apunkt/GPy
1
0
Fork 0
mirror of https://github.com/SheffieldML/GPy.git synced 2026-05-09 12:02:38 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions

new slicing done and first attempts at copy and pickling full models

Browse source
This commit is contained in:
Max Zwiessele 2014-03-27 09:28:44 +00:00
parent c65a1e3544
commit e26e737014
2 changed files with 103 additions and 86 deletions
Download patch file Download diff file Expand all files Collapse all files

184
GPy/kern/_src/kernel_slice_operations.py
View file

@ -5,118 +5,134 @@ Created on 11 Mar 2014
''' '''
from ...core.parameterization.parameterized import ParametersChangedMeta from ...core.parameterization.parameterized import ParametersChangedMeta
import numpy as np import numpy as np
from functools import wraps
def put_clean(dct, name, *args, **kw): def put_clean(dct, name, func):
if name in dct: if name in dct:
dct['_clean_{}'.format(name)] = dct[name] dct['_clean_{}'.format(name)] = dct[name]
dct[name] = _slice_wrapper(None, dct[name], *args, **kw) dct[name] = func(dct[name])
class KernCallsViaSlicerMeta(ParametersChangedMeta): class KernCallsViaSlicerMeta(ParametersChangedMeta):
def __new__(cls, name, bases, dct): def __new__(cls, name, bases, dct):
put_clean(dct, 'K') put_clean(dct, 'K', _slice_K)
put_clean(dct, 'Kdiag', diag=True) put_clean(dct, 'Kdiag', _slice_Kdiag)
put_clean(dct, 'update_gradients_full', diag=False, derivative=True) put_clean(dct, 'update_gradients_full', _slice_update_gradients_full)
put_clean(dct, 'gradients_X', diag=False, derivative=True, ret_X=True) put_clean(dct, 'update_gradients_diag', _slice_update_gradients_diag)
put_clean(dct, 'gradients_X_diag', diag=True, derivative=True, ret_X=True) put_clean(dct, 'gradients_X', _slice_gradients_X)
put_clean(dct, 'psi0', diag=False, derivative=False) put_clean(dct, 'gradients_X_diag', _slice_gradients_X_diag)
put_clean(dct, 'psi1', diag=False, derivative=False)
put_clean(dct, 'psi2', diag=False, derivative=False) put_clean(dct, 'psi0', _slice_psi)
put_clean(dct, 'update_gradients_expectations', derivative=True, psi_stat=True) put_clean(dct, 'psi1', _slice_psi)
put_clean(dct, 'gradients_Z_expectations', derivative=True, psi_stat_Z=True, ret_X=True) put_clean(dct, 'psi2', _slice_psi)
put_clean(dct, 'gradients_qX_expectations', derivative=True, psi_stat=True, ret_X=True) put_clean(dct, 'update_gradients_expectations', _slice_update_gradients_expectations)
put_clean(dct, 'gradients_Z_expectations', _slice_gradients_Z_expectations)
put_clean(dct, 'gradients_qX_expectations', _slice_gradients_qX_expectations)
return super(KernCallsViaSlicerMeta, cls).__new__(cls, name, bases, dct) return super(KernCallsViaSlicerMeta, cls).__new__(cls, name, bases, dct)
class _Slice_wrap(object): class _Slice_wrap(object):
def __init__(self, instance, f): def __init__(self, k, X, X2=None):
self.k = instance self.k = k
self.f = f self.shape = X.shape
def copy_to(self, new_instance): if self.k._sliced_X == 0:
return self.__class__(new_instance, self.f) self.X = self.k._slice_X(X)
def _slice_X(self, X): self.X2 = self.k._slice_X(X2) if X2 is not None else None
return self.k._slice_X(X) if not self.k._sliced_X else X self.ret = True
def _slice_X_X2(self, X, X2): else:
return self.k._slice_X(X) if not self.k._sliced_X else X, self.k._slice_X(X2) if X2 is not None and not self.k._sliced_X else X2 self.X = X
self.X2 = X2
self.ret = False
def __enter__(self): def __enter__(self):
self.k._sliced_X += 1 self.k._sliced_X += 1
return self return self
def __exit__(self, *a): def __exit__(self, *a):
self.k._sliced_X -= 1 self.k._sliced_X -= 1
def handle_return_array(self, return_val):
class _Slice_wrapper(_Slice_wrap): if self.ret:
def __call__(self, X, X2 = None, *a, **kw): ret = np.zeros(self.shape)
X, X2 = self._slice_X_X2(X, X2) ret[:, self.k.active_dims] = return_val
with self:
ret = self.f(X, X2, *a, **kw)
return ret return ret
return return_val
class _Slice_wrapper_diag(_Slice_wrap): def _slice_K(f):
def __call__(self, X, *a, **kw): @wraps(f)
X = self._slice_X(X) def wrap(self, X, X2 = None, *a, **kw):
with self: with _Slice_wrap(self, X, X2) as s:
ret = self.f(X, *a, **kw) ret = f(self, s.X, s.X2, *a, **kw)
return ret return ret
return wrap
class _Slice_wrapper_derivative(_Slice_wrap): def _slice_Kdiag(f):
def __call__(self, dL_dK, X, X2=None): @wraps(f)
self._slice_X(X) def wrap(self, X, *a, **kw):
with self: with _Slice_wrap(self, X, None) as s:
ret = self.f(dL_dK, X, X2) ret = f(self, s.X, *a, **kw)
return ret return ret
return wrap
class _Slice_wrapper_diag_derivative(_Slice_wrap): def _slice_update_gradients_full(f):
def __call__(self, dL_dKdiag, X): @wraps(f)
X = self._slice_X(X) def wrap(self, dL_dK, X, X2=None):
with self: with _Slice_wrap(self, X, X2) as s:
ret = self.f(dL_dKdiag, X) ret = f(self, dL_dK, s.X, s.X2)
return ret return ret
return wrap
class _Slice_wrapper_grad_X(_Slice_wrap): def _slice_update_gradients_diag(f):
def __call__(self, dL_dK, X, X2=None): @wraps(f)
ret = np.zeros(X.shape) def wrap(self, dL_dKdiag, X):
X, X2 = self._slice_X_X2(X, X2) with _Slice_wrap(self, X, None) as s:
with self: ret = f(self, dL_dKdiag, s.X)
ret[:, self.k.active_dims] = self.f(dL_dK, X, X2)
return ret return ret
return wrap
class _Slice_wrapper_grad_X_diag(_Slice_wrap): def _slice_gradients_X(f):
def __call__(self, dL_dKdiag, X): @wraps(f)
ret = np.zeros(X.shape) def wrap(self, dL_dK, X, X2=None):
X = self._slice_X(X) with _Slice_wrap(self, X, X2) as s:
with self: ret = s.handle_return_array(f(self, dL_dK, s.X, s.X2))
ret[:, self.k.active_dims] = self.f(dL_dKdiag, X)
return ret return ret
return wrap
class _Slice_wrapper_psi_stat_derivative_no_ret(_Slice_wrap): def _slice_gradients_X_diag(f):
def __call__(self, dL_dpsi0, dL_dpsi1, dL_dpsi2, Z, variational_posterior): @wraps(f)
Z, variational_posterior = self._slice_X_X2(Z, variational_posterior) def wrap(self, dL_dKdiag, X):
with self: with _Slice_wrap(self, X, None) as s:
ret = self.f(dL_dpsi0, dL_dpsi1, dL_dpsi2, Z, variational_posterior) ret = s.handle_return_array(f(self, dL_dKdiag, s.X))
return ret return ret
return wrap
class _Slice_wrapper_psi_stat_derivative(_Slice_wrap): def _slice_psi(f):
def __call__(self, dL_dpsi0, dL_dpsi1, dL_dpsi2, Z, variational_posterior): @wraps(f)
ret1, ret2 = np.zeros(variational_posterior.shape), np.zeros(variational_posterior.shape) def wrap(self, Z, variational_posterior):
Z, variational_posterior = self._slice_X_X2(Z, variational_posterior) with _Slice_wrap(self, Z, variational_posterior) as s:
with self: ret = f(self, s.X, s.X2)
ret = list(self.f(dL_dpsi0, dL_dpsi1, dL_dpsi2, Z, variational_posterior)) return ret
return wrap
def _slice_update_gradients_expectations(f):
@wraps(f)
def wrap(self, dL_dpsi0, dL_dpsi1, dL_dpsi2, Z, variational_posterior):
with _Slice_wrap(self, Z, variational_posterior) as s:
ret = f(self, dL_dpsi0, dL_dpsi1, dL_dpsi2, s.X, s.X2)
return ret
return wrap
def _slice_gradients_Z_expectations(f):
@wraps(f)
def wrap(self, dL_dpsi1, dL_dpsi2, Z, variational_posterior):
with _Slice_wrap(self, Z, variational_posterior) as s:
ret = s.handle_return_array(f(self, dL_dpsi1, dL_dpsi2, s.X, s.X2))
return ret
return wrap
def _slice_gradients_qX_expectations(f):
@wraps(f)
def wrap(self, dL_dpsi0, dL_dpsi1, dL_dpsi2, Z, variational_posterior):
with _Slice_wrap(self, variational_posterior, Z) as s:
ret = list(f(self, dL_dpsi0, dL_dpsi1, dL_dpsi2, s.X2, s.X))
r2 = ret[:2] r2 = ret[:2]
ret[0] = ret1 ret[0] = s.handle_return_array(r2[0])
ret[1] = ret2 ret[1] = s.handle_return_array(r2[1])
ret[0][:, self.k.active_dims] = r2[0]
ret[1][:, self.k.active_dims] = r2[1]
del r2
return ret
class _Slice_wrapper_psi_stat_derivative_Z(_Slice_wrap):
def __call__(self, dL_dpsi0, dL_dpsi1, dL_dpsi2, Z, variational_posterior):
ret1, ret2 = np.zeros(variational_posterior.shape), np.zeros(variational_posterior.shape)
Z, variational_posterior = self._slice_X_X2(Z, variational_posterior)
with self:
ret = list(self.f(dL_dpsi0, dL_dpsi1, dL_dpsi2, Z, variational_posterior))
r2 = ret[:2]
ret[0] = ret1
ret[1] = ret2
ret[0][:, self.k.active_dims] = r2[0]
ret[1][:, self.k.active_dims] = r2[1]
del r2 del r2
return ret return ret
return wrap

3
GPy/util/caching.py
View file

@ -101,7 +101,7 @@ class Cacher(object):
def __name__(self): def __name__(self):
return self.operation.__name__ return self.operation.__name__
from functools import partial from functools import partial, update_wrapper
class Cacher_wrap(object): class Cacher_wrap(object):
def __init__(self, f, limit, ignore_args, force_kwargs): def __init__(self, f, limit, ignore_args, force_kwargs):
@ -109,6 +109,7 @@ class Cacher_wrap(object):
self.ignore_args = ignore_args self.ignore_args = ignore_args
self.force_kwargs = force_kwargs self.force_kwargs = force_kwargs
self.f = f self.f = f
update_wrapper(self, self.f)
def __get__(self, obj, objtype=None): def __get__(self, obj, objtype=None):
return partial(self, obj) return partial(self, obj)
def __call__(self, *args, **kwargs): def __call__(self, *args, **kwargs):