Installation¶

Jaeger can be installed via Bioconda, PyPI, git, or Apptainer/Singularity. Choose the method that best fits your environment.

Quick reference¶

Method	Best for	GPU support	Command
Bioconda	HPC / reproducible environments	Yes	`mamba install jaeger-bio`
PyPI	Development / latest features	Yes	`pip install jaeger-bio[gpu]`
Git	Contributors / bleeding edge	Yes	`pip install ".[gpu]"`
Apptainer	Containerized HPC workflows	Yes	`apptainer build …`

Bioconda¶

The simplest way to install Jaeger on most systems. GPU support requires the CUDA Toolkit and cuDNN to be accessible to conda.

# Add required channels (once)
conda config --add channels defaults
conda config --add channels bioconda
conda config --add channels conda-forge
conda config --set channel_priority strict

# Create environment and install
mamba create -n jaeger -c nvidia -c conda-forge cuda-nvcc "python>=3.11,<3.13" pip jaeger-bio

# Activate
conda activate jaeger

# Verify installation
jaeger health

PyPI¶

Recommended for users who want the latest stable release or need to install into an existing Python environment.

# Create a conda environment
mamba create -n jaeger -c nvidia -c conda-forge cuda-nvcc "python>=3.11,<3.13" pip
conda activate jaeger

# Or use venv
python3 -m venv jaeger
source jaeger/bin/activate

# Install with GPU support
pip install jaeger-bio[gpu]

# Or CPU-only
pip install jaeger-bio[cpu]

# Or on Apple Silicon (Mac ARM)
pip install jaeger-bio[darwin-arm]

# Verify installation
jaeger health

Git¶

For developers, contributors, or anyone who needs the very latest code from the main branch.

# Clone the repository
git clone https://github.com/MGXlab/Jaeger.git
cd Jaeger

# Create a conda environment
mamba create -n jaeger -c nvidia -c conda-forge cuda-nvcc "python>=3.11,<3.13" pip
conda activate jaeger

# Or use venv
python3 -m venv jaeger
source jaeger/bin/activate

# Install in editable mode
pip install -e ".[gpu]"   # GPU
pip install -e ".[cpu]"   # CPU
pip install -e ".[darwin-arm]"  # Mac ARM

Apptainer¶

For HPC clusters or environments where you prefer a self-contained container.

# Download container definition and config
wget -O jaeger_singularity.def https://raw.githubusercontent.com/Yasas1994/Jaeger/main/singularity/jaeger_singularity.def
wget -O config.json https://raw.githubusercontent.com/Yasas1994/Jaeger/main/src/jaeger/data/config.json

# Build the container
apptainer build jaeger.sif singularity/jaeger_singularity.def

# Test the container
apptainer run --nv jaeger.sif jaeger --help
apptainer run --nv jaeger.sif jaeger health

# List available models
apptainer run --nv jaeger.sif jaeger download --list

# Download a model
apptainer run --nv jaeger.sif jaeger download \
  --model_name jaeger_57341_1.5M_fragment \
  --path /path/to/save/model \
  --config /path/to/config.json

# Run prediction
apptainer run --nv jaeger.sif jaeger predict \
  --model jaeger_57341_1.5M_fragment \
  --config /path/to/config.json \
  -i /path/to/input.fasta \
  -o /path/to/save/results

Troubleshooting¶

Jaeger fails to detect the GPU¶

Check CUDA modules (HPC only):
```
module avail
module load cuda/12.0.0
```
Verify NVIDIA drivers:
```
nvidia-smi
```

Manually install CUDA toolkit (if the above fails):

# Create environment
conda create -n jaeger -c conda-forge -c bioconda -c defaults "python>=3.11,<3.13" pip

# Install CUDA and cuDNN
conda install -n jaeger -c "nvidia/label/cuda-11.8.0" cudatoolkit=11
conda install -n jaeger -c conda-forge cudnn

# Install Jaeger
conda install -n jaeger -c conda-forge -c bioconda jaeger-bio
conda activate jaeger

See the TensorFlow installation guide for more details on GPU setup.