_zh-hans/ASR/whisper/train.py b/egs/multi_zh-hans/ASR/whisper/train.py
@@ -147,7 +147,7 @@ def get_parser():
         "--model-name",
         type=str,
         default="large-v2",
-        choices=["large-v2", "large-v3", "medium", "small", "tiny"],
+        choices=["large-v2", "large-v3", "medium", "base", "small", "tiny"],
         help="""The model name to use.
         """,
     )
@@ -870,7 +870,7 @@ def run(rank, world_size, args):
 
     if params.print_diagnostics:
         opts = diagnostics.TensorDiagnosticOptions(
-            2**22
+            512
         )  # allow 4 megabytes per sub-module
         diagnostic = diagnostics.attach_diagnostics(model, opts)
 

diff --git a/egs/wenetspeech/ASR/whisper/train.py b/egs/wenetspeech/ASR/whisper/train.py
@@ -146,7 +146,7 @@ def get_parser():
         "--model-name",
         type=str,
         default="large-v2",
-        choices=["large-v2", "large-v3", "medium", "small", "tiny"],
+        choices=["large-v2", "large-v3", "medium", "base", "small", "tiny"],
         help="""The model name to use.
         """,
     )
@@ -806,7 +806,7 @@ def run(rank, world_size, args):
 
     if params.print_diagnostics:
         opts = diagnostics.TensorDiagnosticOptions(
-            2**22
+            512
         )  # allow 4 megabytes per sub-module
         diagnostic = diagnostics.attach_diagnostics(model, opts)