sweepx

The sweep toolkit: seismic I/O, FWI losses, neural reparameterizations, the YAML task runner and a natural-language agent, on top of the sweep wave-equation engine.

pip install sweepx

That installs the toolkit and the engine it runs on, sweep-solver (import sweep), which ships prebuilt CUDA cores. Optional extras: sweepx[segy], [hdf5], [animate], [agent-ui], [mcp], [vllm], [all]. Installation guide.

What is in it

Module	What it does	Docs
sweepx.io	Seismic I/O — SEG-Y, acquisition geometry, data and model plans.	IO
sweepx.loss	Misfit functions for FWI — two dozen, from L2 to optimal transport.	Loss
sweepx.nn	Neural reparameterizations and priors — INR / hash / SIREN encoders, TV and diffusion priors.	NN
sweepx.opt	FWI optimizers — L-BFGS, truncated Newton / Gauss-Newton, Adam, Shampoo / SOAP, preconditioners and bound parameterizations.	Opt
sweepx.tasks	Production FWI / LSRTM / RTM runner — typed YAML task specs, a CLI, multi-GPU runs, field-data pipelines, preprocessing and plotting.	Tasks
sweepx.agent	Natural-language control layer — chat + files → validated runs, through a local LLM (Ollama / vLLM).	Agent

The modules are layered: agent → tasks → io / loss / nn / opt. A module imports only from the layers below it, and the four libraries neither import each other nor the engine (checked in CI with import-linter).

Quick start

sweep-tasks run examples/tasks/synthetic/09_forward_constant_box.yaml   # or: sweepx tasks run …
sweep-tasks init fwi -o my_fwi.yaml    # an annotated FWI template (any task type) ...
sweep-tasks run my_fwi.yaml            # ... that runs as written: Marmousi, nothing to download

import sweep                         # the engine: equations, propagators
from sweepx.loss import EnvelopeLoss # the toolkit, one module at a time
from sweepx.tasks import load_task, TaskRunner

…or in plain language, via the agent (needs a local LLM):

sweep-agent chat                   # or: sweepx agent chat
>>> load the Marmousi benchmark model and run a forward — show the shot gather

Development

git clone https://github.com/DeepWave-KAUST/sweepx.git && cd sweepx
pip install -e ".[dev]"
pytest tests/loss                  # tests/<module>/ for each module
lint-imports                       # the layering contract

License

MIT — see LICENSE.