workflow

Workflow

We wrote the workflow using CWL and executed using StreamFlow. As said in the principal README, the workflow is based on the jupyter notebook available to the following link.

Applications

Following the notebook cells, we created self-contained Python applications. As the notebook, we used the Devito code library to improve the portability and performance of the applications. The applications are in the cwl/scrips directory.

Execution environment

In our experiment we used Python v3.10.12

    python -m venv venv
    pip install -r requirements.txt
    source venv/bin/activate
    streamflow run streamflow.yml

Before to reproduce the workflow execution, the user can custumize the config file. In particular:

• nshots: Number of shots to used to generate the gradient
• nreceivers: Number of receiver locations per shot
• fwi_iterations: Number of outer FWI iterations In the notebook these parameters are described in detail.

Other important hyperparameters are:

• num_threads: how many threads the reduce step can use.
• nshards: how many instance of the compute residual must be created.