qcd_ml.compat

Compatibility layers for third party code.

qcd_ml.compat.gpt

Compatibility to lehner/gpt.

Using the two functions lattice2ndarray and ndarray2lattice it is possible to use gpt operators in torch or numpy:

w_gpt = g.qcd.fermion.wilson_clover(U_gpt, {"mass": -0.58,
    "csw_r": 1.0,
    "csw_t": 1.0,
    "xi_0": 1.0,
    "nu": 1.0,
    "isAnisotropic": False,
    "boundary_phases": [1,1,1,1]})

w = lambda x: torch.tensor(lattice2ndarray(w_gpt(ndarray2lattice(x.numpy(), U_gpt[0].grid, g.vspincolor))))

qcd_ml.compat.gpt.lattice2ndarray(lattice)

Converts a lehner/gpt lattice to a numpy ndarray keeping the ordering of axes as one would expect. Example:

q_top = g.qcd.gauge.topological_charge_5LI(U_smeared, field=True)
plot_scalar_field(lattice2ndarray(q_top))

qcd_ml.compat.gpt.ndarray2lattice(ndarray, grid, lat_constructor)

Converts an ndarray to a gpt lattice, it is the inverse of lattice2ndarray. Example:

lat = ndarray2lattice(arr, g.grid([4,4,4,8], g.double), g.vspincolor)