Make Ising example configurable #89
Conversation
| @@ -100,7 +105,9 @@ def create_dataset(L, histogram_cutoff, dir): | |||
| for config in multiset_permutations(primal_configuration): | |||
| config = np.array(config) | |||
| config = np.reshape(config, (L, L, L)) | |||
| total_energy, atomic_features = E_dimensionless(config, L) | |||
| total_energy, atomic_features = E_dimensionless( | |||
| config, L, spin_function | |||
There was a problem hiding this comment.
| config, L, spin_function | |
| config, L, spin_function, scale_spin |
There was a problem hiding this comment.
Did we have a unit test for Ising model? Not sure why the checks did not catch this
There was a problem hiding this comment.
there are currently no unit tests that use this model.
| spin = np.zeros_like(config) | ||
|
|
||
| if scale_spin: | ||
| random_scaling = np.random.random((L ** 3,)) |
There was a problem hiding this comment.
| random_scaling = np.random.random((L ** 3,)) | |
| random_scaling = np.random.random((L, L, L)) |
There was a problem hiding this comment.
@pzhanggit
if you propose this change of dimensionality, just make sure that this change of dimensionality does not break the np.multiply in the following line that perform element-wise multiplication between lumpy arrays
There was a problem hiding this comment.
@allaffa That's exactly why I proposed the change. The original code failed with dimension inconsistency error. I changed it to fix the bugs. Same as line 109. I can unfix them if you want
| for z in range(L): | ||
| for y in range(L): | ||
| for x in range(L): | ||
| spin[x, y, z] = math.sin(math.pi * config[x, y, z] / 2) | ||
| spin[x, y, z] = spin_function(config[x, y, z]) |
Make Ising example configurable
Makes the spin value randomization and non-linear spin function optional. With this update running the function with defaults results in the original Ising model