### Install Binaries

Source: https://github.com/mozilla/deepspeech/blob/master/doc/BUILDING.rst

Installs the compiled library and binaries to the system path.

```bash
PREFIX=/usr/local sudo make install
```

--------------------------------

### Install Python Bindings

Source: https://github.com/mozilla/deepspeech/blob/master/doc/BUILDING.rst

Generates and installs the Python language bindings.

```bash
cd native_client/python
make bindings
```

--------------------------------

### Install DeepSpeech and Download Models

Source: https://github.com/mozilla/deepspeech/blob/master/doc/index.rst

Installs DeepSpeech and downloads pre-trained English model files and example audio. Ensure you have Python 3 and virtualenv installed.

```bash
# Create and activate a virtualenv
virtualenv -p python3 $HOME/tmp/deepspeech-venv/
source $HOME/tmp/deepspeech-venv/bin/activate

# Install DeepSpeech
pip3 install deepspeech

# Download pre-trained English model files
curl -LO https://github.com/mozilla/DeepSpeech/releases/download/v0.9.3/deepspeech-0.9.3-models.pbmm
curl -LO https://github.com/mozilla/DeepSpeech/releases/download/v0.9.3/deepspeech-0.9.3-models.scorer

# Download example audio files
curl -LO https://github.com/mozilla/DeepSpeech/releases/download/v0.9.3/audio-0.9.3.tar.gz
tar xvf audio-0.9.3.tar.gz

# Transcribe an audio file
deepspeech --model deepspeech-0.9.3-models.pbmm --scorer deepspeech-0.9.3-models.scorer --audio audio/2830-3980-0043.wav
```

--------------------------------

### Install System Build Tools

Source: https://github.com/mozilla/deepspeech/blob/master/doc/TRAINING.rst

Installs system-level development headers required for building Python modules like webrtcvad.

```bash
sudo apt-get install python3-dev
```

--------------------------------

### Run DeepSpeech Inference

Source: https://github.com/mozilla/deepspeech/blob/master/doc/USING.rst

Example command to perform speech-to-text on an audio file using the installed deepspeech binary. Assumes pre-trained models have been downloaded.

```bash

```

--------------------------------

### Install Header Files

Source: https://github.com/mozilla/deepspeech/blob/master/native_client/ctcdecode/third_party/openfst-1.6.9-win/src/CMakeLists.txt

Installs header files from the include/ directory to the destination include/ directory, matching files with the .h pattern.

```cmake
install(DIRECTORY include/ DESTINATION include/
        FILES_MATCHING PATTERN "*.h")
```

--------------------------------

### Install Training Dependencies

Source: https://github.com/mozilla/deepspeech/blob/master/doc/TRAINING.rst

Installs required Python packages and the DeepSpeech training package in editable mode.

```bash
cd DeepSpeech
pip3 install --upgrade pip==20.2.2 wheel==0.34.2 setuptools==49.6.0
pip3 install --upgrade -e .
```

--------------------------------

### Install MSYS2 Packages

Source: https://github.com/mozilla/deepspeech/blob/master/doc/BUILDING_DotNet.rst

Install necessary packages (patch, unzip) using pacman for MSYS2.

```bash
pacman -Syu
pacman -Su
pacman -S patch unzip
```

--------------------------------

### Example Playback with Reverberation and Volume Maximization

Source: https://github.com/mozilla/deepspeech/blob/master/doc/TRAINING.rst

This bash command uses the `play.py` script to play all samples from `test.sdb` with reverberation and maximized volume. It demonstrates applying specific augmentations during playback for testing purposes.

```bash
bin/play.py --augment reverb[p=0.1,delay=50.0,decay=2.0] --augment volume --random test.sdb
```

--------------------------------

### Check Bazel Version

Source: https://github.com/mozilla/deepspeech/blob/master/doc/BUILDING_DotNet.rst

Verify the installed Bazel version.

```bash
bazel version
```

--------------------------------

### Install KenLM Python Module

Source: https://github.com/mozilla/deepspeech/blob/master/native_client/kenlm/README.md

Install the KenLM Python module using pip. This command downloads and installs the module from the master branch of the GitHub repository.

```bash
pip install https://github.com/kpu/kenlm/archive/master.zip
```

--------------------------------

### Train Custom DeepSpeech Models

Source: https://context7.com/mozilla/deepspeech/llms.txt

Commands for setting up the training environment, installing dependencies, and executing the training process.

```bash
# Clone the repository
git clone --branch v0.9.3 https://github.com/mozilla/DeepSpeech

# Set up virtual environment
python3 -m venv $HOME/tmp/deepspeech-train-venv/
source $HOME/tmp/deepspeech-train-venv/bin/activate

# Install dependencies
cd DeepSpeech
pip3 install --upgrade pip==20.2.2 wheel==0.34.2 setuptools==49.6.0
pip3 install --upgrade -e .

# Install GPU support (optional, but recommended)
pip3 uninstall tensorflow
pip3 install 'tensorflow-gpu==1.15.4'

# Prepare your data in CSV format with columns:
# wav_filename,wav_filesize,transcript

# Basic training command
python3 DeepSpeech.py \
    --train_files train.csv \
    --dev_files dev.csv \
    --test_files test.csv \
    --epochs 75 \
    --learning_rate 0.0001 \
    --export_dir ./export

# Training with checkpointing
python3 DeepSpeech.py \
    --train_files train.csv \
    --dev_files dev.csv \
    --test_files test.csv \
    --checkpoint_dir ./checkpoints \
    --epochs 75

# Export model for TFLite (mobile/embedded)
python3 DeepSpeech.py \
    --checkpoint_dir ./checkpoints \
    --export_dir ./export \
    --export_tflite
```

--------------------------------

### Example Training with All Augmentations

Source: https://github.com/mozilla/deepspeech/blob/master/doc/TRAINING.rst

This bash script demonstrates how to enable and configure multiple audio augmentation techniques during the DeepSpeech training process. It includes parameters for overlay, reverb, resampling, codec, volume, pitch, tempo, warp, frequency mask, time mask, dropout, add, and multiply augmentations.

```bash
python -u DeepSpeech.py \
          --train_files "train.sdb" \
          --feature_cache ./feature.cache \
          --cache_for_epochs 10 \
          --epochs 100 \
          --augment overlay[p=0.5,source=noise.sdb,layers=1,snr=50:20~10] \
          --augment reverb[p=0.1,delay=50.0~30.0,decay=10.0:2.0~1.0] \
          --augment resample[p=0.1,rate=12000:8000~4000] \
          --augment codec[p=0.1,bitrate=48000:16000] \
          --augment volume[p=0.1,dbfs=-10:-40] \
          --augment pitch[p=0.1,pitch=1~0.2] \
          --augment tempo[p=0.1,factor=1~0.5] \
          --augment warp[p=0.1,nt=4,nf=1,wt=0.5:1.0,wf=0.1:0.2] \
          --augment frequency_mask[p=0.1,n=1:3,size=1:5] \
          --augment time_mask[p=0.1,domain=signal,n=3:10~2,size=50:100~40] \
          --augment dropout[p=0.1,rate=0.05] \
          --augment add[p=0.1,domain=signal,stddev=0~0.5] \
          --augment multiply[p=0.1,domain=features,stddev=0~0.5] \
          [...]
```

--------------------------------

### Install DeepSpeech Python Bindings (CPU)

Source: https://github.com/mozilla/deepspeech/blob/master/doc/USING.rst

Install the DeepSpeech package for CPU inference using pip3 after activating the virtual environment. This command also handles dependency installation.

```bash
$ pip3 install deepspeech
```

--------------------------------

### Install DeepSpeech Python package

Source: https://github.com/mozilla/deepspeech/blob/master/doc/BUILDING.rst

Install the DeepSpeech package from the local distribution directory.

```bash
pip install dist/deepspeech*
```

--------------------------------

### Install DeepSpeech Node.js Package

Source: https://context7.com/mozilla/deepspeech/llms.txt

Install the DeepSpeech Node.js bindings using npm. Supports CPU and GPU versions.

```bash
npm install deepspeech

npm install deepspeech-gpu
```

--------------------------------

### Configure fstconst Library

Source: https://github.com/mozilla/deepspeech/blob/master/native_client/ctcdecode/third_party/openfst-1.6.9-win/src/extensions/const/CMakeLists.txt

Builds the fstconst library from multiple source files and sets installation paths.

```cmake
add_library(fstconst 
  const8-fst.cc 
  const16-fst.cc 
  const64-fst.cc)
target_link_libraries(fstconst fst)
set_target_properties(fstconst PROPERTIES
  SOVERSION "${SOVERSION}"
  FOLDER constant
)

install(TARGETS fstconst 
  LIBRARY DESTINATION lib
  ARCHIVE DESTINATION lib
  RUNTIME DESTINATION lib
 )
```

--------------------------------

### Setting up a Conda Environment for Training

Source: https://github.com/mozilla/deepspeech/blob/master/doc/TRAINING.rst

This bash snippet illustrates the initial step of setting up a separate environment for training DeepSpeech using Conda. It emphasizes the importance of using a dedicated environment to prevent conflicts and ensure a stable training setup.

```bash
conda create --name deepspeech python=3.7
conda activate deepspeech
```

--------------------------------

### Run Python Linter

Source: https://github.com/mozilla/deepspeech/blob/master/CONTRIBUTING.rst

Install and execute the project's linter to check for style issues and mistakes in Python code.

```bash
pip install pylint cardboardlint
cardboardlinter --refspec master
```

--------------------------------

### Configure Augmentation Flags

Source: https://github.com/mozilla/deepspeech/blob/master/doc/TRAINING.rst

Syntax for applying augmentation types and an example using the overlay augmentation.

```bash
--augment augmentation_type1[param1=value1,param2=value2,...] --augment augmentation_type2[param1=value1,param2=value2,...] ...
```

```bash
python3 DeepSpeech.py --augment overlay[p=0.1,source=/path/to/audio.sdb,snr=20.0] ...
```

--------------------------------

### Install DeepSpeech Python Package

Source: https://context7.com/mozilla/deepspeech/llms.txt

Install the DeepSpeech Python package for CPU or GPU inference. Requires Python 3 and pip. Also downloads pre-trained model files.

```bash
virtualenv -p python3 $HOME/tmp/deepspeepspeech-venv/
source $HOME/tmp/deepspeech-venv/bin/activate

pip3 install deepspeech

pip3 install deepspeech-gpu

wget https://github.com/mozilla/DeepSpeech/releases/download/v0.9.3/deepspeech-0.9.3-models.pbmm
wget https://github.com/mozilla/DeepSpeech/releases/download/v0.9.3/deepspeech-0.9.3-models.scorer
```

--------------------------------

### Install Node.js DeepSpeech GPU Package

Source: https://github.com/mozilla/deepspeech/blob/master/doc/USING.rst

For Linux users with a supported NVIDIA GPU, install the GPU-specific package using npm. Ensure CUDA dependencies are met.

```bash
npm install deepspeech-gpu
```

--------------------------------

### Build CTC decoder package

Source: https://github.com/mozilla/deepspeech/blob/master/doc/BUILDING.rst

Compile the CTC decoder bindings with specified parallelism and install the resulting wheel.

```bash
cd native_client/ctcdecode
make bindings NUM_PROCESSES=8
pip install dist/*.whl
```

--------------------------------

### Install Node.js DeepSpeech Package

Source: https://github.com/mozilla/deepspeech/blob/master/doc/USING.rst

Use npm to install the Node.js bindings for DeepSpeech. Supports Node.js versions 4-13 and Electron.JS versions 1.6-7.1.

```bash
npm install deepspeech
```

--------------------------------

### Install DeepSpeech Python Bindings (GPU)

Source: https://github.com/mozilla/deepspeech/blob/master/doc/USING.rst

Install the GPU-accelerated version of DeepSpeech for Linux systems with a supported NVIDIA GPU. Ensure CUDA dependencies are met.

```bash
$ pip3 install deepspeech-gpu
```

--------------------------------

### Configure FST Linear Executables

Source: https://github.com/mozilla/deepspeech/blob/master/native_client/ctcdecode/third_party/openfst-1.6.9-win/src/extensions/linear/CMakeLists.txt

Builds and installs command-line tools for linear FST operations if binary support is enabled.

```cmake
if(HAVE_BIN)
  add_executable(fstlinear
    fstlinear.cc)
  target_link_libraries(fstlinear
    fstlinearscript
    fstscript 
    fst
    ${CMAKE_DL_LIBS}
 )

  add_executable(fstloglinearapply
    fstloglinearapply.cc)
  target_link_libraries(fstloglinearapply
    fstlinearscript
    fstscript 
    fst
    ${CMAKE_DL_LIBS}
  )
  install(TARGETS fstlinear fstloglinearapply
    RUNTIME DESTINATION bin
  )
  set_target_properties(fstlinear fstloglinearapply PROPERTIES
    FOLDER linear/bin
  )
endif(HAVE_BIN)
```

--------------------------------

### Add fstngram Library

Source: https://github.com/mozilla/deepspeech/blob/master/native_client/ctcdecode/third_party/openfst-1.6.9-win/src/extensions/ngram/CMakeLists.txt

Defines and configures the 'fstngram' static library, linking it against the 'fst' library and setting its installation properties.

```cmake
file(GLOB HEADER_FILES ../../include/fst/extensions/ngram/*.h)
message(STATUS "${HEADER_FILES}")

add_library(fstngram
    bitmap-index.cc 
    ngram-fst.cc 
    nthbit.cc
    ${HEADER_FILES}
)

target_link_libraries(fstngram
    fst
)

set_target_properties(fstngram PROPERTIES
  SOVERSION "${SOVERSION}"
  FOLDER ngram
)

install(TARGETS fstngram 
	LIBRARY DESTINATION lib
	ARCHIVE DESTINATION lib
	RUNTIME DESTINATION lib
)
```

--------------------------------

### Configure Git Pre-commit Hook for Linting

Source: https://github.com/mozilla/deepspeech/blob/master/CONTRIBUTING.rst

Installs a pre-commit hook that stashes uncommitted changes, runs cardboardlinter on the staged changes, and restores the working tree.

```bash
cat <<\EOF > .git/hooks/pre-commit
#!/bin/bash
if [ ! -x "$(command -v cardboardlinter)" ]; then
    exit 0
fi

# First, stash index and work dir, keeping only the
# to-be-committed changes in the working directory.
echo "Stashing working tree changes..." 1>&2
old_stash=$(git rev-parse -q --verify refs/stash)
git stash save -q --keep-index
new_stash=$(git rev-parse -q --verify refs/stash)

# If there were no changes (e.g., `--amend` or `--allow-empty`)
# then nothing was stashed, and we should skip everything,
# including the tests themselves.  (Presumably the tests passed
# on the previous commit, so there is no need to re-run them.)
if [ "$old_stash" = "$new_stash" ]; then
    echo "No changes, skipping lint." 1>&2
    exit 0
fi

# Run tests
cardboardlinter --refspec HEAD -n auto
status=$?

# Restore changes
echo "Restoring working tree changes..." 1>&2
git reset --hard -q && git stash apply --index -q && git stash drop -q

# Exit with status from test-run: nonzero prevents commit
exit $status
EOF
chmod +x .git/hooks/pre-commit
```

--------------------------------

### Simulate Codec Augmentation at Epoch Start and End

Source: https://github.com/mozilla/deepspeech/blob/master/doc/TRAINING.rst

These bash commands simulate the codec augmentation on a WAV file using the `play.py` script. The first command applies the augmentation at the beginning of an epoch (clock=0.0), and the second applies it at the end (clock=1.0), showing how to control the timing of augmentations.

```bash
bin/play.py --augment codec[p=0.1,bitrate=48000:16000] --clock 0.0 test.wav
```

```bash
bin/play.py --augment codec[p=0.1,bitrate=48000:16000] --clock 1.0 test.wav
```

--------------------------------

### Get TensorFlow Version

Source: https://github.com/mozilla/deepspeech/blob/master/ISSUE_TEMPLATE.md

Use this command to retrieve the installed TensorFlow version. Ensure TensorFlow is installed in your Python environment.

```bash
python -c "import tensorflow as tf; print(tf.GIT_VERSION, tf.VERSION)"
```

--------------------------------

### Run Sample Training Script

Source: https://github.com/mozilla/deepspeech/blob/master/doc/TRAINING.rst

Execute the convenience script to train on the small LDC93S1 sample dataset.

```bash
./bin/run-ldc93s1.sh
```

--------------------------------

### Create and Load DeepSpeech Model Instance (C)

Source: https://github.com/mozilla/deepspeech/blob/master/doc/C-Examples.rst

Demonstrates how to create a DeepSpeech model instance and load a pre-trained model file. Ensure the model file path is correct.

```c
#include "deepspeech.h"
#include <stdio.h>
#include <stdlib.h>

int main(int argc, char **argv) {
    SpeechRecognitionContext *ctx = deepspeech_init("deepspeech-0.9.3-models.pb", "deepspeech-0.9.3-models.scorer");
    if (!ctx) {
        fprintf(stderr, "Failed to initialize DeepSpeech context!\n");
        return 1;
    }

    printf("DeepSpeech context initialized successfully.\n");

    // ... rest of the inference code ...

    // Free the context when done
    // deepspeech_free(ctx);

    return 0;
}

```

--------------------------------

### Initialize and use ThreadPool in C++

Source: https://github.com/mozilla/deepspeech/blob/master/native_client/ctcdecode/third_party/ThreadPool/README.md

Demonstrates creating a thread pool with a specified number of threads and retrieving results using futures.

```c++
// create thread pool with 4 worker threads
ThreadPool pool(4);

// enqueue and store future
auto result = pool.enqueue([](int answer) { return answer; }, 42);

// get result from future
std::cout << result.get() << std::endl;
```

--------------------------------

### Create Virtual Environment

Source: https://github.com/mozilla/deepspeech/blob/master/doc/TRAINING.rst

Initializes a new Python virtual environment in the specified directory.

```bash
$ python3 -m venv $HOME/tmp/deepspeech-train-venv/
```

--------------------------------

### Generate Dockerfile for Building from Source

Source: https://github.com/mozilla/deepspeech/blob/master/doc/USING.rst

Use the make command to generate the Dockerfile.build for automatically building DeepSpeech components from source.

```bash
make Dockerfile.build
```

--------------------------------

### TensorFlow Error Log Example

Source: https://github.com/mozilla/deepspeech/blob/master/doc/TRAINING.rst

Example of a common error message encountered when cuDNN fails to initialize during training.

```text
tensorflow.python.framework.errors_impl.UnknownError: Failed to get convolution algorithm. This is probably because cuDNN failed to initialize, so try looking to see if a warning log message was printed above.
        [[{{node tower_0/conv1d/Conv2D}}]]
```

--------------------------------

### Model Initialization and Configuration

Source: https://context7.com/mozilla/deepspeech/llms.txt

Methods for loading the acoustic model and configuring inference parameters like beam width.

```APIDOC
## Model(model_path)

### Description
Initializes a new DeepSpeech model instance from a .pbmm or .tflite file.

### Parameters
#### Request Body
- **model_path** (string) - Required - Path to the acoustic model file.

## setBeamWidth(beam_width)

### Description
Sets the beam width for the decoder. Larger values improve accuracy but increase computation time.

### Parameters
#### Request Body
- **beam_width** (int) - Required - The beam width value.
```

--------------------------------

### Build Native Client for CPU

Source: https://github.com/mozilla/deepspeech/blob/master/doc/BUILDING_DotNet.rst

Compile the native client library for CPU support, including AVX/AVX2 optimizations. Ensure your CPU supports these instructions.

```bash
bazel build --workspace_status_command="bash native_client/bazel_workspace_status_cmd.sh" -c opt --copt=/arch:AVX --copt=/arch:AVX2 //native_client:libdeepspeech.so
```

--------------------------------

### Install DeepSpeech GPU Version and Transcribe Audio

Source: https://github.com/mozilla/deepspeech/blob/master/doc/index.rst

Installs the GPU-enabled DeepSpeech package and demonstrates how to transcribe an audio file using a supported NVIDIA GPU. Requires CUDA dependencies.

```bash
# Create and activate a virtualenv
virtualenv -p python3 $HOME/tmp/deepspeech-gpu-venv/
source $HOME/tmp/deepspeech-gpu-venv/bin/activate

# Install DeepSpeech CUDA enabled package
pip3 install deepspeech-gpu

# Transcribe an audio file.
deepspeech --model deepspeech-0.9.3-models.pbmm --scorer deepspeech-0.9.3-models.scorer --audio audio/2830-3980-0043.wav
```

--------------------------------

### Cross-build for LePotato ARM64

Source: https://github.com/mozilla/deepspeech/blob/master/doc/BUILDING.rst

Build the DeepSpeech library for LePotato ARM64 using Bazel.

```bash
bazel build --workspace_status_command="bash native_client/bazel_workspace_status_cmd.sh" --config=monolithic --config=rpi3-armv8 --config=rpi3-armv8_opt -c opt --copt=-O3 --copt=-fvisibility=hidden //native_client:libdeepspeech.so
```

--------------------------------

### Configure TensorFlow for GPU

Source: https://github.com/mozilla/deepspeech/blob/master/doc/TRAINING.rst

Switches the standard TensorFlow installation to the GPU-enabled version.

```bash
pip3 uninstall tensorflow
pip3 install 'tensorflow-gpu==1.15.4'
```

--------------------------------

### Enable Testing and Add Test Subdirectory

Source: https://github.com/mozilla/deepspeech/blob/master/native_client/ctcdecode/third_party/openfst-1.6.9-win/src/CMakeLists.txt

Enables the testing framework and includes the 'test' subdirectory for test-related build configurations.

```cmake
enable_testing()
add_subdirectory(test)
```

--------------------------------

### Display DeepSpeech Help

Source: https://github.com/mozilla/deepspeech/blob/master/doc/TRAINING.rst

Access the full list of command line options for the training script.

```bash
python3 DeepSpeech.py --helpfull
```

--------------------------------

### Compile KenLM with CMake

Source: https://github.com/mozilla/deepspeech/blob/master/native_client/kenlm/README.md

Standard build procedure using CMake to prepare the library.

```bash
mkdir -p build
cd build
cmake ..
make -j 4
```

--------------------------------

### Generate Javadoc with SWIG

Source: https://github.com/mozilla/deepspeech/blob/master/native_client/java/libdeepspeech/src/main/java/org/deepspeech/libdeepspeech_doc/README.rst

Run this command from the native_client/java directory to generate Javadoc documentation. Requires SWIG 4.0 or higher.

```bash
swig -c++ -java -doxygen -package org.deepspeech.libdeepspeech -outdir libdeepspeech/src/main/java/org/deepspeech/libdeepspeech_doc -o jni/deepspeech_wrap.cpp jni/deepspeech.i
```

--------------------------------

### Upgrade DeepSpeech Python Bindings (CPU)

Source: https://github.com/mozilla/deepspeech/blob/master/doc/USING.rst

Update an existing DeepSpeech installation to the latest version using pip3.

```bash
$ pip3 install --upgrade deepspeech
```

--------------------------------

### DeepSpeech Inference Error Output

Source: https://github.com/mozilla/deepspeech/wiki/Finding-TensorFlow-op-kernel-targets-when-you-add-new-operations-to-the-graph

Example of the error message generated when a required TensorFlow operation is missing from the binary.

```bash
$ ./deepspeech --model ../ldc93s1_export/output_graph.pb --alphabet ../data/alphabet.txt --audio ../data/ldc93s1/LDC93S1.wav
TensorFlow: v1.14.0-14-g1aad02a78e
DeepSpeech: v0.6.0-alpha.4-7-gff8f405
Warning: reading entire model file into memory. Transform model file into an mmapped graph to reduce heap usage.
2019-07-17 09:57:35.976319: I tensorflow/core/platform/cpu_feature_guard.cc:142] Your CPU supports instructions that this TensorFlow binary was not compiled to use: SSE4.2 AVX AVX2 FMA
Invalid argument: No OpKernel was registered to support Op 'Split' used by {{node cudnn_lstm/rnn/cond/rnn/multi_rnn_cell/cell_0/cudnn_compatible_lstm_cell/split}}with these attrs: [T=DT_FLOAT, num_split=4]
Registered devices: [CPU]
Registered kernels:
  <no registered kernels>

     [[cudnn_lstm/rnn/cond/rnn/multi_rnn_cell/cell_0/cudnn_compatible_lstm_cell/split]]
Could not create model.
```

--------------------------------

### Build libdeepspeech.so for Android

Source: https://github.com/mozilla/deepspeech/blob/master/doc/BUILDING.rst

Commands to build the shared library for ARM and ARM64 architectures using Bazel.

```bash
bazel build --workspace_status_command="bash native_client/bazel_workspace_status_cmd.sh" --config=monolithic --config=android --config=android_arm --define=runtime=tflite --action_env ANDROID_NDK_API_LEVEL=21 --cxxopt=-std=c++14 --copt=-D_GLIBCXX_USE_C99 //native_client:libdeepspeech.so
```

```bash
bazel build --workspace_status_command="bash native_client/bazel_workspace_status_cmd.sh" --config=monolithic --config=android --config=android_arm64 --define=runtime=tflite --action_env ANDROID_NDK_API_LEVEL=21 --cxxopt=-std=c++14 --copt=-D_GLIBCXX_USE_C99 //native_client:libdeepspeech.so
```

--------------------------------

### Configure TensorFlow

Source: https://github.com/mozilla/deepspeech/blob/master/doc/BUILDING.rst

Prepares the TensorFlow environment for the build process.

```bash
cd tensorflow
./configure
```

--------------------------------

### Upgrade DeepSpeech Python Bindings (GPU)

Source: https://github.com/mozilla/deepspeech/blob/master/doc/USING.rst

Update an existing GPU-specific DeepSpeech installation to the latest version using pip3.

```bash
$ pip3 install --upgrade deepspeech-gpu
```

--------------------------------

### Download DeepSpeech model files

Source: https://github.com/mozilla/deepspeech/blob/master/doc/USING.rst

Use wget to download the required .pbmm model and .scorer files into the current directory.

```bash
wget https://github.com/mozilla/DeepSpeech/releases/download/v0.9.3/deepspeech-0.9.3-models.pbmm
wget https://github.com/mozilla/DeepSpeech/releases/download/v0.9.3/deepspeech-0.9.3-models.scorer
```

--------------------------------

### Create and Load DeepSpeech Model Instance (.NET)

Source: https://github.com/mozilla/deepspeech/blob/master/doc/DotNet-Examples.rst

Instantiates a DeepSpeech model and loads the specified model file. Ensure the model file path is correct.

```csharp
var model = new Model("model.pb");
model.LoadStateDict("model.pb");
model.SetScorer("deepspeech_0.6.0-models.scorer");
```

--------------------------------

### Create Model Instance and Load Model in Java

Source: https://github.com/mozilla/deepspeech/blob/master/doc/Java-Examples.rst

Instantiate a DeepSpeech model and load the pre-trained model files. Ensure the model and scorer files are accessible.

```java
DeepSpeechModel model = new DeepSpeechModel("model.pbmm");
model.init("alphabet.txt", "output_graph.pb", 1, 1);
model.loadModel("model.pbmm", "model.scorer");
```

--------------------------------

### Add ngram_fst Module

Source: https://github.com/mozilla/deepspeech/blob/master/native_client/ctcdecode/third_party/openfst-1.6.9-win/src/extensions/ngram/CMakeLists.txt

Defines and configures the 'ngram_fst' shared module, linking it against the 'fst' library and setting its installation properties for Windows.

```cmake
add_library(ngram_fst MODULE
    bitmap-index.cc 
    ngram-fst.cc 
    nthbit.cc
)

set_target_properties(ngram_fst PROPERTIES
    WINDOWS_EXPORT_ALL_SYMBOLS true
    FOLDER ngram/modules
)

target_link_libraries(ngram_fst
    fst
)

install(TARGETS ngram_fst
    LIBRARY DESTINATION lib/fst
)
```

--------------------------------

### Configure FST Linear Library

Source: https://github.com/mozilla/deepspeech/blob/master/native_client/ctcdecode/third_party/openfst-1.6.9-win/src/extensions/linear/CMakeLists.txt

Defines the library build process for linear scripts, including header file globbing and installation paths.

```cmake
file(GLOB HEADER_FILES ../../include/fst/extensions/linear/*.h)
message(STATUS "${HEADER_FILES}")


if(HAVE_SCRIPT)
  add_library(fstlinearscript
    linearscript.cc
    ${HEADER_FILES}
  )
  target_link_libraries(fstlinearscript
    fstscript
    fst
  )
  set_target_properties(fstlinearscript PROPERTIES 
    SOVERSION "${SOVERSION}"
    FOLDER linear
  )
  
  install(TARGETS fstlinearscript
	LIBRARY DESTINATION lib
	ARCHIVE DESTINATION lib
	RUNTIME DESTINATION lib
  )
endif(HAVE_SCRIPT)
```

--------------------------------

### Build Native Client with CUDA Support

Source: https://github.com/mozilla/deepspeech/blob/master/doc/BUILDING_DotNet.rst

Compile the native client library with CUDA support. This requires CUDA to be enabled during the configure.py step.

```bash
bazel build --workspace_status_command="bash native_client/bazel_workspace_status_cmd.sh" -c opt --config=cuda --copt=/arch:AVX --copt=/arch:AVX2 //native_client:libdeepspeech.so
```

--------------------------------

### Define fstlookahead library in CMake

Source: https://github.com/mozilla/deepspeech/blob/master/native_client/ctcdecode/third_party/openfst-1.6.9-win/src/extensions/lookahead/CMakeLists.txt

Configures the fstlookahead library with its source files, links the fst dependency, and sets target properties for installation.

```cmake
add_library(fstlookahead
  arc_lookahead-fst.cc 
  ilabel_lookahead-fst.cc
  olabel_lookahead-fst.cc
)
target_link_libraries(fstlookahead fst)
set_target_properties(fstlookahead PROPERTIES 
  SOVERSION "${SOVERSION}"
  FOLDER lookahead  
)

install(TARGETS fstlookahead 
	LIBRARY DESTINATION lib
	ARCHIVE DESTINATION lib
	RUNTIME DESTINATION lib
)
```

--------------------------------

### Define FST Module Helper Function

Source: https://github.com/mozilla/deepspeech/blob/master/native_client/ctcdecode/third_party/openfst-1.6.9-win/src/extensions/special/CMakeLists.txt

Provides a reusable function to add FST modules with standard linking, property configuration, and installation paths.

```cmake
function (add_module _name)
    add_library(${ARGV})
    if (TARGET ${_name})
        target_link_libraries(${_name} fst)
        set_target_properties(${_name} 
          PROPERTIES WINDOWS_EXPORT_ALL_SYMBOLS true
          FOLDER special/modules
        )
    endif()

    install(TARGETS ${_name} LIBRARY DESTINATION lib/fst)
endfunction()

add_module(phi-fst MODULE phi-fst.cc)
add_module(rho-fst MODULE rho-fst.cc)
add_module(sigma-fst MODULE sigma-fst.cc)
```

--------------------------------

### Define fstfar Library

Source: https://github.com/mozilla/deepspeech/blob/master/native_client/ctcdecode/third_party/openfst-1.6.9-win/src/extensions/far/CMakeLists.txt

Defines the fstfar library by compiling C++ source files and linking against the fst library. Installs the library to the 'lib' directory.

```cmake
file(GLOB HEADER_FILES ../../include/fst/extensions/far/*.h)
message(STATUS "${HEADER_FILES}")

add_library(fstfar
  sttable.cc
  stlist.cc
  ${HEADER_FILES}
)
target_link_libraries(fstfar fst)
set_target_properties(fstfar PROPERTIES 
  SOVERSION "${SOVERSION}"
  FOLDER far
)

install(TARGETS fstfar
  LIBRARY DESTINATION lib
  ARCHIVE DESTINATION lib
  RUNTIME DESTINATION lib
)
```

--------------------------------

### Run DeepSpeech Command-Line Client

Source: https://github.com/mozilla/deepspeech/blob/master/doc/USING.rst

Execute the DeepSpeech C++ command-line client with specified model, scorer, and audio file. Assumes pre-trained models are downloaded.

```bash
./deepspeech --model deepspeech-0.9.3-models.pbmm --scorer deepspeech-0.9.3-models.scorer --audio audio_input.wav
```

--------------------------------

### Build CTC decoder with custom configuration

Source: https://github.com/mozilla/deepspeech/blob/master/doc/BUILDING.rst

Build the CTC decoder using custom SWIG distribution and platform settings.

```bash
SWIG_DIST_URL=[...] PYTHON_PLATFORM_NAME=[...] make bindings
pip install dist/*.whl
```

--------------------------------

### Speech-to-Text with Metadata (Python)

Source: https://context7.com/mozilla/deepspeech/llms.txt

Get detailed transcription results including per-token timing information and confidence scores. Supports multiple candidate transcripts.

```python
from deepspeech import Model
import numpy as np
import wave
import json

ds = Model('deepspeech-0.9.3-models.pbmm')
ds.enableExternalScorer('deepspeech-0.9.3-models.scorer')

fin = wave.open('audio.wav', 'rb')
audio = np.frombuffer(fin.readframes(fin.getnframes()), np.int16)
fin.close()

# Get metadata with multiple candidate transcripts
num_candidates = 3
metadata = ds.sttWithMetadata(audio, num_candidates)


```

--------------------------------

### Generate Initial Scorer Package

Source: https://github.com/mozilla/deepspeech/blob/master/doc/Scorer.rst

Use this command to create an initial scorer package. The alpha and beta values used here will be overridden by lm_optimizer.py.

```bash
generate_scorer_package --alphabet_config_path alphabet.config --vocab_list_file vocab.txt --package_path lm.binary --default_alpha 0.0 --default_beta 0.0
```

--------------------------------

### Build Language Bindings

Source: https://github.com/mozilla/deepspeech/blob/master/doc/BUILDING.rst

Uses the Makefile to compile language-specific bindings.

```bash
cd ../DeepSpeech/native_client
make deepspeech
```

--------------------------------

### Define fstfarscript Library

Source: https://github.com/mozilla/deepspeech/blob/master/native_client/ctcdecode/third_party/openfst-1.6.9-win/src/extensions/far/CMakeLists.txt

Defines the fstfarscript library if HAVE_SCRIPT is enabled. It compiles specific C++ files and links against fstfar, fstscript, and fst. Installs the library to the 'lib' directory.

```cmake
if(HAVE_SCRIPT)
  add_library(fstfarscript
    far-class.cc 
    farscript.cc
    getters.cc 
    script-impl.cc
    strings.cc
  )
  target_link_libraries(fstfarscript fstfar fstscript fst)
  set_target_properties(fstfarscript PROPERTIES 
    SOVERSION "${SOVERSION}"
    FOLDER far
  )

  install(TARGETS fstfarscript
	LIBRARY DESTINATION lib
	ARCHIVE DESTINATION lib
	RUNTIME DESTINATION lib
  )
endif(HAVE_SCRIPT)
```

--------------------------------

### Custom Executable Addition Function

Source: https://github.com/mozilla/deepspeech/blob/master/native_client/ctcdecode/third_party/openfst-1.6.9-win/src/extensions/pdt/CMakeLists.txt

Defines a custom function `add_executable2` to simplify adding executables. This function adds an executable, links it to necessary libraries, sets its folder property, and installs it.

```cmake
function (add_executable2 _name)
    add_executable(${ARGV})
    if (TARGET ${_name})
        target_link_libraries(${_name} fstpdtscript fstscript fst ${CMAKE_DL_LIBS})
        set_target_properties(${_name} PROPERTIES
          FOLDER pdt/bin
        )
    endif()
    install(TARGETS ${_name} RUNTIME DESTINATION bin)
endfunction()
```

--------------------------------

### Generate Dockerfile for Training

Source: https://github.com/mozilla/deepspeech/blob/master/doc/TRAINING.rst

Generates the Dockerfile for training, optionally specifying a custom repository or branch.

```bash
make Dockerfile.train
```

```bash
make Dockerfile.train DEEPSPEECH_REPO=git://your/fork DEEPSPEECH_SHA=origin/your-branch
```

--------------------------------

### Create and Load Model Instance - JavaScript

Source: https://github.com/mozilla/deepspeech/blob/master/doc/NodeJS-Examples.rst

Instantiates a DeepSpeech model and loads the model files. Ensure the model and grammar files are correctly path-ed.

```javascript
const model = new DeepSpeech.Model(modelPath);
model.init();
model.loadGrammar(grammarPath);

```

--------------------------------

### Define Executable Helper Function

Source: https://github.com/mozilla/deepspeech/blob/master/native_client/ctcdecode/third_party/openfst-1.6.9-win/src/extensions/far/CMakeLists.txt

A CMake function to simplify adding executables, linking them with necessary libraries, setting their folder property, and installing them to the 'bin' directory. This is used when HAVE_BIN is enabled.

```cmake
if(HAVE_BIN)
  function (add_executable2 _name)
      add_executable(${ARGV})
      if (TARGET ${_name})
          target_link_libraries(${_name} fstfarscript fstscript fst ${CMAKE_DL_LIBS})
          set_target_properties(${_name} PROPERTIES FOLDER far/bin)
      endif()
      install(TARGETS ${_name} RUNTIME DESTINATION bin)
  endfunction()

  add_executable2(farcompilestrings farcompilestrings.cc)
  add_executable2(farcreate  farcreate.cc)
  add_executable2(farequal  farequal.cc)
  add_executable2(farextract  farextract.cc)
  add_executable2(farinfo  farinfo.cc)
  add_executable2(farisomorphic  farisomorphic.cc)
  add_executable2(farprintstrings  farprintstrings.cc)
endif(HAVE_BIN)
```

--------------------------------

### Define Executable Target with Dependencies

Source: https://github.com/mozilla/deepspeech/blob/master/native_client/ctcdecode/third_party/openfst-1.6.9-win/src/bin/CMakeLists.txt

This CMake function defines an executable target, links it with specified libraries (fstscript, fst, CMAKE_DL_LIBS), and sets its folder property. It also handles the installation of the executable.

```cmake
function (add_executable2 _name)
    add_executable(${ARGV})
    if (TARGET ${_name})
        target_link_libraries(${_name} fstscript fst ${CMAKE_DL_LIBS})
        set_target_properties(${_name} PROPERTIES FOLDER bin)
    endif()

    install(TARGETS ${_name} RUNTIME DESTINATION bin)
endfunction()
```

--------------------------------

### Define and Link Compression Library

Source: https://github.com/mozilla/deepspeech/blob/master/native_client/ctcdecode/third_party/openfst-1.6.9-win/src/extensions/compress/CMakeLists.txt

Defines the `fstcompressscript` library using C++ source files and header files. It links against `fstscript`, `fst`, and ZLIB. The library is versioned and installed to the 'lib' directory.

```cmake
file(GLOB HEADER_FILES ../../include/fst/extensions/compress/*.h)
message(STATUS "${HEADER_FILES}")

add_library(fstcompressscript
  compress-script.cc
  ${HEADER_FILES}
 )

target_link_libraries(fstcompressscript 
  fstscript
  fst
  ${ZLIBS}
)
set_target_properties(fstcompressscript PROPERTIES
  SOVERSION "10"
)
install(TARGETS fstcompressscript 
  LIBRARY DESTINATION lib
  ARCHIVE DESTINATION lib
  RUNTIME DESTINATION lib
 )
```

--------------------------------

### Create Pre-Augmented Test Set

Source: https://github.com/mozilla/deepspeech/blob/master/doc/TRAINING.rst

This bash command uses the `data_set_tool.py` script to create a pre-augmented test set. It applies overlay and resampling augmentations to `test.sdb` and saves the result to `test-augmented.sdb`, useful for preparing data for training or evaluation.

```bash
bin/data_set_tool.py \
          --augment overlay[source=noise.sdb,layers=1,snr=20~10] \
          --augment resample[rate=12000:8000~4000] \
          test.sdb test-augmented.sdb
```

--------------------------------

### Create and Configure Release Branch

Source: https://github.com/mozilla/deepspeech/wiki/Release-checklist

Commands to create a new release branch and update CI configuration files to reference the new branch.

```bash
$ cd DeepSpeech
$ git checkout master && git pull
$ git checkout -b r0.8
$ git push origin r0.8
$ git checkout -b add-r0.8
$ for i in .taskcluster.yml taskcluster/github-events.cyml; do sed -i.bak -e 's/- master/- r0.8/' $i; done
$ git add .taskcluster.yml taskcluster/github-events.cyml
$ git commit -m "Switch CI to r0.8"
$ git push origin add-r0.8
```

--------------------------------

### Create DeepSpeech Model Instance (Python)

Source: https://context7.com/mozilla/deepspeech/llms.txt

Load a DeepSpeech model from a .pbmm file to perform speech-to-text inference. Optionally set beam width and retrieve model properties.

```python
from deepspeech import Model
import numpy as np
import wave

# Load the acoustic model
ds = Model('deepspeech-0.9.3-models.pbmm')

# Optionally set beam width (larger values give better results but slower)
ds.setBeamWidth(500)

# Get the sample rate expected by the model (typically 16000 Hz)
sample_rate = ds.sampleRate()
print(f"Model expects {sample_rate}Hz audio")

# Get current beam width
beam_width = ds.beamWidth()
print(f"Beam width: {beam_width}")
```

--------------------------------

### Cross-build for RPi3 ARMv7

Source: https://github.com/mozilla/deepspeech/blob/master/doc/BUILDING.rst

Build the DeepSpeech library for Raspberry Pi 3 using Bazel.

```bash
bazel build --workspace_status_command="bash native_client/bazel_workspace_status_cmd.sh" --config=monolithic --config=rpi3 --config=rpi3_opt -c opt --copt=-O3 --copt=-fvisibility=hidden //native_client:libdeepspeech.so
```

--------------------------------

### Distributed Training with Horovod

Source: https://github.com/mozilla/deepspeech/blob/master/doc/TRAINING.rst

Command to initiate distributed training across multiple machines and GPUs using Horovod. Ensure homogenous hardware and software configurations across all nodes for stability.

```bash
horovodrun -np 16 -H server1:4,server2:4,server3:4,server4:4 python3 DeepSpeech.py --train_files [...] --horovod
```

--------------------------------

### Define and Message Header Files

Source: https://github.com/mozilla/deepspeech/blob/master/native_client/ctcdecode/third_party/openfst-1.6.9-win/src/extensions/pdt/CMakeLists.txt

This snippet defines a list of header files and prints their paths using the message command.

```cmake
file(GLOB HEADER_FILES ../../include/fst/extensions/pdt/*.h)
message(STATUS "${HEADER_FILES}")
```