Advanced Usage

Visualizing the workflow

You can generate a HTML visualization of the Taiji pipeline using the following command:

taiji view Taiji.html

Open the “Taiji.html” using your web browser to see what the pipeline can do.

Cache manipulation

The workflow cache (default to sciflow.db) stores data that are neccessary for resuming the program at any checkpoints. Taiji provides several commands that allow you to inspect and clear the cache.

• Inspection: taiji show STEP_NAME.
• Deletion: taiji delete STEP_NAME or taiji delete STEP_NAME --delete-all.

Parallelism and Distributed computing

Use -n num_jobs to tell Taiji how many concurrent jobs you want to run.

If you run the program without distributed computing, you need to use +RTS -N10 to specify the number of cores (here we use 10 cores).

Taiji currently supports distributed computing systems like slurm or PBS. To use this feature, first modify your configuration file (see Distributed computing). Then this feature can be turned on by taiji run --config config.yml --cloud.