Skip to content

CLI Usage Guide

Overview

DNALLM provides two ways to use the command-line interface: 1. After package installation: Using the dnallm-* commands 2. Development environment: Running directly from the project root

Usage After Installation

After installing the DNALLM package, you can use the following commands:

# Training
dnallm-train --config config.yaml
dnallm-train --model model_name --data data_path --output output_dir

# Run inference
dnallm-inference --config config.yaml
dnallm-inference --model model_name --input input_file

# Generate configuration files
dnallm-model-config-generator --output config.yaml
dnallm-model-config-generator --preview

# MCP server
dnallm-mcp-server --config config.yaml

Development Environment Usage

In the project root directory, you can use the following methods:

1. Using the Launcher Script

# Main CLI
python run_cli.py --help

# Training
python run_cli.py train --config config.yaml

# Inference
python run_cli.py inference --config config.yaml

# Generate configuration
python run_cli.py model-config-generator --output config.yaml

2. Running CLI Modules Directly

# Main CLI
python cli/cli.py --help

# Training
python cli/train.py config.yaml model_path data_path

# Inference
python cli/inference.py config.yaml model_path

# Configuration generator
python cli/model_config_generator.py --output config.yaml

3. Using Package Modules

# Package CLI
python -m dnallm.cli.cli --help

# Package training
python -m dnallm.cli.train config.yaml model_path data_path

# Package inference
python -m dnallm.cli.inference config.yaml model_path

# Package configuration generator
python -m dnallm.cli.model_config_generator --output config.yaml

Command Reference

dnallm-train

Train a DNA language model with specified configuration.

Options: - --config, -c: Path to training configuration file - --model, -m: Model name or path - --data, -d: Path to training data - --output, -o: Output directory for training results

Examples: 

# Using configuration file
dnallm-train --config finetune_config.yaml

# Using command line arguments
dnallm-train --model zhangtaolab/plant-dnagpt-BPE --data ./data --output ./outputs

dnallm-inference

Run inference with a trained DNA language model.

Options: - --config, -c: Path to inference configuration file - --model, -m: Model name or path - --input, -i: Path to input data file - --output, -o: Output file path

Examples: 

# Using configuration file
dnallm-inference --config inference_config.yaml

# Using command line arguments
dnallm-inference --model ./models/trained_model --input ./test_data.csv

dnallm-model-config-generator

Generate configuration files for DNALLM tasks.

Options: - --output, -o: Output file path for configuration - --preview: Preview configuration without saving - --template: Template type (training, inference, benchmark)

Examples: 

# Generate training configuration
dnallm-model-config-generator --output training_config.yaml

# Preview configuration
dnallm-model-config-generator --preview

dnallm-mcp-server

Start MCP (Model Context Protocol) server.

Options: - --config, -c: Path to configuration file - --port, -p: Server port (default: 8000) - --host, -h: Server host (default: localhost)

Examples: 

# Start server with configuration
dnallm-mcp-server --config mcp_config.yaml

# Start server on specific port
dnallm-mcp-server --port 9000

Configuration Examples

Training Configuration (config.yaml)

model:
  name_or_path: "zhangtaolab/plant-dnagpt-BPE"
  source: "huggingface"

task:
  task_type: "binary_classification"
  num_labels: 2
  label_names: ["negative", "positive"]

data:
  train_file: "path/to/train.csv"
  eval_file: "path/to/eval.csv"
  text_column: "sequence"
  label_column: "label"

training:
  num_train_epochs: 3
  per_device_train_batch_size: 8
  learning_rate: 5e-5
  save_steps: 1000
  eval_steps: 1000

Inference Configuration (config.yaml)

model:
  name_or_path: "path/to/trained/model"
  source: "local"

task:
  task_type: "binary_classification"
  num_labels: 2

data:
  input_file: "path/to/input/data"
  output_file: "predictions.csv"

Benchmark Configuration (config.yaml)

benchmark:
  name: "DNA Model Benchmark"
  description: "Comparing DNA language models on various tasks"

models:
  - name: "Model 1"
    source: "huggingface"
    path: "zhangtaolab/plant-dnagpt-BPE"
    task_type: "binary_classification"
  - name: "Model 2"
    source: "modelscope"
    path: "zhangtaolab/plant-dnabert"
    task_type: "binary_classification"

datasets:
  - name: "Test Dataset"
    path: "path/to/dataset.csv"
    format: "csv"
    task: "binary_classification"

evaluation:
  metrics: ["accuracy", "precision", "recall", "f1", "mcc"]
  save_predictions: true
  output_dir: "./benchmark_results"

Project Structure

DNALLM/
├── cli/                    # Root directory CLI entry points
│   ├── cli.py            # Main CLI
│   ├── train.py          # Training CLI
│   ├── inference.py      # Inference CLI
│   └── model_config_generator.py # Configuration generator
├── ui/                    # UI applications
│   ├── run_config_app.py # Configuration generator launcher
│   └── ...
├── dnallm/               # Core package
│   ├── cli/             # Package CLI modules
│   │   ├── cli.py       # Package CLI implementation
│   │   ├── train.py     # Package training module
│   │   ├── inference.py # Package inference module
│   │   └── model_config_generator.py # Package config generator
│   └── ...
├── run_cli.py           # Root directory CLI launcher
└── pyproject.toml       # Package configuration

Important Notes

  1. Development Environment: Ensure you're running commands from the project root directory
  2. Dependencies: Make sure all dependencies are properly installed
  3. Path Configuration: Use absolute paths or paths relative to the project root
  4. Python Version: Requires Python 3.10 or higher

Troubleshooting

Import Errors

  • Ensure you're running from the project root directory
  • Check Python path settings
  • Verify the package is properly installed

Configuration Errors

  • Check configuration file format
  • Verify file paths are correct
  • Ensure all required configuration parameters are present

Permission Errors

  • Check file and directory permissions
  • Ensure you have write permissions for output directories

Getting Help

# Show help for a specific command
dnallm-train --help

# Show help for configuration generator
dnallm-model-config-generator --help

# Show help for MCP server
dnallm-mcp-server --help

Next Steps