update

README.md

@@ -13,17 +13,21 @@ Check out the configuration reference at https://huggingface.co/docs/hub/spaces-
 ### ASR
 
 ```text
-docker build -t asr:v20240510_1734 .
+docker build -t asr:v20240511_1022 .
 
 docker run -itd --name ASR \
 --network host \
 -v /data/tianxing/PycharmProjects/asr/data/:/home/user/app/data/ \
 -v /data/tianxing/PycharmProjects/asr/pretrained_models/:/home/user/app/pretrained_models/ \
-asr:v20240510_1734 /bin/bash
+asr:v20240511_1022 /bin/bash
 
 ```
 
 
+```text
+docker cp ASR:/tmp/asr/si_chuan_hua.wav /home/nlp
+```
+
 ### 异常1
 
 https://blog.csdn.net/sinat_28442665/article/details/126154991

examples/wenet/toolbox_infer.py

@@ -0,0 +1,79 @@
+#!/usr/bin/python3
+# -*- coding: utf-8 -*-
+import argparse
+import os
+from pathlib import Path
+import sys
+import tempfile
+
+pwd = os.path.abspath(os.path.dirname(__file__))
+sys.path.append(os.path.join(pwd, "../../"))
+
+import librosa
+import numpy as np
+import sherpa
+from scipy.io import wavfile
+import torch
+import torchaudio
+
+from project_settings import project_path, temp_directory
+from toolbox.k2_sherpa.utils import audio_convert
+from toolbox.k2_sherpa import decode, models
+
+
+def get_args():
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "--model_dir",
+        default=(project_path / "pretrained_models/huggingface/csukuangfj/wenet-chinese-model").as_posix(),
+        type=str
+    )
+    parser.add_argument(
+        "--in_filename",
+        default=(project_path / "data/test_wavs/paraformer-zh/si_chuan_hua.wav").as_posix(),
+        type=str
+    )
+    parser.add_argument("--sample_rate", default=16000, type=int)
+    args = parser.parse_args()
+    return args
+
+
+def main():
+    args = get_args()
+
+    # audio convert
+    in_filename = Path(args.in_filename)
+    out_filename = Path(tempfile.gettempdir()) / "asr" / in_filename.name
+    out_filename.parent.mkdir(parents=True, exist_ok=True)
+
+    audio_convert(in_filename=in_filename.as_posix(),
+                  out_filename=out_filename.as_posix(),
+                  )
+
+    # load recognizer
+    m_dict = models.model_map["Chinese"][0]
+
+    local_model_dir = Path(args.model_dir)
+    nn_model_file = local_model_dir / m_dict["nn_model_file"]
+    tokens_file = local_model_dir / m_dict["tokens_file"]
+
+    recognizer = models.load_recognizer(
+        repo_id=m_dict["repo_id"],
+        nn_model_file=nn_model_file.as_posix(),
+        tokens_file=tokens_file.as_posix(),
+        sub_folder=m_dict["sub_folder"],
+        local_model_dir=local_model_dir,
+        recognizer_type=m_dict["recognizer_type"],
+        decoding_method="greedy_search",
+        num_active_paths=2,
+    )
+
+    text = decode.decode_by_recognizer(recognizer=recognizer,
+                                       filename=out_filename.as_posix(),
+                                       )
+    print("text: {}".format(text))
+    return
+
+
+if __name__ == "__main__":
+    main()

start.sh

@@ -0,0 +1,7 @@
+#!/usr/bin/env bash
+
+export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/lib/python3.8/site-packages/k2/lib/
+
+rm -rf logs/
+
+python3 run_asr_server.py

stop.sh

@@ -0,0 +1,3 @@
+#!/usr/bin/env bash
+
+kill -9 `ps -aef | grep 'run_asr_server.py' | grep -v grep | awk '{print $2}'`

toolbox/k2_sherpa/models.py

@@ -80,6 +80,7 @@ def load_sherpa_offline_recognizer(nn_model_file: str,
     )
 
     recognizer = sherpa.OfflineRecognizer(config)
+
     return recognizer