diff --git a/_sources/python_api/wenet.paraformer.convert_paraformer_to_wenet_config.rst.txt b/_sources/python_api/wenet.paraformer.convert_paraformer_to_wenet_config.rst.txt
deleted file mode 100644
index 5dee656c1..000000000
--- a/_sources/python_api/wenet.paraformer.convert_paraformer_to_wenet_config.rst.txt
+++ /dev/null
@@ -1,7 +0,0 @@
-wenet.paraformer.convert\_paraformer\_to\_wenet\_config module
-==============================================================
-
-.. automodule:: wenet.paraformer.convert_paraformer_to_wenet_config
-   :members:
-   :undoc-members:
-   :show-inheritance:
diff --git a/_sources/python_api/wenet.paraformer.convert_paraformer_to_wenet_config_and_ckpt.rst.txt b/_sources/python_api/wenet.paraformer.convert_paraformer_to_wenet_config_and_ckpt.rst.txt
new file mode 100644
index 000000000..d234b224e
--- /dev/null
+++ b/_sources/python_api/wenet.paraformer.convert_paraformer_to_wenet_config_and_ckpt.rst.txt
@@ -0,0 +1,7 @@
+wenet.paraformer.convert\_paraformer\_to\_wenet\_config\_and\_ckpt module
+=========================================================================
+
+.. automodule:: wenet.paraformer.convert_paraformer_to_wenet_config_and_ckpt
+   :members:
+   :undoc-members:
+   :show-inheritance:
diff --git a/_sources/python_api/wenet.paraformer.rst.txt b/_sources/python_api/wenet.paraformer.rst.txt
index 77d4a8b27..1e38e97ae 100644
--- a/_sources/python_api/wenet.paraformer.rst.txt
+++ b/_sources/python_api/wenet.paraformer.rst.txt
@@ -14,7 +14,7 @@ Submodules
 
    wenet.paraformer.attention
    wenet.paraformer.cif
-   wenet.paraformer.convert_paraformer_to_wenet_config
+   wenet.paraformer.convert_paraformer_to_wenet_config_and_ckpt
    wenet.paraformer.layers
    wenet.paraformer.paraformer
    wenet.paraformer.search
diff --git a/objects.inv b/objects.inv
index d0d67ff80..4609ee8a8 100644
Binary files a/objects.inv and b/objects.inv differ
diff --git a/python_api/wenet.paraformer.cif.html b/python_api/wenet.paraformer.cif.html
index 0b7388dbb..0fa841eb5 100644
--- a/python_api/wenet.paraformer.cif.html
+++ b/python_api/wenet.paraformer.cif.html
@@ -20,7 +20,7 @@
     <script src="../_static/js/theme.js"></script>
     <link rel="index" title="Index" href="../genindex.html" />
     <link rel="search" title="Search" href="../search.html" />
-    <link rel="next" title="wenet.paraformer.convert_paraformer_to_wenet_config module" href="wenet.paraformer.convert_paraformer_to_wenet_config.html" />
+    <link rel="next" title="wenet.paraformer.convert_paraformer_to_wenet_config_and_ckpt module" href="wenet.paraformer.convert_paraformer_to_wenet_config_and_ckpt.html" />
     <link rel="prev" title="wenet.paraformer.attention module" href="wenet.paraformer.attention.html" /> 
 </head>
 
@@ -98,7 +98,7 @@ <h1>wenet.paraformer.cif module<a class="headerlink" href="#wenet-paraformer-cif
           </div>
           <footer><div class="rst-footer-buttons" role="navigation" aria-label="Footer">
         <a href="wenet.paraformer.attention.html" class="btn btn-neutral float-left" title="wenet.paraformer.attention module" accesskey="p" rel="prev"><span class="fa fa-arrow-circle-left" aria-hidden="true"></span> Previous</a>
-        <a href="wenet.paraformer.convert_paraformer_to_wenet_config.html" class="btn btn-neutral float-right" title="wenet.paraformer.convert_paraformer_to_wenet_config module" accesskey="n" rel="next">Next <span class="fa fa-arrow-circle-right" aria-hidden="true"></span></a>
+        <a href="wenet.paraformer.convert_paraformer_to_wenet_config_and_ckpt.html" class="btn btn-neutral float-right" title="wenet.paraformer.convert_paraformer_to_wenet_config_and_ckpt module" accesskey="n" rel="next">Next <span class="fa fa-arrow-circle-right" aria-hidden="true"></span></a>
     </div>
 
   <hr/>
diff --git a/python_api/wenet.paraformer.convert_paraformer_to_wenet_config.html b/python_api/wenet.paraformer.convert_paraformer_to_wenet_config_and_ckpt.html
similarity index 92%
rename from python_api/wenet.paraformer.convert_paraformer_to_wenet_config.html
rename to python_api/wenet.paraformer.convert_paraformer_to_wenet_config_and_ckpt.html
index 0b5115ffc..f36a3ab50 100644
--- a/python_api/wenet.paraformer.convert_paraformer_to_wenet_config.html
+++ b/python_api/wenet.paraformer.convert_paraformer_to_wenet_config_and_ckpt.html
@@ -4,7 +4,7 @@
   <meta charset="utf-8" /><meta name="generator" content="Docutils 0.18.1: http://docutils.sourceforge.net/" />
 
   <meta name="viewport" content="width=device-width, initial-scale=1.0" />
-  <title>wenet.paraformer.convert_paraformer_to_wenet_config module &mdash; wenet  documentation</title>
+  <title>wenet.paraformer.convert_paraformer_to_wenet_config_and_ckpt module &mdash; wenet  documentation</title>
       <link rel="stylesheet" href="../_static/pygments.css" type="text/css" />
       <link rel="stylesheet" href="../_static/css/theme.css" type="text/css" />
   <!--[if lt IE 9]>
@@ -79,9 +79,9 @@
           <li class="breadcrumb-item"><a href="modules.html">Python API Reference</a></li>
           <li class="breadcrumb-item"><a href="wenet.html">wenet package</a></li>
           <li class="breadcrumb-item"><a href="wenet.paraformer.html">wenet.paraformer package</a></li>
-      <li class="breadcrumb-item active">wenet.paraformer.convert_paraformer_to_wenet_config module</li>
+      <li class="breadcrumb-item active">wenet.paraformer.convert_paraformer_to_wenet_config_and_ckpt module</li>
       <li class="wy-breadcrumbs-aside">
-            <a href="../_sources/python_api/wenet.paraformer.convert_paraformer_to_wenet_config.rst.txt" rel="nofollow"> View page source</a>
+            <a href="../_sources/python_api/wenet.paraformer.convert_paraformer_to_wenet_config_and_ckpt.rst.txt" rel="nofollow"> View page source</a>
       </li>
   </ul>
   <hr/>
@@ -89,8 +89,8 @@
           <div role="main" class="document" itemscope="itemscope" itemtype="http://schema.org/Article">
            <div itemprop="articleBody">
              
-  <section id="wenet-paraformer-convert-paraformer-to-wenet-config-module">
-<h1>wenet.paraformer.convert_paraformer_to_wenet_config module<a class="headerlink" href="#wenet-paraformer-convert-paraformer-to-wenet-config-module" title="Permalink to this heading"></a></h1>
+  <section id="wenet-paraformer-convert-paraformer-to-wenet-config-and-ckpt-module">
+<h1>wenet.paraformer.convert_paraformer_to_wenet_config_and_ckpt module<a class="headerlink" href="#wenet-paraformer-convert-paraformer-to-wenet-config-and-ckpt-module" title="Permalink to this heading"></a></h1>
 </section>
 
 
diff --git a/python_api/wenet.paraformer.html b/python_api/wenet.paraformer.html
index a174584b6..18a1cce2b 100644
--- a/python_api/wenet.paraformer.html
+++ b/python_api/wenet.paraformer.html
@@ -96,7 +96,7 @@ <h2>Submodules<a class="headerlink" href="#submodules" title="Permalink to this
 <ul>
 <li class="toctree-l1"><a class="reference internal" href="wenet.paraformer.attention.html">wenet.paraformer.attention module</a></li>
 <li class="toctree-l1"><a class="reference internal" href="wenet.paraformer.cif.html">wenet.paraformer.cif module</a></li>
-<li class="toctree-l1"><a class="reference internal" href="wenet.paraformer.convert_paraformer_to_wenet_config.html">wenet.paraformer.convert_paraformer_to_wenet_config module</a></li>
+<li class="toctree-l1"><a class="reference internal" href="wenet.paraformer.convert_paraformer_to_wenet_config_and_ckpt.html">wenet.paraformer.convert_paraformer_to_wenet_config_and_ckpt module</a></li>
 <li class="toctree-l1"><a class="reference internal" href="wenet.paraformer.layers.html">wenet.paraformer.layers module</a></li>
 <li class="toctree-l1"><a class="reference internal" href="wenet.paraformer.paraformer.html">wenet.paraformer.paraformer module</a></li>
 <li class="toctree-l1"><a class="reference internal" href="wenet.paraformer.search.html">wenet.paraformer.search module</a></li>
diff --git a/python_api/wenet.paraformer.layers.html b/python_api/wenet.paraformer.layers.html
index b73bb9ee7..d99ffc9d1 100644
--- a/python_api/wenet.paraformer.layers.html
+++ b/python_api/wenet.paraformer.layers.html
@@ -21,7 +21,7 @@
     <link rel="index" title="Index" href="../genindex.html" />
     <link rel="search" title="Search" href="../search.html" />
     <link rel="next" title="wenet.paraformer.paraformer module" href="wenet.paraformer.paraformer.html" />
-    <link rel="prev" title="wenet.paraformer.convert_paraformer_to_wenet_config module" href="wenet.paraformer.convert_paraformer_to_wenet_config.html" /> 
+    <link rel="prev" title="wenet.paraformer.convert_paraformer_to_wenet_config_and_ckpt module" href="wenet.paraformer.convert_paraformer_to_wenet_config_and_ckpt.html" /> 
 </head>
 
 <body class="wy-body-for-nav"> 
@@ -97,7 +97,7 @@ <h1>wenet.paraformer.layers module<a class="headerlink" href="#wenet-paraformer-
            </div>
           </div>
           <footer><div class="rst-footer-buttons" role="navigation" aria-label="Footer">
-        <a href="wenet.paraformer.convert_paraformer_to_wenet_config.html" class="btn btn-neutral float-left" title="wenet.paraformer.convert_paraformer_to_wenet_config module" accesskey="p" rel="prev"><span class="fa fa-arrow-circle-left" aria-hidden="true"></span> Previous</a>
+        <a href="wenet.paraformer.convert_paraformer_to_wenet_config_and_ckpt.html" class="btn btn-neutral float-left" title="wenet.paraformer.convert_paraformer_to_wenet_config_and_ckpt module" accesskey="p" rel="prev"><span class="fa fa-arrow-circle-left" aria-hidden="true"></span> Previous</a>
         <a href="wenet.paraformer.paraformer.html" class="btn btn-neutral float-right" title="wenet.paraformer.paraformer module" accesskey="n" rel="next">Next <span class="fa fa-arrow-circle-right" aria-hidden="true"></span></a>
     </div>
 
diff --git a/searchindex.js b/searchindex.js
index c3791ef45..4b8904569 100644
--- a/searchindex.js
+++ b/searchindex.js
@@ -1 +1 @@
-Search.setIndex({"docnames": ["UIO", "context", "index", "jit_in_wenet", "lm", "papers", "pretrained_models", "pretrained_models.en", "production", "python_api/modules", "python_api/wenet", "python_api/wenet.branchformer", "python_api/wenet.branchformer.cgmlp", "python_api/wenet.branchformer.encoder", "python_api/wenet.branchformer.encoder_layer", "python_api/wenet.cli", "python_api/wenet.cli.hub", "python_api/wenet.cli.model", "python_api/wenet.cli.paraformer_model", "python_api/wenet.cli.transcribe", "python_api/wenet.dataset", "python_api/wenet.dataset.dataset", "python_api/wenet.dataset.kaldi_io", "python_api/wenet.dataset.processor", "python_api/wenet.dataset.wav_distortion", "python_api/wenet.efficient_conformer", "python_api/wenet.efficient_conformer.attention", "python_api/wenet.efficient_conformer.convolution", "python_api/wenet.efficient_conformer.encoder", "python_api/wenet.efficient_conformer.encoder_layer", "python_api/wenet.efficient_conformer.subsampling", "python_api/wenet.k2", "python_api/wenet.k2.model", "python_api/wenet.paraformer", "python_api/wenet.paraformer.attention", "python_api/wenet.paraformer.cif", "python_api/wenet.paraformer.convert_paraformer_to_wenet_config", "python_api/wenet.paraformer.layers", "python_api/wenet.paraformer.paraformer", "python_api/wenet.paraformer.search", "python_api/wenet.squeezeformer", "python_api/wenet.squeezeformer.attention", "python_api/wenet.squeezeformer.conv2d", "python_api/wenet.squeezeformer.convolution", "python_api/wenet.squeezeformer.encoder", "python_api/wenet.squeezeformer.encoder_layer", "python_api/wenet.squeezeformer.positionwise_feed_forward", "python_api/wenet.squeezeformer.subsampling", "python_api/wenet.transducer", "python_api/wenet.transducer.joint", "python_api/wenet.transducer.predictor", "python_api/wenet.transducer.transducer", "python_api/wenet.transformer", "python_api/wenet.transformer.asr_model", "python_api/wenet.transformer.attention", "python_api/wenet.transformer.cmvn", "python_api/wenet.transformer.convolution", "python_api/wenet.transformer.ctc", "python_api/wenet.transformer.decoder", "python_api/wenet.transformer.decoder_layer", "python_api/wenet.transformer.embedding", "python_api/wenet.transformer.encoder", "python_api/wenet.transformer.encoder_layer", "python_api/wenet.transformer.label_smoothing_loss", "python_api/wenet.transformer.positionwise_feed_forward", "python_api/wenet.transformer.search", "python_api/wenet.transformer.subsampling", "python_api/wenet.transformer.swish", "python_api/wenet.utils", "python_api/wenet.utils.checkpoint", "python_api/wenet.utils.cmvn", "python_api/wenet.utils.common", "python_api/wenet.utils.config", "python_api/wenet.utils.context_graph", "python_api/wenet.utils.ctc_utils", "python_api/wenet.utils.executor", "python_api/wenet.utils.file_utils", "python_api/wenet.utils.init_ali_paraformer", "python_api/wenet.utils.init_model", "python_api/wenet.utils.mask", "python_api/wenet.utils.scheduler", "python_api/wenet.utils.tokenize_utils", "python_api/wenet.utils.train_utils", "python_api/wenet.whisper", "python_api/wenet.whisper.convert_whisper_to_wenet_config_and_ckpt", "python_api/wenet.whisper.whisper", "python_package", "reference", "requirements", "runtime", "train", "tutorial_aishell", "tutorial_librispeech"], "filenames": ["UIO.md", "context.md", "index.rst", "jit_in_wenet.md", "lm.md", "papers.md", "pretrained_models.md", "pretrained_models.en.md", "production.rst", "python_api/modules.rst", "python_api/wenet.rst", "python_api/wenet.branchformer.rst", "python_api/wenet.branchformer.cgmlp.rst", "python_api/wenet.branchformer.encoder.rst", "python_api/wenet.branchformer.encoder_layer.rst", "python_api/wenet.cli.rst", "python_api/wenet.cli.hub.rst", "python_api/wenet.cli.model.rst", "python_api/wenet.cli.paraformer_model.rst", "python_api/wenet.cli.transcribe.rst", "python_api/wenet.dataset.rst", "python_api/wenet.dataset.dataset.rst", "python_api/wenet.dataset.kaldi_io.rst", "python_api/wenet.dataset.processor.rst", "python_api/wenet.dataset.wav_distortion.rst", "python_api/wenet.efficient_conformer.rst", "python_api/wenet.efficient_conformer.attention.rst", "python_api/wenet.efficient_conformer.convolution.rst", "python_api/wenet.efficient_conformer.encoder.rst", "python_api/wenet.efficient_conformer.encoder_layer.rst", "python_api/wenet.efficient_conformer.subsampling.rst", "python_api/wenet.k2.rst", "python_api/wenet.k2.model.rst", "python_api/wenet.paraformer.rst", "python_api/wenet.paraformer.attention.rst", "python_api/wenet.paraformer.cif.rst", "python_api/wenet.paraformer.convert_paraformer_to_wenet_config.rst", "python_api/wenet.paraformer.layers.rst", "python_api/wenet.paraformer.paraformer.rst", "python_api/wenet.paraformer.search.rst", "python_api/wenet.squeezeformer.rst", "python_api/wenet.squeezeformer.attention.rst", "python_api/wenet.squeezeformer.conv2d.rst", "python_api/wenet.squeezeformer.convolution.rst", "python_api/wenet.squeezeformer.encoder.rst", "python_api/wenet.squeezeformer.encoder_layer.rst", "python_api/wenet.squeezeformer.positionwise_feed_forward.rst", "python_api/wenet.squeezeformer.subsampling.rst", "python_api/wenet.transducer.rst", "python_api/wenet.transducer.joint.rst", "python_api/wenet.transducer.predictor.rst", "python_api/wenet.transducer.transducer.rst", "python_api/wenet.transformer.rst", "python_api/wenet.transformer.asr_model.rst", "python_api/wenet.transformer.attention.rst", "python_api/wenet.transformer.cmvn.rst", "python_api/wenet.transformer.convolution.rst", "python_api/wenet.transformer.ctc.rst", "python_api/wenet.transformer.decoder.rst", "python_api/wenet.transformer.decoder_layer.rst", "python_api/wenet.transformer.embedding.rst", "python_api/wenet.transformer.encoder.rst", "python_api/wenet.transformer.encoder_layer.rst", "python_api/wenet.transformer.label_smoothing_loss.rst", "python_api/wenet.transformer.positionwise_feed_forward.rst", "python_api/wenet.transformer.search.rst", "python_api/wenet.transformer.subsampling.rst", "python_api/wenet.transformer.swish.rst", "python_api/wenet.utils.rst", "python_api/wenet.utils.checkpoint.rst", "python_api/wenet.utils.cmvn.rst", "python_api/wenet.utils.common.rst", "python_api/wenet.utils.config.rst", "python_api/wenet.utils.context_graph.rst", "python_api/wenet.utils.ctc_utils.rst", "python_api/wenet.utils.executor.rst", "python_api/wenet.utils.file_utils.rst", "python_api/wenet.utils.init_ali_paraformer.rst", "python_api/wenet.utils.init_model.rst", "python_api/wenet.utils.mask.rst", "python_api/wenet.utils.scheduler.rst", "python_api/wenet.utils.tokenize_utils.rst", "python_api/wenet.utils.train_utils.rst", "python_api/wenet.whisper.rst", "python_api/wenet.whisper.convert_whisper_to_wenet_config_and_ckpt.rst", "python_api/wenet.whisper.whisper.rst", "python_package.md", "reference.rst", "requirements.txt", "runtime.md", "train.rst", "tutorial_aishell.md", "tutorial_librispeech.md"], "titles": ["Big Data Training: UIO", "Context Biasing", "Welcome to wenet\u2019s documentation!", "JIT in WeNet", "LM for WeNet", "Papers", "Pretrained Models in WeNet", "Pretrained Models in WeNet", "Production Runtime", "Python API Reference", "wenet package", "wenet.branchformer package", "wenet.branchformer.cgmlp module", "wenet.branchformer.encoder module", "wenet.branchformer.encoder_layer module", "wenet.cli package", "wenet.cli.hub module", "wenet.cli.model module", "wenet.cli.paraformer_model module", "wenet.cli.transcribe module", "wenet.dataset package", "wenet.dataset.dataset module", "wenet.dataset.kaldi_io module", "wenet.dataset.processor module", "wenet.dataset.wav_distortion module", "wenet.efficient_conformer package", "wenet.efficient_conformer.attention module", "wenet.efficient_conformer.convolution module", "wenet.efficient_conformer.encoder module", "wenet.efficient_conformer.encoder_layer module", "wenet.efficient_conformer.subsampling module", "wenet.k2 package", "wenet.k2.model module", "wenet.paraformer package", "wenet.paraformer.attention module", "wenet.paraformer.cif module", "wenet.paraformer.convert_paraformer_to_wenet_config module", "wenet.paraformer.layers module", "wenet.paraformer.paraformer module", "wenet.paraformer.search module", "wenet.squeezeformer package", "wenet.squeezeformer.attention module", "wenet.squeezeformer.conv2d module", "wenet.squeezeformer.convolution module", "wenet.squeezeformer.encoder module", "wenet.squeezeformer.encoder_layer module", "wenet.squeezeformer.positionwise_feed_forward module", "wenet.squeezeformer.subsampling module", "wenet.transducer package", "wenet.transducer.joint module", "wenet.transducer.predictor module", "wenet.transducer.transducer module", "wenet.transformer package", "wenet.transformer.asr_model module", "wenet.transformer.attention module", "wenet.transformer.cmvn module", "wenet.transformer.convolution module", "wenet.transformer.ctc module", "wenet.transformer.decoder module", "wenet.transformer.decoder_layer module", "wenet.transformer.embedding module", "wenet.transformer.encoder module", "wenet.transformer.encoder_layer module", "wenet.transformer.label_smoothing_loss module", "wenet.transformer.positionwise_feed_forward module", "wenet.transformer.search module", "wenet.transformer.subsampling module", "wenet.transformer.swish module", "wenet.utils package", "wenet.utils.checkpoint module", "wenet.utils.cmvn module", "wenet.utils.common module", "wenet.utils.config module", "wenet.utils.context_graph module", "wenet.utils.ctc_utils module", "wenet.utils.executor module", "wenet.utils.file_utils module", "wenet.utils.init_ali_paraformer module", "wenet.utils.init_model module", "wenet.utils.mask module", "wenet.utils.scheduler module", "wenet.utils.tokenize_utils module", "wenet.utils.train_utils module", "wenet.whisper package", "wenet.whisper.convert_whisper_to_wenet_config_and_ckpt module", "wenet.whisper.whisper module", "Python Package", "Reference", "&lt;no title&gt;", "Runtime for WeNet", "How to train models?", "Tutorial on AIShell", "Tutorial on LibriSpeech"], "terms": {"In": [0, 1, 4, 89, 91, 92], "order": [0, 1], "support": [0, 1, 4, 86, 92], "model": [0, 1, 2, 3, 4, 15, 31, 86, 89], "industri": [0, 4, 89], "ten": 0, "million": 0, "hour": [0, 91, 92], "speech": [0, 4, 5, 89, 91, 92], "dataset": [0, 4, 6, 7, 10, 91, 92], "process": [0, 1, 91, 92], "method": [0, 3, 89, 91, 92], "unifi": [0, 4, 89, 92], "ha": [0, 4, 6, 7, 89], "been": [0, 4, 6, 7], "updat": [0, 1, 3], "wenet": [0, 1, 5, 8, 9, 86, 90], "The": [0, 1, 4, 6, 7, 89, 91, 92], "document": 0, "introduc": 0, "from": [0, 1, 4, 89, 91, 92], "follow": [0, 1, 3, 4, 6, 7, 86, 89, 91, 92], "section": [0, 89], "old": [0, 92], "i": [0, 1, 2, 3, 4, 6, 7, 89, 91, 92], "base": [0, 1, 2, 4, 89, 91, 92], "pytorch": [0, 3, 6, 7, 89], "": [0, 1, 4, 6, 7, 89, 92], "nativ": 0, "dure": [0, 1, 92], "need": [0, 1, 4], "load": [0, 3], "all": [0, 4, 89], "audio": [0, 86, 91, 92], "path": [0, 1, 3, 91, 92], "correspondingli": 0, "label": [0, 1, 86], "memori": [0, 91, 92], "one": [0, 1, 4, 89, 91, 92], "time": [0, 1, 89, 91, 92], "randomli": 0, "read": 0, "case": [0, 89, 92], "grade": 0, "ultra": 0, "larg": [0, 1, 89, 91, 92], "scale": 0, "eg": 0, "more": [0, 5, 89, 91, 92], "than": [0, 89, 91, 92], "50": [0, 91, 92], "000": 0, "thi": [0, 1, 4, 91, 92], "caus": [0, 91, 92], "fail": [0, 1, 3, 89], "run": [0, 91, 92], "two": [0, 3, 4, 6, 7, 89, 91, 92], "reason": [0, 4, 91, 92], "out": [0, 4, 91, 92], "oom": 0, "physic": 0, "gener": [0, 4, 89], "machin": [0, 91, 92], "difficult": 0, "slow": 0, "down": 0, "perform": [0, 1], "cannot": 0, "us": [0, 1, 3, 6, 7, 86, 89], "file": [0, 1, 91, 92], "cach": 0, "speed": [0, 4], "greatli": 0, "reduc": [0, 6, 7], "inspir": 0, "webdataset": 0, "tfrecord": 0, "redesign": 0, "Its": 0, "core": [0, 4], "idea": 0, "make": [0, 89], "multipl": 0, "small": [0, 89], "1000": 0, "piec": [0, 92], "compress": 0, "packet": 0, "tar": 0, "them": [0, 92], "iterabledataset": 0, "advantag": [0, 89], "onli": [0, 1, 4, 89, 92], "index": [0, 2, 92], "inform": [0, 1, 86], "packag": [0, 2, 9], "maintain": 0, "which": [0, 1, 3, 4, 89, 91, 92], "save": [0, 3, 6, 7, 92], "solv": 0, "problem": [0, 1, 91, 92], "fly": [0, 1, 91, 92], "decompress": 0, "same": [0, 1, 4, 91, 92], "sequenc": [0, 4, 89, 91, 92], "random": 0, "differ": [0, 1, 4, 6, 7, 89], "can": [0, 1, 3, 4, 6, 7, 86, 89, 91, 92], "ensur": [0, 3], "global": [0, 91, 92], "new": [0, 1], "take": [0, 91, 92], "account": 0, "both": [0, 4, 89], "provid": [0, 6, 7, 91, 92], "we": [0, 1, 3, 4, 6, 7, 89, 91, 92], "call": [0, 1, 89, 91, 92], "shown": [0, 4, 92], "figur": 0, "below": [0, 4], "some": [0, 1, 91, 92], "necessari": [0, 1], "explan": 0, "about": [0, 91, 92], "abov": [0, 1], "mode": [0, 91, 92], "local": [0, 4, 91, 92], "requir": [0, 3, 4, 6, 7, 89, 91, 92], "must": [0, 91, 92], "sort": [0, 92], "kaldi": [0, 1, 4, 92], "style": [0, 4, 91, 92], "wav": [0, 1, 4, 86], "scp": [0, 4, 91, 92], "text": [0, 1, 4, 86, 91, 92], "It": [0, 1, 4, 89, 91, 92], "befor": [0, 1, 3, 92], "share": [0, 89, 91, 92], "network": [0, 6, 7], "cloud": 0, "storag": 0, "txt": [0, 1, 4, 91, 92], "ar": [0, 1, 4, 7, 86, 89, 91, 92], "store": 0, "singl": [0, 1], "also": [0, 1, 4, 89, 91, 92], "adopt": [0, 1, 89], "implement": 0, "practic": [0, 1, 89], "flexibl": 0, "easier": 0, "expand": 0, "debug": 0, "exampl": [0, 1, 4, 6, 7, 91, 92], "def": 0, "read_dataset": 0, "filenam": 0, "batch_siz": [0, 91, 92], "tf": 0, "tfrecorddataset": 0, "map": [0, 91, 92], "_parse_image_funct": 0, "num_parallel_cal": 0, "experiment": 0, "autotun": 0, "shuffl": 0, "500": [0, 92], "batch": 0, "drop_remaind": 0, "true": [0, 1, 91, 92], "repeat": 0, "prefetch": 0, "buffer_s": 0, "return": [0, 1], "refer": [0, 2, 3, 91, 92], "dataflow": 0, "includ": [0, 1, 4], "modul": [0, 2, 3, 11, 15, 20, 25, 31, 33, 40, 48, 52, 68, 83, 89], "token": [0, 1, 86], "convert": 0, "specifi": [0, 1, 86, 92], "unit": [0, 1, 4, 91, 92], "char": [0, 4, 91, 92], "bpe": [0, 4, 92], "filter": 0, "unqualifi": 0, "resampl": 0, "option": [0, 4], "comput": [0, 1, 89, 91, 92], "_": [0, 1, 4, 89, 91, 92], "fbank": 0, "featur": 0, "extract": 0, "spec": 0, "augment": 0, "enhanc": 0, "disrupt": 0, "organ": [0, 4, 91, 92], "pad": 0, "what": [0, 3, 4], "There": [0, 4], "sever": [0, 89, 91, 92], "paramet": [0, 86, 91, 92], "note": 0, "first": [0, 4, 89], "buffer": 0, "size": [0, 6, 7, 89, 91, 92], "recommend": 0, "larger": 0, "number": [0, 91, 92], "contain": [0, 1, 91, 92], "each": [0, 1, 91, 92], "equival": 0, "between": [0, 1, 91], "increas": 0, "you": [0, 3, 4, 6, 7, 86, 89, 91, 92], "set": [0, 89, 91, 92], "1500": 0, "accord": [0, 1, 4], "frame": [0, 1, 4, 89], "oper": 0, "veri": [0, 4, 89], "import": [0, 86], "improv": [0, 1], "dataload": [0, 91], "pre": 0, "granular": 0, "final": [0, 1, 4, 89, 92], "default": [0, 86], "2": [0, 1, 4, 5, 89], "due": 0, "exist": 0, "mai": [0, 1, 4, 91, 92], "alreadi": [0, 91, 92], "so": [0, 4, 89, 91, 92], "real": 0, "when": [0, 1, 3, 89, 91, 92], "insuffici": 0, "next": [0, 1], "fill": 0, "At": 0, "block": 0, "word": [0, 1, 4, 89, 91, 92], "part": [0, 4, 89, 92], "becaus": 0, "previou": [0, 89], "still": 0, "therefor": [0, 4, 89], "avoid": [0, 1, 89], "present": 0, "have": [0, 1, 4, 91, 92], "verifi": 0, "accuraci": [0, 1, 89], "200": 0, "10000": 0, "respect": 0, "cer": 0, "4": [0, 4, 6, 7], "61": 0, "63": 0, "67": 0, "espnet": [0, 92], "similar": 0, "structur": [0, 4, 7, 91, 92], "configur": 0, "thei": [0, 1, 91, 92], "achiev": [0, 89], "recognit": [0, 1, 4, 5, 89, 91, 92], "rate": [0, 89, 91, 92], "show": [0, 4, 86, 89, 91, 92], "correct": 0, "And": [0, 4, 6, 7], "observ": 0, "util": [0, 10, 92], "gpu": [0, 86, 91, 92], "80": 0, "90": 0, "indic": [0, 1, 91], "effici": 0, "high": 0, "For": [0, 4, 6, 7, 89, 91, 92], "detail": [0, 3, 4, 91, 92], "pleas": [0, 4, 6, 89, 91, 92], "http": [0, 3, 4, 6, 7, 86, 92], "github": [0, 4, 86, 91, 92], "com": [0, 4, 86], "e2": [0, 4, 86, 92], "blob": [0, 4], "main": [0, 4], "s0": [0, 91, 92], "sh": [0, 4, 91, 92], "prepar": [0, 4], "three": 0, "relat": [0, 4, 92], "script": [0, 3, 91, 92], "py": [0, 3, 4, 89, 91, 92], "train_data": [0, 92], "cv_data": [0, 92], "test_data": [0, 92], "list": [0, 1, 89, 91, 92], "data_typ": [0, 92], "symbol_t": [0, 92], "python": [0, 2, 3, 6, 7, 87, 89, 92], "bin": [0, 3, 91, 92], "gpu_id": [0, 92], "config": [0, 68, 91, 92], "train_config": [0, 92], "dict": [0, 4, 91, 92], "feat_dir": 0, "train_set": [0, 92], "dev": [0, 92], "If": [0, 1, 3, 91, 92], "type": [0, 3, 89, 91, 92], "format": [0, 4], "kei": [0, 91, 92], "bac009s0002w0122": [0, 91], "export": [0, 1, 3, 6, 7, 89], "asr": [0, 1, 2, 89, 91, 92], "openslr": [0, 6, 7, 91, 92], "33": [0, 91], "data_aishel": [0, 4, 91], "s0002": [0, 91], "\u800c\u5bf9\u697c\u5e02\u6210\u4ea4\u6291\u5236\u4f5c\u7528\u6700\u5927\u7684\u9650\u8d2d": [0, 91], "bac009s0002w0123": [0, 91], "\u4e5f\u6210\u4e3a\u5730\u65b9\u653f\u5e9c\u7684\u773c\u4e2d\u9489": [0, 91], "bac009s0002w0124": [0, 91], "\u81ea\u516d\u6708\u5e95\u547c\u548c\u6d69\u7279\u5e02\u7387\u5148\u5ba3\u5e03\u53d6\u6d88\u9650\u8d2d\u540e": [0, 91], "line": [0, 1, 2, 91, 92], "json": [0, 91, 92], "serial": 0, "string": 0, "field": [0, 91, 92], "1": [0, 1, 4, 86, 89], "maryland": 0, "binbinzhang": 0, "code": [0, 1, 3, 4, 6, 7, 89, 92], "s3": 0, "raw_wav": [0, 91], "shards_000000000": 0, "gz": [0, 92], "shards_000000001": 0, "shards_000000002": 0, "oss": 0, "examplebucket": 0, "cn": [0, 6, 7], "hangzhou": 0, "aliyunc": 0, "exampledir": 0, "q1": 0, "how": [0, 1, 2, 3, 89, 91, 92], "distribut": 0, "partit": 0, "rank": [0, 92], "num": [0, 92], "worker": 0, "segment": 0, "class": [0, 4], "distributedsampl": 0, "__init__": 0, "self": 0, "epoch": [0, 91, 92], "set_epoch": 0, "sampl": 0, "rang": [0, 4], "len": 0, "world_siz": [0, 92], "worker_id": 0, "num_work": [0, 92], "q2": 0, "deal": 0, "unbalanc": 0, "join": [0, 3], "handl": [0, 89], "imbal": 0, "alloc": 0, "applic": [1, 4], "effect": 1, "commonli": 1, "better": [1, 89, 91, 92], "uniqu": 1, "low": 1, "contextu": 1, "inject": 1, "prior": 1, "knowledg": 1, "an": [1, 2, 4, 89, 91, 92], "system": [1, 89, 92], "infer": [1, 89, 91, 92], "user": [1, 6, 7], "favorit": 1, "song": 1, "contact": [1, 4], "app": 1, "locat": 1, "convent": 1, "build": [1, 4, 92], "n": [1, 4, 89, 91, 92], "gram": [1, 4, 92], "finit": 1, "state": 1, "transduc": [1, 10], "fst": [1, 4, 92], "phrase": 1, "compos": [1, 4], "decod": [1, 4, 52, 89, 91, 92], "help": 1, "bia": 1, "result": [1, 6, 7, 86, 89, 91, 92], "toward": 1, "thu": 1, "certain": 1, "scenario": [1, 4, 89], "score": [1, 89], "p": [1, 4, 92], "c": [1, 3, 89, 91, 92], "mathbf": 1, "y": [1, 89], "interpol": 1, "x": [1, 89, 92], "shallow": 1, "fusion": 1, "mathrm": 1, "arg": [1, 3, 89], "max": 1, "log": [1, 4, 89], "lambda": 1, "where": [1, 89], "tunabl": [1, 91, 92], "hyperparamet": 1, "control": 1, "much": [1, 3, 89], "lm": [1, 2, 8], "influenc": 1, "overal": [1, 89], "want": [1, 3, 91, 92], "cat": [1, 92], "charact": [1, 91], "0": [1, 4, 5, 6, 7], "25": 1, "construct": 1, "correspond": [1, 4, 6, 7], "match": 1, "reward": 1, "obtain": 1, "artifici": 1, "boost": [1, 91, 92], "earli": 1, "do": [1, 3, 4, 91, 92], "entir": 1, "add": 1, "special": 1, "failur": 1, "arc": 1, "remov": 1, "record": [1, 91, 92], "easili": [1, 89, 91, 92], "determin": 1, "boundari": 1, "hot": 1, "That": 1, "after": [1, 91, 92], "succe": 1, "other": [1, 4, 89, 91, 92], "start": [1, 91, 92], "int": 1, "contextgraph": 1, "getnextst": 1, "cur_stat": 1, "word_id": 1, "float": 1, "bool": 1, "is_start_boundari": 1, "is_end_boundari": 1, "next_stat": 1, "travers": 1, "current": [1, 89], "arciter": 1, "stdfst": 1, "aiter": 1, "graph_": 1, "done": [1, 92], "const": 1, "stdarc": 1, "valu": [1, 86], "ilabel": 1, "backoff": 1, "might": 1, "cover": 1, "weight": 1, "els": 1, "nextstat": 1, "check": [1, 3, 91, 92], "whether": [1, 4], "One": 1, "break": 1, "append": 1, "output": [1, 89, 91, 92], "chang": [1, 4, 89, 91, 92], "function": [1, 3, 4, 91, 92], "end": [1, 2, 4, 5, 89, 91, 92], "posit": 1, "insert": 1, "tag": 1, "bag": 1, "lattic": 1, "faster": [1, 89], "onlin": 1, "modifi": [1, 3, 4], "cc": [1, 4, 6, 7, 89], "tlg": [1, 4, 92], "input": [1, 86, 89, 91, 92], "decid": 1, "processemit": 1, "processnonemit": 1, "elem": 1, "e_next": 1, "findoraddtoken": 1, "tot_cost": 1, "tok": 1, "null": 1, "begin": [1, 92], "fals": [1, 91, 92], "context_scor": 1, "context_graph_": 1, "olabel": 1, "val": 1, "context_st": 1, "graph_cost": 1, "forwardlink": 1, "next_tok": 1, "put": [1, 4], "head": 1, "link": [1, 7], "forwardlinkt": 1, "ac_cost": 1, "accumul": [1, 4], "lead": 1, "cost": [1, 89, 91, 92], "too": 1, "void": 1, "latticefasterdecodertpl": 1, "pruneforwardlink": 1, "int32": 1, "frame_plus_on": 1, "extra_costs_chang": 1, "links_prun": 1, "basefloat": 1, "delta": 1, "link_extra_cost": 1, "extra_cost": 1, "acoustic_cost": 1, "bracket": 1, "mean": [1, 91, 92], "link_exta_cost": 1, "best": [1, 4, 86, 89, 91, 92], "through": [1, 91, 92], "sourc": 1, "destin": 1, "context_path": 1, "split": [1, 4], "symbol": [1, 89, 91, 92], "tabl": [1, 88], "oov": 1, "cd": [1, 86, 91, 92], "home": [1, 91], "runtim": [1, 2, 4, 6, 7], "libtorch": [1, 89, 91, 92], "glog_logtostderr": 1, "glog_v": 1, "wav_path": [1, 91, 92], "docker_resourc": 1, "test": [1, 4], "model_dir": [1, 3, 86, 92], "decoder_main": 1, "chunk_siz": 1, "model_path": 1, "zip": [1, 3, 4, 6, 7, 92], "3": [1, 4, 89], "unit_path": 1, "tee": 1, "tansform": 2, "toolkit": [2, 5], "instal": [2, 92], "develop": [2, 89], "command": 2, "usag": 2, "program": [2, 91], "train": [2, 3, 4, 6, 7, 89], "tutori": [2, 90], "librispeech": [2, 4, 6, 7, 90], "aishel": [2, 4, 6, 7, 90], "pretrain": [2, 90, 92], "big": [2, 90, 91, 92], "data": [2, 4, 90], "uio": [2, 90], "product": [2, 3, 4, 5], "context": [2, 8, 89], "bias": [2, 8], "jit": [2, 6, 7, 8, 89], "paper": [2, 87, 91, 92], "api": [2, 87], "search": [2, 4, 33, 52, 89, 91, 92], "page": 2, "our": [3, 6, 7, 89, 92], "directli": [3, 4, 6, 7, 89], "torch": [3, 88, 89, 91, 92], "essenti": 3, "deploi": 3, "see": [3, 4, 6, 7, 89, 91, 92], "resourc": [3, 92], "introduct": [3, 89], "TO": [3, 92], "torchscript": 3, "languag": [3, 4, 6, 7, 86, 91, 92], "A": [3, 4, 92], "IN": [3, 92], "deep": 3, "dive": 3, "research": [3, 4], "To": [3, 4], "try": [3, 91, 92], "stage": 3, "should": [3, 89], "satisfi": 3, "script_model": 3, "o": 3, "init": [3, 92], "principl": [3, 4], "taken": 3, "consider": 3, "contribut": 3, "especi": [3, 4], "subclass": 3, "nn": [3, 91, 92], "forward": [3, 89], "know": 3, "allow": 3, "disallow": 3, "tensor": 3, "unsupport": 3, "attribut": 3, "coverag": 3, "explicit": 3, "possibl": [3, 92], "forc": [3, 86], "typeguard": [3, 88], "readthedoc": 3, "io": 3, "en": [3, 6, 7, 92], "latest": 3, "userguid": 3, "html": 3, "statist": [4, 91, 92], "wfst": 4, "framework": [4, 89], "custom": [4, 91, 92], "why": 4, "question": 4, "mani": 4, "peopl": [4, 91, 92], "ask": [4, 91, 92], "now": [4, 86], "rnn": 4, "transform": [4, 10, 89, 91, 92], "full": [4, 92], "swing": 4, "doe": [4, 91, 92], "go": [4, 91, 92], "backward": 4, "simpl": [4, 89, 91, 92], "matur": 4, "complet": 4, "tool": [4, 91, 92], "ani": [4, 91, 92], "amount": 4, "corpu": [4, 92], "fast": [4, 89], "hotfix": 4, "easi": 4, "wide": 4, "actual": [4, 91, 92], "second": 4, "sinc": [4, 6, 7, 89], "work": [4, 89, 91, 92], "hard": 4, "abandon": 4, "tradit": [4, 89], "complex": 4, "technologi": 4, "back": 4, "standard": [4, 89], "power": 4, "solut": 4, "bug": 4, "fix": 4, "replac": 4, "person": 4, "just": [4, 89, 91, 92], "like": [4, 91, 92], "goal": 4, "readi": 4, "prioriti": 4, "draw": 4, "pure": 4, "As": 4, "describ": [4, 89], "choos": 4, "your": [4, 86, 91, 92], "whole": [4, 89, 91, 92], "bellow": 4, "pictur": [4, 89], "wai": [4, 89], "without": [4, 91], "ctc": [4, 52, 89, 91, 92], "prefix": [4, 89, 91, 92], "beam": [4, 89, 91, 92], "With": [4, 89], "graph": 4, "t": [4, 86, 89], "lexicon": [4, 92], "l": [4, 86, 92], "g": [4, 86, 91, 92], "typic": [4, 89, 92], "chines": [4, 6, 86], "english": [4, 86, 92], "its": 4, "\u6211\u4eec": 4, "\u6211": 4, "\u4eec": 4, "appl": 4, "five": 4, "letter": 4, "e": [4, 86, 89, 91, 92], "phonem": 4, "pronunci": 4, "purpos": 4, "name": [4, 91, 92], "compil": 4, "represent": 4, "viterbi": 4, "algorithm": [4, 91, 92], "eas": 4, "keep": 4, "independ": 4, "migrat": 4, "directori": [4, 92], "minim": 4, "remain": 4, "origin": [4, 91, 92], "glog": 4, "meet": [4, 91, 92], "lint": 4, "wrap": 4, "latticefasterdecod": 4, "blank": [4, 91, 92], "skip": [4, 91, 92], "up": 4, "addit": [4, 89], "arpa2fst": 4, "fstdeterminizestar": 4, "fsttablecompos": 4, "fstminimizeencod": 4, "built": 4, "box": 4, "get": [4, 89, 91, 92], "consist": 4, "gain": 4, "10": [4, 92], "aishell2": [4, 6, 7], "here": [4, 6, 7, 89, 91, 92], "dictionari": 4, "7": [4, 91], "unit_fil": [4, 92], "mkdir": [4, 92], "cp": [4, 92], "prepare_dict": [4, 92], "resource_aishel": 4, "filter_scp": 4, "pl": 4, "transcript": [4, 86, 91, 92], "aishell_transcript_v0": 4, "8": [4, 91, 92], "aishell_train_lm": 4, "compile_lexicon_token_fst": [4, 92], "tmp": [4, 92], "lang": [4, 92], "make_tlg": [4, 92], "lang_test": [4, 92], "exit": [4, 92], "nj": [4, 92], "16": [4, 92], "15": 4, "lattice_beam": [4, 92], "5": 4, "max_act": [4, 92], "7000": [4, 92], "blank_skip_thresh": [4, 92], "98": [4, 92], "ctc_weight": [4, 92], "rescoring_weight": [4, 92], "fst_path": [4, 92], "dict_path": [4, 92], "dir": [4, 86, 91, 92], "lm_with_runtim": 4, "orient": 5, "stream": [5, 89], "non": [5, 6, 89, 92], "accept": 5, "interspeech": 5, "2021": 5, "2022": 5, "facilit": [6, 7], "checkpoint": [6, 7, 68, 91, 92], "suffix": [6, 7], "pt": [6, 7, 91, 92], "reproduc": [6, 7], "publish": [6, 7], "continu": [6, 7, 89, 91, 92], "x86": [6, 7, 89], "android": [6, 7, 89], "quantiz": [6, 7], "traffic": [6, 7], "BY": [6, 7], "org": [6, 7, 92], "12": [6, 7, 92], "visit": 6, "download": [6, 7], "contributor": [6, 7], "conform": [6, 7, 89, 91, 92], "gigaspeech": [6, 7, 91, 92], "multi_cn": [6, 7], "wenetspeech": [6, 7, 91, 92], "given": [7, 92], "subpackag": 9, "branchform": 10, "submodul": 10, "cli": 10, "efficient_conform": 10, "k2": 10, "paraform": [10, 86], "squeezeform": 10, "whisper": 10, "cgmlp": 11, "encod": [11, 25, 40, 52, 89, 91, 92], "encoder_lay": [11, 25, 40, 52], "hub": 15, "paraformer_model": 15, "transcrib": [15, 86], "kaldi_io": 20, "processor": 20, "wav_distort": 20, "attent": [25, 33, 40, 52, 89, 91, 92], "convolut": [25, 40, 52], "subsampl": [25, 40, 52, 89], "cif": 33, "convert_paraformer_to_wenet_config": 33, "layer": [33, 89], "conv2d": 40, "positionwise_feed_forward": [40, 52], "joint": [48, 89], "predictor": 48, "asr_model": 52, "cmvn": [52, 68], "decoder_lay": 52, "embed": 52, "label_smoothing_loss": 52, "swish": 52, "common": 68, "context_graph": 68, "ctc_util": 68, "executor": 68, "file_util": 68, "init_ali_paraform": 68, "init_model": 68, "mask": 68, "schedul": 68, "tokenize_util": 68, "train_util": 68, "convert_whisper_to_wenet_config_and_ckpt": 83, "pip": 86, "git": 86, "clone": 86, "m": 86, "own": 86, "devic": [86, 89, 91, 92], "id": [86, 89, 91, 92], "repres": 86, "cpu": 86, "show_tokens_info": 86, "level": [86, 89], "timestamp": 86, "confid": 86, "etc": 86, "align": 86, "load_model": 86, "xxx": 86, "print": [86, 92], "jinja2": 88, "nbsphinx": 88, "sphinx": 88, "recommonmark": 88, "markdown": 88, "rtd": 88, "theme": 88, "torchaudio": [88, 91], "pass": 89, "u2": [89, 91, 92], "balanc": 89, "latenc": 89, "chunk": [89, 92], "accur": 89, "rescor": [89, 91, 92], "autoregress": 89, "benefit": 89, "find": [89, 91, 92], "stabl": 89, "long": 89, "form": 89, "usual": 89, "degrad": 89, "lot": 89, "stamp": 89, "spike": 89, "aspect": 89, "favor": 89, "server": 89, "finish": [89, 91, 92], "fed": 89, "activ": 89, "linear": 89, "softmax": 89, "prob": 89, "further": 89, "partial": 89, "everi": 89, "select": 89, "group": 89, "affect": 89, "infinit": 89, "give": 89, "limit": 89, "less": [89, 91, 92], "promis": 89, "proper": 89, "brief": 89, "descript": 89, "subsampling_r": 89, "right_context": 89, "right": 89, "sos_symbol": 89, "eos_symbol": 89, "eo": [89, 91, 92], "forward_encoder_chunk": 89, "ctc_activ": 89, "forward_attention_decod": 89, "increment": 89, "requri": 89, "histori": 89, "att": 89, "cnn": 89, "left": 89, "causal": 89, "howev": 89, "stride": 89, "tircki": 89, "simpli": 89, "overlap": 89, "straightforward": 89, "neglig": 89, "fraction": 89, "blue": 89, "color": 89, "vocabulari": [89, 91, 92], "bi": 89, "direct": 89, "recurr": 89, "dnn": 89, "feel": [91, 92], "free": [91, 92], "issu": [91, 92], "thank": [91, 92], "kind": [91, 92], "feedback": [91, 92], "recip": [91, 92], "suggest": [91, 92], "manual": [91, 92], "understand": [91, 92], "bash": [91, 92], "stop_stag": [91, 92], "could": [91, 92], "variabl": [91, 92], "absolut": 91, "usernam": 91, "aishell_data_prep": 91, "tab": [91, 92], "separ": [91, 92], "column": [91, 92], "wav_id": [91, 92], "text_label": [91, 92], "bac009s0002w0125": 91, "\u5404\u5730\u653f\u5e9c\u4fbf\u7eb7\u7eb7\u8ddf\u8fdb": 91, "raw": [91, 92], "step": [91, 92], "copi": 91, "compute_cmvn_stat": [91, 92], "cepstral": [91, 92], "varianc": [91, 92], "normal": [91, 92], "These": [91, 92], "acoust": [91, 92], "integ": [91, 92], "unk": [91, 92], "\u4e00": 91, "\u4e01": 91, "\u9f9a": 91, "4230": 91, "\u9f9f": 91, "4231": 91, "4232": 91, "denot": [91, 92], "unknown": [91, 92], "utter": [91, 92], "aslo": [91, 92], "design": [91, 92], "anoth": [91, 92], "shard": [91, 92], "10k": [91, 92], "appli": [91, 92], "5k": [91, 92], "multi": [91, 92], "ddp": [91, 92], "dist_backend": [91, 92], "nccl": [91, 92], "gloo": [91, 92], "cuda": [91, 92], "visibl": [91, 92], "cuda_visible_devic": [91, 92], "card": [91, 92], "resum": [91, 92], "termin": [91, 92], "accident": [91, 92], "exp": [91, 92], "your_exp": [91, 92], "Then": [91, 92], "optim": [91, 92], "loss": [91, 92], "yaml": [91, 92], "conf": [91, 92], "train_conform": [91, 92], "tensorboard": [91, 92], "depend": [91, 92], "2080": [91, 92], "ti": [91, 92], "dai": [91, 92], "monitor": [91, 92], "logdir": [91, 92], "your_exp_nam": [91, 92], "port": [91, 92], "12598": [91, 92], "bind_al": [91, 92], "averag": [91, 92], "average_checkpoint": [91, 92], "average_num": [91, 92], "cross": [91, 92], "valid": [91, 92], "four": [91, 92], "ctc_greedy_search": [91, 92], "greedi": [91, 92], "ctc_prefix_beam_search": [91, 92], "attention_rescor": [91, 92], "candid": [91, 92], "beam_siz": [91, 92], "higher": [91, 92], "greater": [91, 92], "wer": [91, 92], "evalu": [91, 92], "calcul": [91, 92], "error": [91, 92], "export_jit": [91, 92], "data_url": 92, "www": 92, "datadir": 92, "le": 92, "ge": 92, "echo": 92, "clean": 92, "100": 92, "360": 92, "download_and_untar": 92, "fi": 92, "wave_data": 92, "underscor": 92, "data_prep_torchaudio": 92, "1867": 92, "154075": 92, "0014": 92, "flac": 92, "1970": 92, "26100": 92, "0022": 92, "him": 92, "THAT": 92, "IT": 92, "did": 92, "AT": 92, "task": 92, "yourself": 92, "But": 92, "most": 92, "train_960": 92, "merg": 92, "total": 92, "train_clean_100": 92, "train_clean_360": 92, "train_other_500": 92, "f": 92, "dev_clean": 92, "dev_oth": 92, "in_scp": 92, "out_cmvn": 92, "global_cmvn": 92, "subset": 92, "lang_char": 92, "bpemod": 92, "nbpe": 92, "_unit": 92, "bpemodel": 92, "linguist": 92, "borrow": 92, "cut": 92, "d": 92, "spm_train": 92, "vocab_s": 92, "model_typ": 92, "model_prefix": 92, "input_sentence_s": 92, "100000000": 92, "spm_encod": 92, "output_format": 92, "tr": 92, "uniq": 92, "awk": 92, "nr": 92, "num_token": 92, "wc": 92, "byte": 92, "pair": 92, "sentencepiec": 92, "defin": 92, "ab": 92, "4995": 92, "young": 92, "4996": 92, "4997": 92, "youth": 92, "4998": 92, "z": 92, "4999": 92, "zz": 92, "5000": 92, "5001": 92, "recog_set": 92, "make_raw_list": 92, "1455": 92, "134435": 92, "0000": 92, "mnt": 92, "nf": 92, "ptm1": 92, "open": 92, "THE": 92, "girl": 92, "who": 92, "came": 92, "INTO": 92, "world": 92, "ON": 92, "night": 92, "jess": 92, "ran": 92, "cry": 92, "god": 92, "he": 92, "BE": 92, "son": 92, "had": 92, "grown": 92, "womanhood": 92, "farm": 92, "0001": 92, "AND": 92, "NOT": 92, "angri": 92, "she": 92, "wa": 92, "often": 92, "moros": 92, "silent": 92, "winesburg": 92, "said": 92, "drank": 92, "her": 92, "husband": 92, "banker": 92, "0002": 92, "BUT": 92, "louis": 92, "made": 92, "happi": 92, "flew": 92, "half": 92, "insan": 92, "fit": 92, "OF": 92, "temper": 92, "sometim": 92, "noisi": 92, "quarrelsom": 92, "swore": 92, "cri": 92, "anger": 92, "got": 92, "knife": 92, "kitchen": 92, "threaten": 92, "life": 92, "init_fil": 92, "ddp_init": 92, "rm": 92, "delet": 92, "init_method": 92, "readlink": 92, "num_gpu": 92, "otherwis": 92, "cmvn_opt": 92, "write": 92, "dimens": 92, "later": 92, "pin_memori": 92, "wait": 92, "todo": 92, "decode_checkpoint": 92, "avg_": 92, "average_model": 92, "dst_model": 92, "src_path": 92, "val_best": 92, "decoding_chunk_s": 92, "dynam": 92, "poll": 92, "idx": 92, "decode_mod": 92, "test_dir": 92, "penalti": 92, "result_fil": 92, "text_bp": 92, "spm_decod": 92, "input_format": 92, "sed": 92, "v": 92, "eq": 92, "output_fil": 92, "integr": 92, "which_lm": 92, "prune": 92, "1e": 92, "arpa": 92, "wget": 92, "11": 92, "unzip": 92, "gunzip": 92, "k": 92, "train_960_unigram5000": 92, "fst_dir": 92, "acoustic_scal": 92, "lm_with_runtime_": 92, "tail": 92}, "objects": {"wenet": [[15, 0, 0, "-", "cli"]]}, "objtypes": {"0": "py:module"}, "objnames": {"0": ["py", "module", "Python module"]}, "titleterms": {"big": 0, "data": [0, 91, 92], "train": [0, 90, 91, 92], "uio": 0, "necess": 0, "upgrad": 0, "io": 0, "mothod": 0, "system": [0, 4], "design": [0, 4, 89], "overal": 0, "chain": 0, "valid": 0, "experi": [0, 91, 92], "aishel": [0, 91], "raw": 0, "v": 0, "shard": 0, "wenetspeech": 0, "usag": [0, 1, 86], "q": 0, "A": 0, "context": 1, "bias": 1, "graph": 1, "ctc": [1, 57], "prefix": 1, "beam": 1, "search": [1, 39, 65], "wfst": 1, "prune": 1, "welcom": 2, "wenet": [2, 3, 4, 6, 7, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 89, 91, 92], "": 2, "document": 2, "content": [2, 8, 87, 90], "indic": 2, "tabl": 2, "jit": 3, "lm": [4, 92], "motiv": 4, "implement": [4, 89], "result": 4, "how": [4, 90], "us": [4, 91, 92], "paper": 5, "pretrain": [6, 7], "model": [6, 7, 17, 32, 90, 91, 92], "type": [6, 7], "licens": [6, 7], "list": [6, 7], "product": 8, "runtim": [8, 89, 92], "python": [9, 86], "api": 9, "refer": [9, 87, 89], "packag": [10, 11, 15, 20, 25, 31, 33, 40, 48, 52, 68, 83, 86], "subpackag": 10, "branchform": [11, 12, 13, 14], "submodul": [11, 15, 20, 25, 31, 33, 40, 48, 52, 68, 83], "cgmlp": 12, "modul": [12, 13, 14, 16, 17, 18, 19, 21, 22, 23, 24, 26, 27, 28, 29, 30, 32, 34, 35, 36, 37, 38, 39, 41, 42, 43, 44, 45, 46, 47, 49, 50, 51, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 84, 85], "encod": [13, 28, 44, 61], "encoder_lay": [14, 29, 45, 62], "cli": [15, 16, 17, 18, 19], "hub": 16, "paraformer_model": 18, "transcrib": 19, "dataset": [20, 21, 22, 23, 24], "kaldi_io": 22, "processor": 23, "wav_distort": 24, "efficient_conform": [25, 26, 27, 28, 29, 30], "attent": [26, 34, 41, 54], "convolut": [27, 43, 56], "subsampl": [30, 47, 66], "k2": [31, 32], "paraform": [33, 34, 35, 36, 37, 38, 39], "cif": 35, "convert_paraformer_to_wenet_config": 36, "layer": 37, "squeezeform": [40, 41, 42, 43, 44, 45, 46, 47], "conv2d": 42, "positionwise_feed_forward": [46, 64], "transduc": [48, 49, 50, 51], "joint": 49, "predictor": 50, "transform": [52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67], "asr_model": 53, "cmvn": [55, 70, 91, 92], "decod": 58, "decoder_lay": 59, "embed": 60, "label_smoothing_loss": 63, "swish": 67, "util": [68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82], "checkpoint": 69, "common": 71, "config": 72, "context_graph": 73, "ctc_util": 74, "executor": 75, "file_util": 76, "init_ali_paraform": 77, "init_model": 78, "mask": 79, "schedul": 80, "tokenize_util": 81, "train_util": 82, "whisper": [83, 84, 85], "convert_whisper_to_wenet_config_and_ckpt": 84, "instal": 86, "develop": 86, "command": 86, "line": 86, "program": 86, "platform": 89, "support": 89, "architectur": 89, "interfac": 89, "cach": 89, "detail": 89, "tutori": [91, 92], "first": [91, 92], "stage": [91, 92], "1": [91, 92], "download": [91, 92], "0": [91, 92], "prepar": [91, 92], "extract": [91, 92], "optin": [91, 92], "featur": [91, 92], "2": [91, 92], "gener": [91, 92], "label": [91, 92], "token": [91, 92], "dictionari": [91, 92], "3": [91, 92], "format": [91, 92], "4": [91, 92], "neural": [91, 92], "network": [91, 92], "5": [91, 92], "recogn": [91, 92], "wav": [91, 92], "6": [91, 92], "export": [91, 92], "librispeech": 92, "option": 92, "7": 92, "add": 92, "test": 92}, "envversion": {"sphinx.domains.c": 3, "sphinx.domains.changeset": 1, "sphinx.domains.citation": 1, "sphinx.domains.cpp": 9, "sphinx.domains.index": 1, "sphinx.domains.javascript": 3, "sphinx.domains.math": 2, "sphinx.domains.python": 4, "sphinx.domains.rst": 2, "sphinx.domains.std": 2, "nbsphinx": 4, "sphinx.ext.viewcode": 1, "sphinx.ext.todo": 2, "sphinx": 58}, "alltitles": {"Big Data Training: UIO": [[0, "big-data-training-uio"]], "Necessity of upgrading IO mothod": [[0, "necessity-of-upgrading-io-mothod"]], "System design of UIO": [[0, "system-design-of-uio"]], "Overall design": [[0, "overall-design"]], "Chain IO": [[0, "chain-io"]], "Validation experiments": [[0, "validation-experiments"]], "Aishell(raw vs shard)": [[0, "aishell-raw-vs-shard"]], "WenetSpeech(shard)": [[0, "wenetspeech-shard"]], "Usage of UIO": [[0, "usage-of-uio"]], "Q&A": [[0, "q-a"]], "Context Biasing": [[1, "context-biasing"]], "Context Graph": [[1, "context-graph"]], "CTC Prefix Beam Search": [[1, "ctc-prefix-beam-search"]], "CTC WFST Beam Search": [[1, "ctc-wfst-beam-search"]], "Pruning": [[1, "pruning"]], "Usage": [[1, "usage"]], "Welcome to wenet\u2019s documentation!": [[2, "welcome-to-wenet-s-documentation"]], "Contents:": [[2, null], [8, null], [87, null], [90, null]], "Indices and tables": [[2, "indices-and-tables"]], "JIT in WeNet": [[3, "jit-in-wenet"]], "LM for WeNet": [[4, "lm-for-wenet"]], "Motivation": [[4, "motivation"]], "System Design": [[4, "system-design"]], "Implementation": [[4, "implementation"]], "Results": [[4, "results"]], "How to use?": [[4, "how-to-use"]], "Papers": [[5, "papers"]], "Pretrained Models in WeNet": [[6, "pretrained-models-in-wenet"], [7, "pretrained-models-in-wenet"]], "Model Types": [[6, "model-types"], [7, "model-types"]], "Model License": [[6, "model-license"], [7, "model-license"]], "Model List": [[6, "model-list"], [7, "model-list"]], "Production Runtime": [[8, "production-runtime"]], "Python API Reference": [[9, "python-api-reference"]], "wenet package": [[10, "wenet-package"]], "Subpackages": [[10, "subpackages"]], "wenet.branchformer package": [[11, "wenet-branchformer-package"]], "Submodules": [[11, "submodules"], [15, "submodules"], [20, "submodules"], [25, "submodules"], [31, "submodules"], [33, "submodules"], [40, "submodules"], [48, "submodules"], [52, "submodules"], [68, "submodules"], [83, "submodules"]], "wenet.branchformer.cgmlp module": [[12, "wenet-branchformer-cgmlp-module"]], "wenet.branchformer.encoder module": [[13, "wenet-branchformer-encoder-module"]], "wenet.branchformer.encoder_layer module": [[14, "wenet-branchformer-encoder-layer-module"]], "wenet.cli package": [[15, "module-wenet.cli"]], "wenet.cli.hub module": [[16, "wenet-cli-hub-module"]], "wenet.cli.model module": [[17, "wenet-cli-model-module"]], "wenet.cli.paraformer_model module": [[18, "wenet-cli-paraformer-model-module"]], "wenet.cli.transcribe module": [[19, "wenet-cli-transcribe-module"]], "wenet.dataset package": [[20, "wenet-dataset-package"]], "wenet.dataset.dataset module": [[21, "wenet-dataset-dataset-module"]], "wenet.dataset.kaldi_io module": [[22, "wenet-dataset-kaldi-io-module"]], "wenet.dataset.processor module": [[23, "wenet-dataset-processor-module"]], "wenet.dataset.wav_distortion module": [[24, "wenet-dataset-wav-distortion-module"]], "wenet.efficient_conformer package": [[25, "wenet-efficient-conformer-package"]], "wenet.efficient_conformer.attention module": [[26, "wenet-efficient-conformer-attention-module"]], "wenet.efficient_conformer.convolution module": [[27, "wenet-efficient-conformer-convolution-module"]], "wenet.efficient_conformer.encoder module": [[28, "wenet-efficient-conformer-encoder-module"]], "wenet.efficient_conformer.encoder_layer module": [[29, "wenet-efficient-conformer-encoder-layer-module"]], "wenet.efficient_conformer.subsampling module": [[30, "wenet-efficient-conformer-subsampling-module"]], "wenet.k2 package": [[31, "wenet-k2-package"]], "wenet.k2.model module": [[32, "wenet-k2-model-module"]], "wenet.paraformer package": [[33, "wenet-paraformer-package"]], "wenet.paraformer.attention module": [[34, "wenet-paraformer-attention-module"]], "wenet.paraformer.cif module": [[35, "wenet-paraformer-cif-module"]], "wenet.paraformer.convert_paraformer_to_wenet_config module": [[36, "wenet-paraformer-convert-paraformer-to-wenet-config-module"]], "wenet.paraformer.layers module": [[37, "wenet-paraformer-layers-module"]], "wenet.paraformer.paraformer module": [[38, "wenet-paraformer-paraformer-module"]], "wenet.paraformer.search module": [[39, "wenet-paraformer-search-module"]], "wenet.squeezeformer package": [[40, "wenet-squeezeformer-package"]], "wenet.squeezeformer.attention module": [[41, "wenet-squeezeformer-attention-module"]], "wenet.squeezeformer.conv2d module": [[42, "wenet-squeezeformer-conv2d-module"]], "wenet.squeezeformer.convolution module": [[43, "wenet-squeezeformer-convolution-module"]], "wenet.squeezeformer.encoder module": [[44, "wenet-squeezeformer-encoder-module"]], "wenet.squeezeformer.encoder_layer module": [[45, "wenet-squeezeformer-encoder-layer-module"]], "wenet.squeezeformer.positionwise_feed_forward module": [[46, "wenet-squeezeformer-positionwise-feed-forward-module"]], "wenet.squeezeformer.subsampling module": [[47, "wenet-squeezeformer-subsampling-module"]], "wenet.transducer package": [[48, "wenet-transducer-package"]], "wenet.transducer.joint module": [[49, "wenet-transducer-joint-module"]], "wenet.transducer.predictor module": [[50, "wenet-transducer-predictor-module"]], "wenet.transducer.transducer module": [[51, "wenet-transducer-transducer-module"]], "wenet.transformer package": [[52, "wenet-transformer-package"]], "wenet.transformer.asr_model module": [[53, "wenet-transformer-asr-model-module"]], "wenet.transformer.attention module": [[54, "wenet-transformer-attention-module"]], "wenet.transformer.cmvn module": [[55, "wenet-transformer-cmvn-module"]], "wenet.transformer.convolution module": [[56, "wenet-transformer-convolution-module"]], "wenet.transformer.ctc module": [[57, "wenet-transformer-ctc-module"]], "wenet.transformer.decoder module": [[58, "wenet-transformer-decoder-module"]], "wenet.transformer.decoder_layer module": [[59, "wenet-transformer-decoder-layer-module"]], "wenet.transformer.embedding module": [[60, "wenet-transformer-embedding-module"]], "wenet.transformer.encoder module": [[61, "wenet-transformer-encoder-module"]], "wenet.transformer.encoder_layer module": [[62, "wenet-transformer-encoder-layer-module"]], "wenet.transformer.label_smoothing_loss module": [[63, "wenet-transformer-label-smoothing-loss-module"]], "wenet.transformer.positionwise_feed_forward module": [[64, "wenet-transformer-positionwise-feed-forward-module"]], "wenet.transformer.search module": [[65, "wenet-transformer-search-module"]], "wenet.transformer.subsampling module": [[66, "wenet-transformer-subsampling-module"]], "wenet.transformer.swish module": [[67, "wenet-transformer-swish-module"]], "wenet.utils package": [[68, "wenet-utils-package"]], "wenet.utils.checkpoint module": [[69, "wenet-utils-checkpoint-module"]], "wenet.utils.cmvn module": [[70, "wenet-utils-cmvn-module"]], "wenet.utils.common module": [[71, "wenet-utils-common-module"]], "wenet.utils.config module": [[72, "wenet-utils-config-module"]], "wenet.utils.context_graph module": [[73, "wenet-utils-context-graph-module"]], "wenet.utils.ctc_utils module": [[74, "wenet-utils-ctc-utils-module"]], "wenet.utils.executor module": [[75, "wenet-utils-executor-module"]], "wenet.utils.file_utils module": [[76, "wenet-utils-file-utils-module"]], "wenet.utils.init_ali_paraformer module": [[77, "wenet-utils-init-ali-paraformer-module"]], "wenet.utils.init_model module": [[78, "wenet-utils-init-model-module"]], "wenet.utils.mask module": [[79, "wenet-utils-mask-module"]], "wenet.utils.scheduler module": [[80, "wenet-utils-scheduler-module"]], "wenet.utils.tokenize_utils module": [[81, "wenet-utils-tokenize-utils-module"]], "wenet.utils.train_utils module": [[82, "wenet-utils-train-utils-module"]], "wenet.whisper package": [[83, "wenet-whisper-package"]], "wenet.whisper.convert_whisper_to_wenet_config_and_ckpt module": [[84, "wenet-whisper-convert-whisper-to-wenet-config-and-ckpt-module"]], "wenet.whisper.whisper module": [[85, "wenet-whisper-whisper-module"]], "Python Package": [[86, "python-package"]], "Install": [[86, "install"]], "Development Install": [[86, "development-install"]], "Command line Usage": [[86, "command-line-usage"]], "Python Programming Usage": [[86, "python-programming-usage"]], "Reference": [[87, "reference"]], "Runtime for WeNet": [[89, "runtime-for-wenet"]], "Platforms Supported": [[89, "platforms-supported"]], "Architecture and Implementation": [[89, "architecture-and-implementation"]], "Architecture": [[89, "architecture"]], "Interface Design": [[89, "interface-design"]], "Cache in Details": [[89, "cache-in-details"]], "References": [[89, "references"]], "How to train models?": [[90, "how-to-train-models"]], "Tutorial on AIShell": [[91, "tutorial-on-aishell"]], "First Experiment": [[91, "first-experiment"], [92, "first-experiment"]], "Stage -1: Download data": [[91, "stage-1-download-data"], [92, "stage-1-download-data"]], "Stage 0: Prepare Training data": [[91, "stage-0-prepare-training-data"], [92, "stage-0-prepare-training-data"]], "Stage 1: Extract optinal cmvn features": [[91, "stage-1-extract-optinal-cmvn-features"], [92, "stage-1-extract-optinal-cmvn-features"]], "Stage 2: Generate label token dictionary": [[91, "stage-2-generate-label-token-dictionary"], [92, "stage-2-generate-label-token-dictionary"]], "Stage 3: Prepare WeNet data format": [[91, "stage-3-prepare-wenet-data-format"], [92, "stage-3-prepare-wenet-data-format"]], "Stage 4: Neural Network training": [[91, "stage-4-neural-network-training"], [92, "stage-4-neural-network-training"]], "Stage 5: Recognize wav using the trained model": [[91, "stage-5-recognize-wav-using-the-trained-model"], [92, "stage-5-recognize-wav-using-the-trained-model"]], "Stage 6: Export the trained model": [[91, "stage-6-export-the-trained-model"]], "Tutorial on LibriSpeech": [[92, "tutorial-on-librispeech"]], "Stage 6(Optional): Export the trained model": [[92, "stage-6-optional-export-the-trained-model"]], "Stage 7(Optional): Add LM and test it with runtime": [[92, "stage-7-optional-add-lm-and-test-it-with-runtime"]]}, "indexentries": {"module": [[15, "module-wenet.cli"]], "wenet.cli": [[15, "module-wenet.cli"]]}})
\ No newline at end of file
+Search.setIndex({"docnames": ["UIO", "context", "index", "jit_in_wenet", "lm", "papers", "pretrained_models", "pretrained_models.en", "production", "python_api/modules", "python_api/wenet", "python_api/wenet.branchformer", "python_api/wenet.branchformer.cgmlp", "python_api/wenet.branchformer.encoder", "python_api/wenet.branchformer.encoder_layer", "python_api/wenet.cli", "python_api/wenet.cli.hub", "python_api/wenet.cli.model", "python_api/wenet.cli.paraformer_model", "python_api/wenet.cli.transcribe", "python_api/wenet.dataset", "python_api/wenet.dataset.dataset", "python_api/wenet.dataset.kaldi_io", "python_api/wenet.dataset.processor", "python_api/wenet.dataset.wav_distortion", "python_api/wenet.efficient_conformer", "python_api/wenet.efficient_conformer.attention", "python_api/wenet.efficient_conformer.convolution", "python_api/wenet.efficient_conformer.encoder", "python_api/wenet.efficient_conformer.encoder_layer", "python_api/wenet.efficient_conformer.subsampling", "python_api/wenet.k2", "python_api/wenet.k2.model", "python_api/wenet.paraformer", "python_api/wenet.paraformer.attention", "python_api/wenet.paraformer.cif", "python_api/wenet.paraformer.convert_paraformer_to_wenet_config_and_ckpt", "python_api/wenet.paraformer.layers", "python_api/wenet.paraformer.paraformer", "python_api/wenet.paraformer.search", "python_api/wenet.squeezeformer", "python_api/wenet.squeezeformer.attention", "python_api/wenet.squeezeformer.conv2d", "python_api/wenet.squeezeformer.convolution", "python_api/wenet.squeezeformer.encoder", "python_api/wenet.squeezeformer.encoder_layer", "python_api/wenet.squeezeformer.positionwise_feed_forward", "python_api/wenet.squeezeformer.subsampling", "python_api/wenet.transducer", "python_api/wenet.transducer.joint", "python_api/wenet.transducer.predictor", "python_api/wenet.transducer.transducer", "python_api/wenet.transformer", "python_api/wenet.transformer.asr_model", "python_api/wenet.transformer.attention", "python_api/wenet.transformer.cmvn", "python_api/wenet.transformer.convolution", "python_api/wenet.transformer.ctc", "python_api/wenet.transformer.decoder", "python_api/wenet.transformer.decoder_layer", "python_api/wenet.transformer.embedding", "python_api/wenet.transformer.encoder", "python_api/wenet.transformer.encoder_layer", "python_api/wenet.transformer.label_smoothing_loss", "python_api/wenet.transformer.positionwise_feed_forward", "python_api/wenet.transformer.search", "python_api/wenet.transformer.subsampling", "python_api/wenet.transformer.swish", "python_api/wenet.utils", "python_api/wenet.utils.checkpoint", "python_api/wenet.utils.cmvn", "python_api/wenet.utils.common", "python_api/wenet.utils.config", "python_api/wenet.utils.context_graph", "python_api/wenet.utils.ctc_utils", "python_api/wenet.utils.executor", "python_api/wenet.utils.file_utils", "python_api/wenet.utils.init_ali_paraformer", "python_api/wenet.utils.init_model", "python_api/wenet.utils.mask", "python_api/wenet.utils.scheduler", "python_api/wenet.utils.tokenize_utils", "python_api/wenet.utils.train_utils", "python_api/wenet.whisper", "python_api/wenet.whisper.convert_whisper_to_wenet_config_and_ckpt", "python_api/wenet.whisper.whisper", "python_package", "reference", "requirements", "runtime", "train", "tutorial_aishell", "tutorial_librispeech"], "filenames": ["UIO.md", "context.md", "index.rst", "jit_in_wenet.md", "lm.md", "papers.md", "pretrained_models.md", "pretrained_models.en.md", "production.rst", "python_api/modules.rst", "python_api/wenet.rst", "python_api/wenet.branchformer.rst", "python_api/wenet.branchformer.cgmlp.rst", "python_api/wenet.branchformer.encoder.rst", "python_api/wenet.branchformer.encoder_layer.rst", "python_api/wenet.cli.rst", "python_api/wenet.cli.hub.rst", "python_api/wenet.cli.model.rst", "python_api/wenet.cli.paraformer_model.rst", "python_api/wenet.cli.transcribe.rst", "python_api/wenet.dataset.rst", "python_api/wenet.dataset.dataset.rst", "python_api/wenet.dataset.kaldi_io.rst", "python_api/wenet.dataset.processor.rst", "python_api/wenet.dataset.wav_distortion.rst", "python_api/wenet.efficient_conformer.rst", "python_api/wenet.efficient_conformer.attention.rst", "python_api/wenet.efficient_conformer.convolution.rst", "python_api/wenet.efficient_conformer.encoder.rst", "python_api/wenet.efficient_conformer.encoder_layer.rst", "python_api/wenet.efficient_conformer.subsampling.rst", "python_api/wenet.k2.rst", "python_api/wenet.k2.model.rst", "python_api/wenet.paraformer.rst", "python_api/wenet.paraformer.attention.rst", "python_api/wenet.paraformer.cif.rst", "python_api/wenet.paraformer.convert_paraformer_to_wenet_config_and_ckpt.rst", "python_api/wenet.paraformer.layers.rst", "python_api/wenet.paraformer.paraformer.rst", "python_api/wenet.paraformer.search.rst", "python_api/wenet.squeezeformer.rst", "python_api/wenet.squeezeformer.attention.rst", "python_api/wenet.squeezeformer.conv2d.rst", "python_api/wenet.squeezeformer.convolution.rst", "python_api/wenet.squeezeformer.encoder.rst", "python_api/wenet.squeezeformer.encoder_layer.rst", "python_api/wenet.squeezeformer.positionwise_feed_forward.rst", "python_api/wenet.squeezeformer.subsampling.rst", "python_api/wenet.transducer.rst", "python_api/wenet.transducer.joint.rst", "python_api/wenet.transducer.predictor.rst", "python_api/wenet.transducer.transducer.rst", "python_api/wenet.transformer.rst", "python_api/wenet.transformer.asr_model.rst", "python_api/wenet.transformer.attention.rst", "python_api/wenet.transformer.cmvn.rst", "python_api/wenet.transformer.convolution.rst", "python_api/wenet.transformer.ctc.rst", "python_api/wenet.transformer.decoder.rst", "python_api/wenet.transformer.decoder_layer.rst", "python_api/wenet.transformer.embedding.rst", "python_api/wenet.transformer.encoder.rst", "python_api/wenet.transformer.encoder_layer.rst", "python_api/wenet.transformer.label_smoothing_loss.rst", "python_api/wenet.transformer.positionwise_feed_forward.rst", "python_api/wenet.transformer.search.rst", "python_api/wenet.transformer.subsampling.rst", "python_api/wenet.transformer.swish.rst", "python_api/wenet.utils.rst", "python_api/wenet.utils.checkpoint.rst", "python_api/wenet.utils.cmvn.rst", "python_api/wenet.utils.common.rst", "python_api/wenet.utils.config.rst", "python_api/wenet.utils.context_graph.rst", "python_api/wenet.utils.ctc_utils.rst", "python_api/wenet.utils.executor.rst", "python_api/wenet.utils.file_utils.rst", "python_api/wenet.utils.init_ali_paraformer.rst", "python_api/wenet.utils.init_model.rst", "python_api/wenet.utils.mask.rst", "python_api/wenet.utils.scheduler.rst", "python_api/wenet.utils.tokenize_utils.rst", "python_api/wenet.utils.train_utils.rst", "python_api/wenet.whisper.rst", "python_api/wenet.whisper.convert_whisper_to_wenet_config_and_ckpt.rst", "python_api/wenet.whisper.whisper.rst", "python_package.md", "reference.rst", "requirements.txt", "runtime.md", "train.rst", "tutorial_aishell.md", "tutorial_librispeech.md"], "titles": ["Big Data Training: UIO", "Context Biasing", "Welcome to wenet\u2019s documentation!", "JIT in WeNet", "LM for WeNet", "Papers", "Pretrained Models in WeNet", "Pretrained Models in WeNet", "Production Runtime", "Python API Reference", "wenet package", "wenet.branchformer package", "wenet.branchformer.cgmlp module", "wenet.branchformer.encoder module", "wenet.branchformer.encoder_layer module", "wenet.cli package", "wenet.cli.hub module", "wenet.cli.model module", "wenet.cli.paraformer_model module", "wenet.cli.transcribe module", "wenet.dataset package", "wenet.dataset.dataset module", "wenet.dataset.kaldi_io module", "wenet.dataset.processor module", "wenet.dataset.wav_distortion module", "wenet.efficient_conformer package", "wenet.efficient_conformer.attention module", "wenet.efficient_conformer.convolution module", "wenet.efficient_conformer.encoder module", "wenet.efficient_conformer.encoder_layer module", "wenet.efficient_conformer.subsampling module", "wenet.k2 package", "wenet.k2.model module", "wenet.paraformer package", "wenet.paraformer.attention module", "wenet.paraformer.cif module", "wenet.paraformer.convert_paraformer_to_wenet_config_and_ckpt module", "wenet.paraformer.layers module", "wenet.paraformer.paraformer module", "wenet.paraformer.search module", "wenet.squeezeformer package", "wenet.squeezeformer.attention module", "wenet.squeezeformer.conv2d module", "wenet.squeezeformer.convolution module", "wenet.squeezeformer.encoder module", "wenet.squeezeformer.encoder_layer module", "wenet.squeezeformer.positionwise_feed_forward module", "wenet.squeezeformer.subsampling module", "wenet.transducer package", "wenet.transducer.joint module", "wenet.transducer.predictor module", "wenet.transducer.transducer module", "wenet.transformer package", "wenet.transformer.asr_model module", "wenet.transformer.attention module", "wenet.transformer.cmvn module", "wenet.transformer.convolution module", "wenet.transformer.ctc module", "wenet.transformer.decoder module", "wenet.transformer.decoder_layer module", "wenet.transformer.embedding module", "wenet.transformer.encoder module", "wenet.transformer.encoder_layer module", "wenet.transformer.label_smoothing_loss module", "wenet.transformer.positionwise_feed_forward module", "wenet.transformer.search module", "wenet.transformer.subsampling module", "wenet.transformer.swish module", "wenet.utils package", "wenet.utils.checkpoint module", "wenet.utils.cmvn module", "wenet.utils.common module", "wenet.utils.config module", "wenet.utils.context_graph module", "wenet.utils.ctc_utils module", "wenet.utils.executor module", "wenet.utils.file_utils module", "wenet.utils.init_ali_paraformer module", "wenet.utils.init_model module", "wenet.utils.mask module", "wenet.utils.scheduler module", "wenet.utils.tokenize_utils module", "wenet.utils.train_utils module", "wenet.whisper package", "wenet.whisper.convert_whisper_to_wenet_config_and_ckpt module", "wenet.whisper.whisper module", "Python Package", "Reference", "&lt;no title&gt;", "Runtime for WeNet", "How to train models?", "Tutorial on AIShell", "Tutorial on LibriSpeech"], "terms": {"In": [0, 1, 4, 89, 91, 92], "order": [0, 1], "support": [0, 1, 4, 86, 92], "model": [0, 1, 2, 3, 4, 15, 31, 86, 89], "industri": [0, 4, 89], "ten": 0, "million": 0, "hour": [0, 91, 92], "speech": [0, 4, 5, 89, 91, 92], "dataset": [0, 4, 6, 7, 10, 91, 92], "process": [0, 1, 91, 92], "method": [0, 3, 89, 91, 92], "unifi": [0, 4, 89, 92], "ha": [0, 4, 6, 7, 89], "been": [0, 4, 6, 7], "updat": [0, 1, 3], "wenet": [0, 1, 5, 8, 9, 86, 90], "The": [0, 1, 4, 6, 7, 89, 91, 92], "document": 0, "introduc": 0, "from": [0, 1, 4, 89, 91, 92], "follow": [0, 1, 3, 4, 6, 7, 86, 89, 91, 92], "section": [0, 89], "old": [0, 92], "i": [0, 1, 2, 3, 4, 6, 7, 89, 91, 92], "base": [0, 1, 2, 4, 89, 91, 92], "pytorch": [0, 3, 6, 7, 89], "": [0, 1, 4, 6, 7, 89, 92], "nativ": 0, "dure": [0, 1, 92], "need": [0, 1, 4], "load": [0, 3], "all": [0, 4, 89], "audio": [0, 86, 91, 92], "path": [0, 1, 3, 91, 92], "correspondingli": 0, "label": [0, 1, 86], "memori": [0, 91, 92], "one": [0, 1, 4, 89, 91, 92], "time": [0, 1, 89, 91, 92], "randomli": 0, "read": 0, "case": [0, 89, 92], "grade": 0, "ultra": 0, "larg": [0, 1, 89, 91, 92], "scale": 0, "eg": 0, "more": [0, 5, 89, 91, 92], "than": [0, 89, 91, 92], "50": [0, 91, 92], "000": 0, "thi": [0, 1, 4, 91, 92], "caus": [0, 91, 92], "fail": [0, 1, 3, 89], "run": [0, 91, 92], "two": [0, 3, 4, 6, 7, 89, 91, 92], "reason": [0, 4, 91, 92], "out": [0, 4, 91, 92], "oom": 0, "physic": 0, "gener": [0, 4, 89], "machin": [0, 91, 92], "difficult": 0, "slow": 0, "down": 0, "perform": [0, 1], "cannot": 0, "us": [0, 1, 3, 6, 7, 86, 89], "file": [0, 1, 91, 92], "cach": 0, "speed": [0, 4], "greatli": 0, "reduc": [0, 6, 7], "inspir": 0, "webdataset": 0, "tfrecord": 0, "redesign": 0, "Its": 0, "core": [0, 4], "idea": 0, "make": [0, 89], "multipl": 0, "small": [0, 89], "1000": 0, "piec": [0, 92], "compress": 0, "packet": 0, "tar": 0, "them": [0, 92], "iterabledataset": 0, "advantag": [0, 89], "onli": [0, 1, 4, 89, 92], "index": [0, 2, 92], "inform": [0, 1, 86], "packag": [0, 2, 9], "maintain": 0, "which": [0, 1, 3, 4, 89, 91, 92], "save": [0, 3, 6, 7, 92], "solv": 0, "problem": [0, 1, 91, 92], "fly": [0, 1, 91, 92], "decompress": 0, "same": [0, 1, 4, 91, 92], "sequenc": [0, 4, 89, 91, 92], "random": 0, "differ": [0, 1, 4, 6, 7, 89], "can": [0, 1, 3, 4, 6, 7, 86, 89, 91, 92], "ensur": [0, 3], "global": [0, 91, 92], "new": [0, 1], "take": [0, 91, 92], "account": 0, "both": [0, 4, 89], "provid": [0, 6, 7, 91, 92], "we": [0, 1, 3, 4, 6, 7, 89, 91, 92], "call": [0, 1, 89, 91, 92], "shown": [0, 4, 92], "figur": 0, "below": [0, 4], "some": [0, 1, 91, 92], "necessari": [0, 1], "explan": 0, "about": [0, 91, 92], "abov": [0, 1], "mode": [0, 91, 92], "local": [0, 4, 91, 92], "requir": [0, 3, 4, 6, 7, 89, 91, 92], "must": [0, 91, 92], "sort": [0, 92], "kaldi": [0, 1, 4, 92], "style": [0, 4, 91, 92], "wav": [0, 1, 4, 86], "scp": [0, 4, 91, 92], "text": [0, 1, 4, 86, 91, 92], "It": [0, 1, 4, 89, 91, 92], "befor": [0, 1, 3, 92], "share": [0, 89, 91, 92], "network": [0, 6, 7], "cloud": 0, "storag": 0, "txt": [0, 1, 4, 91, 92], "ar": [0, 1, 4, 7, 86, 89, 91, 92], "store": 0, "singl": [0, 1], "also": [0, 1, 4, 89, 91, 92], "adopt": [0, 1, 89], "implement": 0, "practic": [0, 1, 89], "flexibl": 0, "easier": 0, "expand": 0, "debug": 0, "exampl": [0, 1, 4, 6, 7, 91, 92], "def": 0, "read_dataset": 0, "filenam": 0, "batch_siz": [0, 91, 92], "tf": 0, "tfrecorddataset": 0, "map": [0, 91, 92], "_parse_image_funct": 0, "num_parallel_cal": 0, "experiment": 0, "autotun": 0, "shuffl": 0, "500": [0, 92], "batch": 0, "drop_remaind": 0, "true": [0, 1, 91, 92], "repeat": 0, "prefetch": 0, "buffer_s": 0, "return": [0, 1], "refer": [0, 2, 3, 91, 92], "dataflow": 0, "includ": [0, 1, 4], "modul": [0, 2, 3, 11, 15, 20, 25, 31, 33, 40, 48, 52, 68, 83, 89], "token": [0, 1, 86], "convert": 0, "specifi": [0, 1, 86, 92], "unit": [0, 1, 4, 91, 92], "char": [0, 4, 91, 92], "bpe": [0, 4, 92], "filter": 0, "unqualifi": 0, "resampl": 0, "option": [0, 4], "comput": [0, 1, 89, 91, 92], "_": [0, 1, 4, 89, 91, 92], "fbank": 0, "featur": 0, "extract": 0, "spec": 0, "augment": 0, "enhanc": 0, "disrupt": 0, "organ": [0, 4, 91, 92], "pad": 0, "what": [0, 3, 4], "There": [0, 4], "sever": [0, 89, 91, 92], "paramet": [0, 86, 91, 92], "note": 0, "first": [0, 4, 89], "buffer": 0, "size": [0, 6, 7, 89, 91, 92], "recommend": 0, "larger": 0, "number": [0, 91, 92], "contain": [0, 1, 91, 92], "each": [0, 1, 91, 92], "equival": 0, "between": [0, 1, 91], "increas": 0, "you": [0, 3, 4, 6, 7, 86, 89, 91, 92], "set": [0, 89, 91, 92], "1500": 0, "accord": [0, 1, 4], "frame": [0, 1, 4, 89], "oper": 0, "veri": [0, 4, 89], "import": [0, 86], "improv": [0, 1], "dataload": [0, 91], "pre": 0, "granular": 0, "final": [0, 1, 4, 89, 92], "default": [0, 86], "2": [0, 1, 4, 5, 89], "due": 0, "exist": 0, "mai": [0, 1, 4, 91, 92], "alreadi": [0, 91, 92], "so": [0, 4, 89, 91, 92], "real": 0, "when": [0, 1, 3, 89, 91, 92], "insuffici": 0, "next": [0, 1], "fill": 0, "At": 0, "block": 0, "word": [0, 1, 4, 89, 91, 92], "part": [0, 4, 89, 92], "becaus": 0, "previou": [0, 89], "still": 0, "therefor": [0, 4, 89], "avoid": [0, 1, 89], "present": 0, "have": [0, 1, 4, 91, 92], "verifi": 0, "accuraci": [0, 1, 89], "200": 0, "10000": 0, "respect": 0, "cer": 0, "4": [0, 4, 6, 7], "61": 0, "63": 0, "67": 0, "espnet": [0, 92], "similar": 0, "structur": [0, 4, 7, 91, 92], "configur": 0, "thei": [0, 1, 91, 92], "achiev": [0, 89], "recognit": [0, 1, 4, 5, 89, 91, 92], "rate": [0, 89, 91, 92], "show": [0, 4, 86, 89, 91, 92], "correct": 0, "And": [0, 4, 6, 7], "observ": 0, "util": [0, 10, 92], "gpu": [0, 86, 91, 92], "80": 0, "90": 0, "indic": [0, 1, 91], "effici": 0, "high": 0, "For": [0, 4, 6, 7, 89, 91, 92], "detail": [0, 3, 4, 91, 92], "pleas": [0, 4, 6, 89, 91, 92], "http": [0, 3, 4, 6, 7, 86, 92], "github": [0, 4, 86, 91, 92], "com": [0, 4, 86], "e2": [0, 4, 86, 92], "blob": [0, 4], "main": [0, 4], "s0": [0, 91, 92], "sh": [0, 4, 91, 92], "prepar": [0, 4], "three": 0, "relat": [0, 4, 92], "script": [0, 3, 91, 92], "py": [0, 3, 4, 89, 91, 92], "train_data": [0, 92], "cv_data": [0, 92], "test_data": [0, 92], "list": [0, 1, 89, 91, 92], "data_typ": [0, 92], "symbol_t": [0, 92], "python": [0, 2, 3, 6, 7, 87, 89, 92], "bin": [0, 3, 91, 92], "gpu_id": [0, 92], "config": [0, 68, 91, 92], "train_config": [0, 92], "dict": [0, 4, 91, 92], "feat_dir": 0, "train_set": [0, 92], "dev": [0, 92], "If": [0, 1, 3, 91, 92], "type": [0, 3, 89, 91, 92], "format": [0, 4], "kei": [0, 91, 92], "bac009s0002w0122": [0, 91], "export": [0, 1, 3, 6, 7, 89], "asr": [0, 1, 2, 89, 91, 92], "openslr": [0, 6, 7, 91, 92], "33": [0, 91], "data_aishel": [0, 4, 91], "s0002": [0, 91], "\u800c\u5bf9\u697c\u5e02\u6210\u4ea4\u6291\u5236\u4f5c\u7528\u6700\u5927\u7684\u9650\u8d2d": [0, 91], "bac009s0002w0123": [0, 91], "\u4e5f\u6210\u4e3a\u5730\u65b9\u653f\u5e9c\u7684\u773c\u4e2d\u9489": [0, 91], "bac009s0002w0124": [0, 91], "\u81ea\u516d\u6708\u5e95\u547c\u548c\u6d69\u7279\u5e02\u7387\u5148\u5ba3\u5e03\u53d6\u6d88\u9650\u8d2d\u540e": [0, 91], "line": [0, 1, 2, 91, 92], "json": [0, 91, 92], "serial": 0, "string": 0, "field": [0, 91, 92], "1": [0, 1, 4, 86, 89], "maryland": 0, "binbinzhang": 0, "code": [0, 1, 3, 4, 6, 7, 89, 92], "s3": 0, "raw_wav": [0, 91], "shards_000000000": 0, "gz": [0, 92], "shards_000000001": 0, "shards_000000002": 0, "oss": 0, "examplebucket": 0, "cn": [0, 6, 7], "hangzhou": 0, "aliyunc": 0, "exampledir": 0, "q1": 0, "how": [0, 1, 2, 3, 89, 91, 92], "distribut": 0, "partit": 0, "rank": [0, 92], "num": [0, 92], "worker": 0, "segment": 0, "class": [0, 4], "distributedsampl": 0, "__init__": 0, "self": 0, "epoch": [0, 91, 92], "set_epoch": 0, "sampl": 0, "rang": [0, 4], "len": 0, "world_siz": [0, 92], "worker_id": 0, "num_work": [0, 92], "q2": 0, "deal": 0, "unbalanc": 0, "join": [0, 3], "handl": [0, 89], "imbal": 0, "alloc": 0, "applic": [1, 4], "effect": 1, "commonli": 1, "better": [1, 89, 91, 92], "uniqu": 1, "low": 1, "contextu": 1, "inject": 1, "prior": 1, "knowledg": 1, "an": [1, 2, 4, 89, 91, 92], "system": [1, 89, 92], "infer": [1, 89, 91, 92], "user": [1, 6, 7], "favorit": 1, "song": 1, "contact": [1, 4], "app": 1, "locat": 1, "convent": 1, "build": [1, 4, 92], "n": [1, 4, 89, 91, 92], "gram": [1, 4, 92], "finit": 1, "state": 1, "transduc": [1, 10], "fst": [1, 4, 92], "phrase": 1, "compos": [1, 4], "decod": [1, 4, 52, 89, 91, 92], "help": 1, "bia": 1, "result": [1, 6, 7, 86, 89, 91, 92], "toward": 1, "thu": 1, "certain": 1, "scenario": [1, 4, 89], "score": [1, 89], "p": [1, 4, 92], "c": [1, 3, 89, 91, 92], "mathbf": 1, "y": [1, 89], "interpol": 1, "x": [1, 89, 92], "shallow": 1, "fusion": 1, "mathrm": 1, "arg": [1, 3, 89], "max": 1, "log": [1, 4, 89], "lambda": 1, "where": [1, 89], "tunabl": [1, 91, 92], "hyperparamet": 1, "control": 1, "much": [1, 3, 89], "lm": [1, 2, 8], "influenc": 1, "overal": [1, 89], "want": [1, 3, 91, 92], "cat": [1, 92], "charact": [1, 91], "0": [1, 4, 5, 6, 7], "25": 1, "construct": 1, "correspond": [1, 4, 6, 7], "match": 1, "reward": 1, "obtain": 1, "artifici": 1, "boost": [1, 91, 92], "earli": 1, "do": [1, 3, 4, 91, 92], "entir": 1, "add": 1, "special": 1, "failur": 1, "arc": 1, "remov": 1, "record": [1, 91, 92], "easili": [1, 89, 91, 92], "determin": 1, "boundari": 1, "hot": 1, "That": 1, "after": [1, 91, 92], "succe": 1, "other": [1, 4, 89, 91, 92], "start": [1, 91, 92], "int": 1, "contextgraph": 1, "getnextst": 1, "cur_stat": 1, "word_id": 1, "float": 1, "bool": 1, "is_start_boundari": 1, "is_end_boundari": 1, "next_stat": 1, "travers": 1, "current": [1, 89], "arciter": 1, "stdfst": 1, "aiter": 1, "graph_": 1, "done": [1, 92], "const": 1, "stdarc": 1, "valu": [1, 86], "ilabel": 1, "backoff": 1, "might": 1, "cover": 1, "weight": 1, "els": 1, "nextstat": 1, "check": [1, 3, 91, 92], "whether": [1, 4], "One": 1, "break": 1, "append": 1, "output": [1, 89, 91, 92], "chang": [1, 4, 89, 91, 92], "function": [1, 3, 4, 91, 92], "end": [1, 2, 4, 5, 89, 91, 92], "posit": 1, "insert": 1, "tag": 1, "bag": 1, "lattic": 1, "faster": [1, 89], "onlin": 1, "modifi": [1, 3, 4], "cc": [1, 4, 6, 7, 89], "tlg": [1, 4, 92], "input": [1, 86, 89, 91, 92], "decid": 1, "processemit": 1, "processnonemit": 1, "elem": 1, "e_next": 1, "findoraddtoken": 1, "tot_cost": 1, "tok": 1, "null": 1, "begin": [1, 92], "fals": [1, 91, 92], "context_scor": 1, "context_graph_": 1, "olabel": 1, "val": 1, "context_st": 1, "graph_cost": 1, "forwardlink": 1, "next_tok": 1, "put": [1, 4], "head": 1, "link": [1, 7], "forwardlinkt": 1, "ac_cost": 1, "accumul": [1, 4], "lead": 1, "cost": [1, 89, 91, 92], "too": 1, "void": 1, "latticefasterdecodertpl": 1, "pruneforwardlink": 1, "int32": 1, "frame_plus_on": 1, "extra_costs_chang": 1, "links_prun": 1, "basefloat": 1, "delta": 1, "link_extra_cost": 1, "extra_cost": 1, "acoustic_cost": 1, "bracket": 1, "mean": [1, 91, 92], "link_exta_cost": 1, "best": [1, 4, 86, 89, 91, 92], "through": [1, 91, 92], "sourc": 1, "destin": 1, "context_path": 1, "split": [1, 4], "symbol": [1, 89, 91, 92], "tabl": [1, 88], "oov": 1, "cd": [1, 86, 91, 92], "home": [1, 91], "runtim": [1, 2, 4, 6, 7], "libtorch": [1, 89, 91, 92], "glog_logtostderr": 1, "glog_v": 1, "wav_path": [1, 91, 92], "docker_resourc": 1, "test": [1, 4], "model_dir": [1, 3, 86, 92], "decoder_main": 1, "chunk_siz": 1, "model_path": 1, "zip": [1, 3, 4, 6, 7, 92], "3": [1, 4, 89], "unit_path": 1, "tee": 1, "tansform": 2, "toolkit": [2, 5], "instal": [2, 92], "develop": [2, 89], "command": 2, "usag": 2, "program": [2, 91], "train": [2, 3, 4, 6, 7, 89], "tutori": [2, 90], "librispeech": [2, 4, 6, 7, 90], "aishel": [2, 4, 6, 7, 90], "pretrain": [2, 90, 92], "big": [2, 90, 91, 92], "data": [2, 4, 90], "uio": [2, 90], "product": [2, 3, 4, 5], "context": [2, 8, 89], "bias": [2, 8], "jit": [2, 6, 7, 8, 89], "paper": [2, 87, 91, 92], "api": [2, 87], "search": [2, 4, 33, 52, 89, 91, 92], "page": 2, "our": [3, 6, 7, 89, 92], "directli": [3, 4, 6, 7, 89], "torch": [3, 88, 89, 91, 92], "essenti": 3, "deploi": 3, "see": [3, 4, 6, 7, 89, 91, 92], "resourc": [3, 92], "introduct": [3, 89], "TO": [3, 92], "torchscript": 3, "languag": [3, 4, 6, 7, 86, 91, 92], "A": [3, 4, 92], "IN": [3, 92], "deep": 3, "dive": 3, "research": [3, 4], "To": [3, 4], "try": [3, 91, 92], "stage": 3, "should": [3, 89], "satisfi": 3, "script_model": 3, "o": 3, "init": [3, 92], "principl": [3, 4], "taken": 3, "consider": 3, "contribut": 3, "especi": [3, 4], "subclass": 3, "nn": [3, 91, 92], "forward": [3, 89], "know": 3, "allow": 3, "disallow": 3, "tensor": 3, "unsupport": 3, "attribut": 3, "coverag": 3, "explicit": 3, "possibl": [3, 92], "forc": [3, 86], "typeguard": [3, 88], "readthedoc": 3, "io": 3, "en": [3, 6, 7, 92], "latest": 3, "userguid": 3, "html": 3, "statist": [4, 91, 92], "wfst": 4, "framework": [4, 89], "custom": [4, 91, 92], "why": 4, "question": 4, "mani": 4, "peopl": [4, 91, 92], "ask": [4, 91, 92], "now": [4, 86], "rnn": 4, "transform": [4, 10, 89, 91, 92], "full": [4, 92], "swing": 4, "doe": [4, 91, 92], "go": [4, 91, 92], "backward": 4, "simpl": [4, 89, 91, 92], "matur": 4, "complet": 4, "tool": [4, 91, 92], "ani": [4, 91, 92], "amount": 4, "corpu": [4, 92], "fast": [4, 89], "hotfix": 4, "easi": 4, "wide": 4, "actual": [4, 91, 92], "second": 4, "sinc": [4, 6, 7, 89], "work": [4, 89, 91, 92], "hard": 4, "abandon": 4, "tradit": [4, 89], "complex": 4, "technologi": 4, "back": 4, "standard": [4, 89], "power": 4, "solut": 4, "bug": 4, "fix": 4, "replac": 4, "person": 4, "just": [4, 89, 91, 92], "like": [4, 91, 92], "goal": 4, "readi": 4, "prioriti": 4, "draw": 4, "pure": 4, "As": 4, "describ": [4, 89], "choos": 4, "your": [4, 86, 91, 92], "whole": [4, 89, 91, 92], "bellow": 4, "pictur": [4, 89], "wai": [4, 89], "without": [4, 91], "ctc": [4, 52, 89, 91, 92], "prefix": [4, 89, 91, 92], "beam": [4, 89, 91, 92], "With": [4, 89], "graph": 4, "t": [4, 86, 89], "lexicon": [4, 92], "l": [4, 86, 92], "g": [4, 86, 91, 92], "typic": [4, 89, 92], "chines": [4, 6, 86], "english": [4, 86, 92], "its": 4, "\u6211\u4eec": 4, "\u6211": 4, "\u4eec": 4, "appl": 4, "five": 4, "letter": 4, "e": [4, 86, 89, 91, 92], "phonem": 4, "pronunci": 4, "purpos": 4, "name": [4, 91, 92], "compil": 4, "represent": 4, "viterbi": 4, "algorithm": [4, 91, 92], "eas": 4, "keep": 4, "independ": 4, "migrat": 4, "directori": [4, 92], "minim": 4, "remain": 4, "origin": [4, 91, 92], "glog": 4, "meet": [4, 91, 92], "lint": 4, "wrap": 4, "latticefasterdecod": 4, "blank": [4, 91, 92], "skip": [4, 91, 92], "up": 4, "addit": [4, 89], "arpa2fst": 4, "fstdeterminizestar": 4, "fsttablecompos": 4, "fstminimizeencod": 4, "built": 4, "box": 4, "get": [4, 89, 91, 92], "consist": 4, "gain": 4, "10": [4, 92], "aishell2": [4, 6, 7], "here": [4, 6, 7, 89, 91, 92], "dictionari": 4, "7": [4, 91], "unit_fil": [4, 92], "mkdir": [4, 92], "cp": [4, 92], "prepare_dict": [4, 92], "resource_aishel": 4, "filter_scp": 4, "pl": 4, "transcript": [4, 86, 91, 92], "aishell_transcript_v0": 4, "8": [4, 91, 92], "aishell_train_lm": 4, "compile_lexicon_token_fst": [4, 92], "tmp": [4, 92], "lang": [4, 92], "make_tlg": [4, 92], "lang_test": [4, 92], "exit": [4, 92], "nj": [4, 92], "16": [4, 92], "15": 4, "lattice_beam": [4, 92], "5": 4, "max_act": [4, 92], "7000": [4, 92], "blank_skip_thresh": [4, 92], "98": [4, 92], "ctc_weight": [4, 92], "rescoring_weight": [4, 92], "fst_path": [4, 92], "dict_path": [4, 92], "dir": [4, 86, 91, 92], "lm_with_runtim": 4, "orient": 5, "stream": [5, 89], "non": [5, 6, 89, 92], "accept": 5, "interspeech": 5, "2021": 5, "2022": 5, "facilit": [6, 7], "checkpoint": [6, 7, 68, 91, 92], "suffix": [6, 7], "pt": [6, 7, 91, 92], "reproduc": [6, 7], "publish": [6, 7], "continu": [6, 7, 89, 91, 92], "x86": [6, 7, 89], "android": [6, 7, 89], "quantiz": [6, 7], "traffic": [6, 7], "BY": [6, 7], "org": [6, 7, 92], "12": [6, 7, 92], "visit": 6, "download": [6, 7], "contributor": [6, 7], "conform": [6, 7, 89, 91, 92], "gigaspeech": [6, 7, 91, 92], "multi_cn": [6, 7], "wenetspeech": [6, 7, 91, 92], "given": [7, 92], "subpackag": 9, "branchform": 10, "submodul": 10, "cli": 10, "efficient_conform": 10, "k2": 10, "paraform": [10, 86], "squeezeform": 10, "whisper": 10, "cgmlp": 11, "encod": [11, 25, 40, 52, 89, 91, 92], "encoder_lay": [11, 25, 40, 52], "hub": 15, "paraformer_model": 15, "transcrib": [15, 86], "kaldi_io": 20, "processor": 20, "wav_distort": 20, "attent": [25, 33, 40, 52, 89, 91, 92], "convolut": [25, 40, 52], "subsampl": [25, 40, 52, 89], "cif": 33, "convert_paraformer_to_wenet_config_and_ckpt": 33, "layer": [33, 89], "conv2d": 40, "positionwise_feed_forward": [40, 52], "joint": [48, 89], "predictor": 48, "asr_model": 52, "cmvn": [52, 68], "decoder_lay": 52, "embed": 52, "label_smoothing_loss": 52, "swish": 52, "common": 68, "context_graph": 68, "ctc_util": 68, "executor": 68, "file_util": 68, "init_ali_paraform": 68, "init_model": 68, "mask": 68, "schedul": 68, "tokenize_util": 68, "train_util": 68, "convert_whisper_to_wenet_config_and_ckpt": 83, "pip": 86, "git": 86, "clone": 86, "m": 86, "own": 86, "devic": [86, 89, 91, 92], "id": [86, 89, 91, 92], "repres": 86, "cpu": 86, "show_tokens_info": 86, "level": [86, 89], "timestamp": 86, "confid": 86, "etc": 86, "align": 86, "load_model": 86, "xxx": 86, "print": [86, 92], "jinja2": 88, "nbsphinx": 88, "sphinx": 88, "recommonmark": 88, "markdown": 88, "rtd": 88, "theme": 88, "torchaudio": [88, 91], "pass": 89, "u2": [89, 91, 92], "balanc": 89, "latenc": 89, "chunk": [89, 92], "accur": 89, "rescor": [89, 91, 92], "autoregress": 89, "benefit": 89, "find": [89, 91, 92], "stabl": 89, "long": 89, "form": 89, "usual": 89, "degrad": 89, "lot": 89, "stamp": 89, "spike": 89, "aspect": 89, "favor": 89, "server": 89, "finish": [89, 91, 92], "fed": 89, "activ": 89, "linear": 89, "softmax": 89, "prob": 89, "further": 89, "partial": 89, "everi": 89, "select": 89, "group": 89, "affect": 89, "infinit": 89, "give": 89, "limit": 89, "less": [89, 91, 92], "promis": 89, "proper": 89, "brief": 89, "descript": 89, "subsampling_r": 89, "right_context": 89, "right": 89, "sos_symbol": 89, "eos_symbol": 89, "eo": [89, 91, 92], "forward_encoder_chunk": 89, "ctc_activ": 89, "forward_attention_decod": 89, "increment": 89, "requri": 89, "histori": 89, "att": 89, "cnn": 89, "left": 89, "causal": 89, "howev": 89, "stride": 89, "tircki": 89, "simpli": 89, "overlap": 89, "straightforward": 89, "neglig": 89, "fraction": 89, "blue": 89, "color": 89, "vocabulari": [89, 91, 92], "bi": 89, "direct": 89, "recurr": 89, "dnn": 89, "feel": [91, 92], "free": [91, 92], "issu": [91, 92], "thank": [91, 92], "kind": [91, 92], "feedback": [91, 92], "recip": [91, 92], "suggest": [91, 92], "manual": [91, 92], "understand": [91, 92], "bash": [91, 92], "stop_stag": [91, 92], "could": [91, 92], "variabl": [91, 92], "absolut": 91, "usernam": 91, "aishell_data_prep": 91, "tab": [91, 92], "separ": [91, 92], "column": [91, 92], "wav_id": [91, 92], "text_label": [91, 92], "bac009s0002w0125": 91, "\u5404\u5730\u653f\u5e9c\u4fbf\u7eb7\u7eb7\u8ddf\u8fdb": 91, "raw": [91, 92], "step": [91, 92], "copi": 91, "compute_cmvn_stat": [91, 92], "cepstral": [91, 92], "varianc": [91, 92], "normal": [91, 92], "These": [91, 92], "acoust": [91, 92], "integ": [91, 92], "unk": [91, 92], "\u4e00": 91, "\u4e01": 91, "\u9f9a": 91, "4230": 91, "\u9f9f": 91, "4231": 91, "4232": 91, "denot": [91, 92], "unknown": [91, 92], "utter": [91, 92], "aslo": [91, 92], "design": [91, 92], "anoth": [91, 92], "shard": [91, 92], "10k": [91, 92], "appli": [91, 92], "5k": [91, 92], "multi": [91, 92], "ddp": [91, 92], "dist_backend": [91, 92], "nccl": [91, 92], "gloo": [91, 92], "cuda": [91, 92], "visibl": [91, 92], "cuda_visible_devic": [91, 92], "card": [91, 92], "resum": [91, 92], "termin": [91, 92], "accident": [91, 92], "exp": [91, 92], "your_exp": [91, 92], "Then": [91, 92], "optim": [91, 92], "loss": [91, 92], "yaml": [91, 92], "conf": [91, 92], "train_conform": [91, 92], "tensorboard": [91, 92], "depend": [91, 92], "2080": [91, 92], "ti": [91, 92], "dai": [91, 92], "monitor": [91, 92], "logdir": [91, 92], "your_exp_nam": [91, 92], "port": [91, 92], "12598": [91, 92], "bind_al": [91, 92], "averag": [91, 92], "average_checkpoint": [91, 92], "average_num": [91, 92], "cross": [91, 92], "valid": [91, 92], "four": [91, 92], "ctc_greedy_search": [91, 92], "greedi": [91, 92], "ctc_prefix_beam_search": [91, 92], "attention_rescor": [91, 92], "candid": [91, 92], "beam_siz": [91, 92], "higher": [91, 92], "greater": [91, 92], "wer": [91, 92], "evalu": [91, 92], "calcul": [91, 92], "error": [91, 92], "export_jit": [91, 92], "data_url": 92, "www": 92, "datadir": 92, "le": 92, "ge": 92, "echo": 92, "clean": 92, "100": 92, "360": 92, "download_and_untar": 92, "fi": 92, "wave_data": 92, "underscor": 92, "data_prep_torchaudio": 92, "1867": 92, "154075": 92, "0014": 92, "flac": 92, "1970": 92, "26100": 92, "0022": 92, "him": 92, "THAT": 92, "IT": 92, "did": 92, "AT": 92, "task": 92, "yourself": 92, "But": 92, "most": 92, "train_960": 92, "merg": 92, "total": 92, "train_clean_100": 92, "train_clean_360": 92, "train_other_500": 92, "f": 92, "dev_clean": 92, "dev_oth": 92, "in_scp": 92, "out_cmvn": 92, "global_cmvn": 92, "subset": 92, "lang_char": 92, "bpemod": 92, "nbpe": 92, "_unit": 92, "bpemodel": 92, "linguist": 92, "borrow": 92, "cut": 92, "d": 92, "spm_train": 92, "vocab_s": 92, "model_typ": 92, "model_prefix": 92, "input_sentence_s": 92, "100000000": 92, "spm_encod": 92, "output_format": 92, "tr": 92, "uniq": 92, "awk": 92, "nr": 92, "num_token": 92, "wc": 92, "byte": 92, "pair": 92, "sentencepiec": 92, "defin": 92, "ab": 92, "4995": 92, "young": 92, "4996": 92, "4997": 92, "youth": 92, "4998": 92, "z": 92, "4999": 92, "zz": 92, "5000": 92, "5001": 92, "recog_set": 92, "make_raw_list": 92, "1455": 92, "134435": 92, "0000": 92, "mnt": 92, "nf": 92, "ptm1": 92, "open": 92, "THE": 92, "girl": 92, "who": 92, "came": 92, "INTO": 92, "world": 92, "ON": 92, "night": 92, "jess": 92, "ran": 92, "cry": 92, "god": 92, "he": 92, "BE": 92, "son": 92, "had": 92, "grown": 92, "womanhood": 92, "farm": 92, "0001": 92, "AND": 92, "NOT": 92, "angri": 92, "she": 92, "wa": 92, "often": 92, "moros": 92, "silent": 92, "winesburg": 92, "said": 92, "drank": 92, "her": 92, "husband": 92, "banker": 92, "0002": 92, "BUT": 92, "louis": 92, "made": 92, "happi": 92, "flew": 92, "half": 92, "insan": 92, "fit": 92, "OF": 92, "temper": 92, "sometim": 92, "noisi": 92, "quarrelsom": 92, "swore": 92, "cri": 92, "anger": 92, "got": 92, "knife": 92, "kitchen": 92, "threaten": 92, "life": 92, "init_fil": 92, "ddp_init": 92, "rm": 92, "delet": 92, "init_method": 92, "readlink": 92, "num_gpu": 92, "otherwis": 92, "cmvn_opt": 92, "write": 92, "dimens": 92, "later": 92, "pin_memori": 92, "wait": 92, "todo": 92, "decode_checkpoint": 92, "avg_": 92, "average_model": 92, "dst_model": 92, "src_path": 92, "val_best": 92, "decoding_chunk_s": 92, "dynam": 92, "poll": 92, "idx": 92, "decode_mod": 92, "test_dir": 92, "penalti": 92, "result_fil": 92, "text_bp": 92, "spm_decod": 92, "input_format": 92, "sed": 92, "v": 92, "eq": 92, "output_fil": 92, "integr": 92, "which_lm": 92, "prune": 92, "1e": 92, "arpa": 92, "wget": 92, "11": 92, "unzip": 92, "gunzip": 92, "k": 92, "train_960_unigram5000": 92, "fst_dir": 92, "acoustic_scal": 92, "lm_with_runtime_": 92, "tail": 92}, "objects": {"wenet": [[15, 0, 0, "-", "cli"]]}, "objtypes": {"0": "py:module"}, "objnames": {"0": ["py", "module", "Python module"]}, "titleterms": {"big": 0, "data": [0, 91, 92], "train": [0, 90, 91, 92], "uio": 0, "necess": 0, "upgrad": 0, "io": 0, "mothod": 0, "system": [0, 4], "design": [0, 4, 89], "overal": 0, "chain": 0, "valid": 0, "experi": [0, 91, 92], "aishel": [0, 91], "raw": 0, "v": 0, "shard": 0, "wenetspeech": 0, "usag": [0, 1, 86], "q": 0, "A": 0, "context": 1, "bias": 1, "graph": 1, "ctc": [1, 57], "prefix": 1, "beam": 1, "search": [1, 39, 65], "wfst": 1, "prune": 1, "welcom": 2, "wenet": [2, 3, 4, 6, 7, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 89, 91, 92], "": 2, "document": 2, "content": [2, 8, 87, 90], "indic": 2, "tabl": 2, "jit": 3, "lm": [4, 92], "motiv": 4, "implement": [4, 89], "result": 4, "how": [4, 90], "us": [4, 91, 92], "paper": 5, "pretrain": [6, 7], "model": [6, 7, 17, 32, 90, 91, 92], "type": [6, 7], "licens": [6, 7], "list": [6, 7], "product": 8, "runtim": [8, 89, 92], "python": [9, 86], "api": 9, "refer": [9, 87, 89], "packag": [10, 11, 15, 20, 25, 31, 33, 40, 48, 52, 68, 83, 86], "subpackag": 10, "branchform": [11, 12, 13, 14], "submodul": [11, 15, 20, 25, 31, 33, 40, 48, 52, 68, 83], "cgmlp": 12, "modul": [12, 13, 14, 16, 17, 18, 19, 21, 22, 23, 24, 26, 27, 28, 29, 30, 32, 34, 35, 36, 37, 38, 39, 41, 42, 43, 44, 45, 46, 47, 49, 50, 51, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 84, 85], "encod": [13, 28, 44, 61], "encoder_lay": [14, 29, 45, 62], "cli": [15, 16, 17, 18, 19], "hub": 16, "paraformer_model": 18, "transcrib": 19, "dataset": [20, 21, 22, 23, 24], "kaldi_io": 22, "processor": 23, "wav_distort": 24, "efficient_conform": [25, 26, 27, 28, 29, 30], "attent": [26, 34, 41, 54], "convolut": [27, 43, 56], "subsampl": [30, 47, 66], "k2": [31, 32], "paraform": [33, 34, 35, 36, 37, 38, 39], "cif": 35, "convert_paraformer_to_wenet_config_and_ckpt": 36, "layer": 37, "squeezeform": [40, 41, 42, 43, 44, 45, 46, 47], "conv2d": 42, "positionwise_feed_forward": [46, 64], "transduc": [48, 49, 50, 51], "joint": 49, "predictor": 50, "transform": [52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67], "asr_model": 53, "cmvn": [55, 70, 91, 92], "decod": 58, "decoder_lay": 59, "embed": 60, "label_smoothing_loss": 63, "swish": 67, "util": [68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82], "checkpoint": 69, "common": 71, "config": 72, "context_graph": 73, "ctc_util": 74, "executor": 75, "file_util": 76, "init_ali_paraform": 77, "init_model": 78, "mask": 79, "schedul": 80, "tokenize_util": 81, "train_util": 82, "whisper": [83, 84, 85], "convert_whisper_to_wenet_config_and_ckpt": 84, "instal": 86, "develop": 86, "command": 86, "line": 86, "program": 86, "platform": 89, "support": 89, "architectur": 89, "interfac": 89, "cach": 89, "detail": 89, "tutori": [91, 92], "first": [91, 92], "stage": [91, 92], "1": [91, 92], "download": [91, 92], "0": [91, 92], "prepar": [91, 92], "extract": [91, 92], "optin": [91, 92], "featur": [91, 92], "2": [91, 92], "gener": [91, 92], "label": [91, 92], "token": [91, 92], "dictionari": [91, 92], "3": [91, 92], "format": [91, 92], "4": [91, 92], "neural": [91, 92], "network": [91, 92], "5": [91, 92], "recogn": [91, 92], "wav": [91, 92], "6": [91, 92], "export": [91, 92], "librispeech": 92, "option": 92, "7": 92, "add": 92, "test": 92}, "envversion": {"sphinx.domains.c": 3, "sphinx.domains.changeset": 1, "sphinx.domains.citation": 1, "sphinx.domains.cpp": 9, "sphinx.domains.index": 1, "sphinx.domains.javascript": 3, "sphinx.domains.math": 2, "sphinx.domains.python": 4, "sphinx.domains.rst": 2, "sphinx.domains.std": 2, "nbsphinx": 4, "sphinx.ext.viewcode": 1, "sphinx.ext.todo": 2, "sphinx": 58}, "alltitles": {"Big Data Training: UIO": [[0, "big-data-training-uio"]], "Necessity of upgrading IO mothod": [[0, "necessity-of-upgrading-io-mothod"]], "System design of UIO": [[0, "system-design-of-uio"]], "Overall design": [[0, "overall-design"]], "Chain IO": [[0, "chain-io"]], "Validation experiments": [[0, "validation-experiments"]], "Aishell(raw vs shard)": [[0, "aishell-raw-vs-shard"]], "WenetSpeech(shard)": [[0, "wenetspeech-shard"]], "Usage of UIO": [[0, "usage-of-uio"]], "Q&A": [[0, "q-a"]], "Context Biasing": [[1, "context-biasing"]], "Context Graph": [[1, "context-graph"]], "CTC Prefix Beam Search": [[1, "ctc-prefix-beam-search"]], "CTC WFST Beam Search": [[1, "ctc-wfst-beam-search"]], "Pruning": [[1, "pruning"]], "Usage": [[1, "usage"]], "Welcome to wenet\u2019s documentation!": [[2, "welcome-to-wenet-s-documentation"]], "Contents:": [[2, null], [8, null], [87, null], [90, null]], "Indices and tables": [[2, "indices-and-tables"]], "JIT in WeNet": [[3, "jit-in-wenet"]], "LM for WeNet": [[4, "lm-for-wenet"]], "Motivation": [[4, "motivation"]], "System Design": [[4, "system-design"]], "Implementation": [[4, "implementation"]], "Results": [[4, "results"]], "How to use?": [[4, "how-to-use"]], "Papers": [[5, "papers"]], "Pretrained Models in WeNet": [[6, "pretrained-models-in-wenet"], [7, "pretrained-models-in-wenet"]], "Model Types": [[6, "model-types"], [7, "model-types"]], "Model License": [[6, "model-license"], [7, "model-license"]], "Model List": [[6, "model-list"], [7, "model-list"]], "Production Runtime": [[8, "production-runtime"]], "Python API Reference": [[9, "python-api-reference"]], "wenet package": [[10, "wenet-package"]], "Subpackages": [[10, "subpackages"]], "wenet.branchformer package": [[11, "wenet-branchformer-package"]], "Submodules": [[11, "submodules"], [15, "submodules"], [20, "submodules"], [25, "submodules"], [31, "submodules"], [33, "submodules"], [40, "submodules"], [48, "submodules"], [52, "submodules"], [68, "submodules"], [83, "submodules"]], "wenet.branchformer.cgmlp module": [[12, "wenet-branchformer-cgmlp-module"]], "wenet.branchformer.encoder module": [[13, "wenet-branchformer-encoder-module"]], "wenet.branchformer.encoder_layer module": [[14, "wenet-branchformer-encoder-layer-module"]], "wenet.cli package": [[15, "module-wenet.cli"]], "wenet.cli.hub module": [[16, "wenet-cli-hub-module"]], "wenet.cli.model module": [[17, "wenet-cli-model-module"]], "wenet.cli.paraformer_model module": [[18, "wenet-cli-paraformer-model-module"]], "wenet.cli.transcribe module": [[19, "wenet-cli-transcribe-module"]], "wenet.dataset package": [[20, "wenet-dataset-package"]], "wenet.dataset.dataset module": [[21, "wenet-dataset-dataset-module"]], "wenet.dataset.kaldi_io module": [[22, "wenet-dataset-kaldi-io-module"]], "wenet.dataset.processor module": [[23, "wenet-dataset-processor-module"]], "wenet.dataset.wav_distortion module": [[24, "wenet-dataset-wav-distortion-module"]], "wenet.efficient_conformer package": [[25, "wenet-efficient-conformer-package"]], "wenet.efficient_conformer.attention module": [[26, "wenet-efficient-conformer-attention-module"]], "wenet.efficient_conformer.convolution module": [[27, "wenet-efficient-conformer-convolution-module"]], "wenet.efficient_conformer.encoder module": [[28, "wenet-efficient-conformer-encoder-module"]], "wenet.efficient_conformer.encoder_layer module": [[29, "wenet-efficient-conformer-encoder-layer-module"]], "wenet.efficient_conformer.subsampling module": [[30, "wenet-efficient-conformer-subsampling-module"]], "wenet.k2 package": [[31, "wenet-k2-package"]], "wenet.k2.model module": [[32, "wenet-k2-model-module"]], "wenet.paraformer package": [[33, "wenet-paraformer-package"]], "wenet.paraformer.attention module": [[34, "wenet-paraformer-attention-module"]], "wenet.paraformer.cif module": [[35, "wenet-paraformer-cif-module"]], "wenet.paraformer.convert_paraformer_to_wenet_config_and_ckpt module": [[36, "wenet-paraformer-convert-paraformer-to-wenet-config-and-ckpt-module"]], "wenet.paraformer.layers module": [[37, "wenet-paraformer-layers-module"]], "wenet.paraformer.paraformer module": [[38, "wenet-paraformer-paraformer-module"]], "wenet.paraformer.search module": [[39, "wenet-paraformer-search-module"]], "wenet.squeezeformer package": [[40, "wenet-squeezeformer-package"]], "wenet.squeezeformer.attention module": [[41, "wenet-squeezeformer-attention-module"]], "wenet.squeezeformer.conv2d module": [[42, "wenet-squeezeformer-conv2d-module"]], "wenet.squeezeformer.convolution module": [[43, "wenet-squeezeformer-convolution-module"]], "wenet.squeezeformer.encoder module": [[44, "wenet-squeezeformer-encoder-module"]], "wenet.squeezeformer.encoder_layer module": [[45, "wenet-squeezeformer-encoder-layer-module"]], "wenet.squeezeformer.positionwise_feed_forward module": [[46, "wenet-squeezeformer-positionwise-feed-forward-module"]], "wenet.squeezeformer.subsampling module": [[47, "wenet-squeezeformer-subsampling-module"]], "wenet.transducer package": [[48, "wenet-transducer-package"]], "wenet.transducer.joint module": [[49, "wenet-transducer-joint-module"]], "wenet.transducer.predictor module": [[50, "wenet-transducer-predictor-module"]], "wenet.transducer.transducer module": [[51, "wenet-transducer-transducer-module"]], "wenet.transformer package": [[52, "wenet-transformer-package"]], "wenet.transformer.asr_model module": [[53, "wenet-transformer-asr-model-module"]], "wenet.transformer.attention module": [[54, "wenet-transformer-attention-module"]], "wenet.transformer.cmvn module": [[55, "wenet-transformer-cmvn-module"]], "wenet.transformer.convolution module": [[56, "wenet-transformer-convolution-module"]], "wenet.transformer.ctc module": [[57, "wenet-transformer-ctc-module"]], "wenet.transformer.decoder module": [[58, "wenet-transformer-decoder-module"]], "wenet.transformer.decoder_layer module": [[59, "wenet-transformer-decoder-layer-module"]], "wenet.transformer.embedding module": [[60, "wenet-transformer-embedding-module"]], "wenet.transformer.encoder module": [[61, "wenet-transformer-encoder-module"]], "wenet.transformer.encoder_layer module": [[62, "wenet-transformer-encoder-layer-module"]], "wenet.transformer.label_smoothing_loss module": [[63, "wenet-transformer-label-smoothing-loss-module"]], "wenet.transformer.positionwise_feed_forward module": [[64, "wenet-transformer-positionwise-feed-forward-module"]], "wenet.transformer.search module": [[65, "wenet-transformer-search-module"]], "wenet.transformer.subsampling module": [[66, "wenet-transformer-subsampling-module"]], "wenet.transformer.swish module": [[67, "wenet-transformer-swish-module"]], "wenet.utils package": [[68, "wenet-utils-package"]], "wenet.utils.checkpoint module": [[69, "wenet-utils-checkpoint-module"]], "wenet.utils.cmvn module": [[70, "wenet-utils-cmvn-module"]], "wenet.utils.common module": [[71, "wenet-utils-common-module"]], "wenet.utils.config module": [[72, "wenet-utils-config-module"]], "wenet.utils.context_graph module": [[73, "wenet-utils-context-graph-module"]], "wenet.utils.ctc_utils module": [[74, "wenet-utils-ctc-utils-module"]], "wenet.utils.executor module": [[75, "wenet-utils-executor-module"]], "wenet.utils.file_utils module": [[76, "wenet-utils-file-utils-module"]], "wenet.utils.init_ali_paraformer module": [[77, "wenet-utils-init-ali-paraformer-module"]], "wenet.utils.init_model module": [[78, "wenet-utils-init-model-module"]], "wenet.utils.mask module": [[79, "wenet-utils-mask-module"]], "wenet.utils.scheduler module": [[80, "wenet-utils-scheduler-module"]], "wenet.utils.tokenize_utils module": [[81, "wenet-utils-tokenize-utils-module"]], "wenet.utils.train_utils module": [[82, "wenet-utils-train-utils-module"]], "wenet.whisper package": [[83, "wenet-whisper-package"]], "wenet.whisper.convert_whisper_to_wenet_config_and_ckpt module": [[84, "wenet-whisper-convert-whisper-to-wenet-config-and-ckpt-module"]], "wenet.whisper.whisper module": [[85, "wenet-whisper-whisper-module"]], "Python Package": [[86, "python-package"]], "Install": [[86, "install"]], "Development Install": [[86, "development-install"]], "Command line Usage": [[86, "command-line-usage"]], "Python Programming Usage": [[86, "python-programming-usage"]], "Reference": [[87, "reference"]], "Runtime for WeNet": [[89, "runtime-for-wenet"]], "Platforms Supported": [[89, "platforms-supported"]], "Architecture and Implementation": [[89, "architecture-and-implementation"]], "Architecture": [[89, "architecture"]], "Interface Design": [[89, "interface-design"]], "Cache in Details": [[89, "cache-in-details"]], "References": [[89, "references"]], "How to train models?": [[90, "how-to-train-models"]], "Tutorial on AIShell": [[91, "tutorial-on-aishell"]], "First Experiment": [[91, "first-experiment"], [92, "first-experiment"]], "Stage -1: Download data": [[91, "stage-1-download-data"], [92, "stage-1-download-data"]], "Stage 0: Prepare Training data": [[91, "stage-0-prepare-training-data"], [92, "stage-0-prepare-training-data"]], "Stage 1: Extract optinal cmvn features": [[91, "stage-1-extract-optinal-cmvn-features"], [92, "stage-1-extract-optinal-cmvn-features"]], "Stage 2: Generate label token dictionary": [[91, "stage-2-generate-label-token-dictionary"], [92, "stage-2-generate-label-token-dictionary"]], "Stage 3: Prepare WeNet data format": [[91, "stage-3-prepare-wenet-data-format"], [92, "stage-3-prepare-wenet-data-format"]], "Stage 4: Neural Network training": [[91, "stage-4-neural-network-training"], [92, "stage-4-neural-network-training"]], "Stage 5: Recognize wav using the trained model": [[91, "stage-5-recognize-wav-using-the-trained-model"], [92, "stage-5-recognize-wav-using-the-trained-model"]], "Stage 6: Export the trained model": [[91, "stage-6-export-the-trained-model"]], "Tutorial on LibriSpeech": [[92, "tutorial-on-librispeech"]], "Stage 6(Optional): Export the trained model": [[92, "stage-6-optional-export-the-trained-model"]], "Stage 7(Optional): Add LM and test it with runtime": [[92, "stage-7-optional-add-lm-and-test-it-with-runtime"]]}, "indexentries": {"module": [[15, "module-wenet.cli"]], "wenet.cli": [[15, "module-wenet.cli"]]}})
\ No newline at end of file