coffea.ml_tools

Tools to interface with various ML inference services

Providing the interfaces to the run ML inference such that user can simply handle data mangling in awkward/numpy formats. Specifics of passing numpy arrays conversion and the handling of dask are mostly abstract away.

Classes

numpy_call_wrapper()	Wrapper for awkward.to_numpy evaluations for dask_awkward array inputs.
torch_wrapper(torch_jit[, expected_output_shape])	Wrapper for running pytorch with awkward/dask-awkward inputs.
triton_wrapper(model_url[, client_args, ...])	Wrapper for running triton inference.
xgboost_wrapper(fname)	Very simple wrapper for xgbooster inference.
tf_wrapper(tf_model[, skip_length_zero])	Wrapper for running tensorflow inference with awkward/dask-awkward inputs.

Class Inheritance Diagram