Rename train2.py to avoid confusion (#1386)

.github/scripts/run-multi-zh_hans-zipformer.sh

@@ -51,6 +51,8 @@ for method in modified_beam_search fast_beam_search; do
   $repo/test_wavs/DEV_T0000000002.wav
 done
 
+rm -rf $repo
+
 log "==== Test icefall-asr-multi-zh-hans-zipformer-ctc-2023-10-24 ===="
 repo_url=https://huggingface.co/zrjin/icefall-asr-multi-zh-hans-zipformer-ctc-2023-10-24/
 
@@ -92,4 +94,4 @@ for method in modified_beam_search fast_beam_search; do
   $repo/test_wavs/DEV_T0000000000.wav \
   $repo/test_wavs/DEV_T0000000001.wav \
   $repo/test_wavs/DEV_T0000000002.wav
-done
+done

egs/aishell/ASR/prepare.sh

@@ -261,10 +261,9 @@ if [ $stage -le 8 ] && [ $stop_stage -ge 8 ]; then
   fi
 
   if [ ! -f $lang_char_dir/HLG.fst ]; then
-    lang_phone_dir=data/lang_phone
     ./local/prepare_lang_fst.py  \
-      --lang-dir $lang_phone_dir \
+      --lang-dir $lang_char_dir \
-      --ngram-G ./data/lm/G_3_gram.fst.txt
+      --ngram-G ./data/lm/G_3_gram_char.fst.txt
   fi
 fi
 

egs/aishell/ASR/pruned_transducer_stateless7/do_not_use_it_directly.py

@@ -1234,6 +1234,7 @@ def scan_pessimistic_batches_for_oom(
 
 
 def main():
+    raise RuntimeError("Please don't use this file directly!")
     parser = get_parser()
     AsrDataModule.add_arguments(parser)
     args = parser.parse_args()

egs/aishell/ASR/pruned_transducer_stateless7/export-onnx.py

@@ -56,7 +56,7 @@ import torch.nn as nn
 from decoder2 import Decoder
 from onnxruntime.quantization import QuantType, quantize_dynamic
 from scaling_converter import convert_scaled_to_non_scaled
-from train2 import add_model_arguments, get_params, get_transducer_model
+from do_not_use_it_directly import add_model_arguments, get_params, get_transducer_model
 from zipformer import Zipformer
 
 from icefall.checkpoint import (

egs/aishell/ASR/pruned_transducer_stateless7_streaming/do_not_use_it_directly.py

@@ -1233,6 +1233,7 @@ def scan_pessimistic_batches_for_oom(
 
 
 def main():
+    raise RuntimeError("Please don't use this file directly!")
     parser = get_parser()
     AishellAsrDataModule.add_arguments(parser)
     args = parser.parse_args()

egs/commonvoice/ASR/pruned_transducer_stateless7_streaming/README.md

@@ -4,6 +4,6 @@ See https://k2-fsa.github.io/icefall/recipes/Streaming-ASR/librispeech/zipformer
 
 [./emformer.py](./emformer.py) and [./train.py](./train.py)
 are basically the same as
-[./emformer2.py](./emformer2.py) and [./train2.py](./train2.py).
+[./emformer2.py](./emformer2.py) and [./do_not_use_it_directly.py](./do_not_use_it_directly.py).
-The only purpose of [./emformer2.py](./emformer2.py) and [./train2.py](./train2.py)
+The only purpose of [./emformer2.py](./emformer2.py) and [./do_not_use_it_directly.py](./do_not_use_it_directly.py)
 is for exporting to [sherpa-ncnn](https://github.com/k2-fsa/sherpa-ncnn).

egs/commonvoice/ASR/pruned_transducer_stateless7_streaming/do_not_use_it_directly.py

@@ -1237,6 +1237,7 @@ def scan_pessimistic_batches_for_oom(
 
 
 def main():
+    raise RuntimeError("Please don't use this file directly!")
     parser = get_parser()
     CommonVoiceAsrDataModule.add_arguments(parser)
     args = parser.parse_args()

egs/csj/ASR/pruned_transducer_stateless7_streaming/do_not_use_it_directly.py

@@ -1274,6 +1274,7 @@ def scan_pessimistic_batches_for_oom(
 
 
 def main():
+    raise RuntimeError("Please don't use this file directly!")
     parser = get_parser()
     CSJAsrDataModule.add_arguments(parser)
     Tokenizer.add_arguments(parser)

egs/csj/ASR/pruned_transducer_stateless7_streaming/export-for-ncnn.py

@@ -72,7 +72,7 @@ from pathlib import Path
 import torch
 from scaling_converter import convert_scaled_to_non_scaled
 from tokenizer import Tokenizer
-from train2 import add_model_arguments, get_params, get_transducer_model
+from do_not_use_it_directly import add_model_arguments, get_params, get_transducer_model
 
 from icefall.checkpoint import (
     average_checkpoints,

egs/librispeech/ASR/conv_emformer_transducer_stateless2/do_not_use_it_directly.py

@@ -1099,6 +1099,7 @@ def scan_pessimistic_batches_for_oom(
 
 
 def main():
+    raise RuntimeError("Please don't use this file directly!")
     parser = get_parser()
     LibriSpeechAsrDataModule.add_arguments(parser)
     args = parser.parse_args()

egs/librispeech/ASR/conv_emformer_transducer_stateless2/export-for-ncnn.py

@@ -39,8 +39,8 @@ from pathlib import Path
 
 import k2
 import torch
+from do_not_use_it_directly import add_model_arguments, get_params, get_transducer_model
 from scaling_converter import convert_scaled_to_non_scaled
-from train2 import add_model_arguments, get_params, get_transducer_model
 
 from icefall.checkpoint import (
     average_checkpoints,

egs/librispeech/ASR/conv_emformer_transducer_stateless2/export-onnx.py

@@ -61,7 +61,7 @@ import torch.nn as nn
 from decoder import Decoder
 from emformer import Emformer
 from scaling_converter import convert_scaled_to_non_scaled
-from train2 import add_model_arguments, get_params, get_transducer_model
+from do_not_use_it_directly import add_model_arguments, get_params, get_transducer_model
 
 from icefall.checkpoint import (
     average_checkpoints,

egs/librispeech/ASR/pruned_transducer_stateless7_streaming/README.md

@@ -4,7 +4,7 @@ See https://k2-fsa.github.io/icefall/recipes/Streaming-ASR/librispeech/zipformer
 
 [./emformer.py](./emformer.py) and [./train.py](./train.py)
 are basically the same as
-[./emformer2.py](./emformer2.py) and [./train2.py](./train2.py).
+[./emformer2.py](./emformer2.py) and [./do_not_use_it_directly.py](./do_not_use_it_directly.py).
-The only purpose of [./emformer2.py](./emformer2.py) and [./train2.py](./train2.py)
+The only purpose of [./emformer2.py](./emformer2.py) and [./do_not_use_it_directly.py](./do_not_use_it_directly.py)
 is for exporting to [sherpa-ncnn](https://github.com/k2-fsa/sherpa-ncnn).
 

egs/librispeech/ASR/pruned_transducer_stateless7_streaming/do_not_use_it_directly.py

@@ -1234,6 +1234,7 @@ def scan_pessimistic_batches_for_oom(
 
 
 def main():
+    raise RuntimeError("Please don't use this file directly!")
     parser = get_parser()
     LibriSpeechAsrDataModule.add_arguments(parser)
     args = parser.parse_args()

egs/librispeech/ASR/pruned_transducer_stateless7_streaming/export-for-ncnn-zh.py

@@ -68,8 +68,8 @@ from pathlib import Path
 
 import k2
 import torch
+from do_not_use_it_directly import add_model_arguments, get_params, get_transducer_model
 from scaling_converter import convert_scaled_to_non_scaled
-from train2 import add_model_arguments, get_params, get_transducer_model
 
 from icefall.checkpoint import (
     average_checkpoints,

egs/librispeech/ASR/pruned_transducer_stateless7_streaming/export-for-ncnn.py

@@ -66,8 +66,8 @@ from pathlib import Path
 
 import k2
 import torch
+from do_not_use_it_directly import add_model_arguments, get_params, get_transducer_model
 from scaling_converter import convert_scaled_to_non_scaled
-from train2 import add_model_arguments, get_params, get_transducer_model
 
 from icefall.checkpoint import (
     average_checkpoints,

egs/librispeech/ASR/pruned_transducer_stateless7_streaming_multi/do_not_use_it_directly.py

@@ -0,0 +1 @@
+../pruned_transducer_stateless7_streaming/do_not_use_it_directly.py

egs/librispeech/ASR/pruned_transducer_stateless7_streaming_multi/export-for-ncnn.py

@@ -66,8 +66,8 @@ from pathlib import Path
 
 import k2
 import torch
+from do_not_use_it_directly import add_model_arguments, get_params, get_transducer_model
 from scaling_converter import convert_scaled_to_non_scaled
-from train2 import add_model_arguments, get_params, get_transducer_model
 
 from icefall.checkpoint import (
     average_checkpoints,

egs/librispeech/ASR/pruned_transducer_stateless7_streaming_multi/train2.py

@@ -1 +0,0 @@
-../pruned_transducer_stateless7_streaming/train2.py