def doStep(w):
grad = ndict.cloneZeros(v)
gw = ndict.cloneZeros(w)
for l in range(n_batch):
n_tot = x.itervalues().next().shape[1]
idx_minibatch = np.random.randint(0, n_tot, n_subbatch)
x_minibatch = {i: x[i][:, idx_minibatch] for i in x}
if convertImgs:
x_minibatch = {i: x_minibatch[i] / 256. for i in x_minibatch}
# Use z ~ q(z|x) to compute d[LB]/d[gw]
_, z, _ = model_q.gen_xz(v, x_minibatch, {}, n_subbatch)
_, logpz_q = model_q.logpxz(v, x_minibatch, z)
logpx_p, logpz_p, _gw, gz_p = model_p.dlogpxz_dwz(
w, x_minibatch, z)
for i in _gw:
gw[i] += _gw[i]
# Compute d[LB]/d[gv] where gv = v (variational params)
_, _, gv, _ = model_q.dlogpxz_dwz(v, x_minibatch, z)
weight = np.sum(logpx_p) + np.sum(logpz_p) - np.sum(logpz_q)
for i in v:
f = gv[i] * weight
h = gv[i]
cv_cov[i] = cv_cov[i] + cv_lr * (f * h - cv_cov[i])
cv_var[i] = cv_var[i] + cv_lr * (h**2 - cv_var[i])
grad[i] += f - (cv_cov[i] / (cv_var[i] + 1e-8)) * h
_, gwprior = model_p.dlogpw_dw(w)
for i in gw:
gw[i] += float(n_subbatch * n_batch) / n_tot * gwprior[i]
def optimize(_w, _gw, gw_ss, stepsize):
reg = 1e-8
for i in _gw:
gw_ss[i] += _gw[i]**2
if nsteps[0] > warmup:
_w[i] += stepsize / np.sqrt(gw_ss[i] + reg) * _gw[i]
optimize(w, gw, gw_ss, ada_stepsize)
optimize(v, grad, gv_ss, ada_stepsize)
nsteps[0] += 1
if ndict.hasNaN(grad):
raise Exception()
if ndict.hasNaN(v):
raise Exception()
return z.copy(), logpx_p + logpz_p - logpz_q
def checknan(self, v, w, gv, gw):
if ndict.hasNaN(gv) or ndict.hasNaN(gw):
raise Exception("dL_dw(): NaN found in gradients")
print('v:')
ndict.p(v)
print('w:')
ndict.p(w)
print('gv:')
ndict.p(gv)
print('gw:')
ndict.p(gw)
raise Exception("dL_dw(): NaN found in gradients")
def doStep(w):
grad = ndict.cloneZeros(v)
gw = ndict.cloneZeros(w)
for l in range(n_batch):
n_tot = x.itervalues().next().shape[1]
idx_minibatch = np.random.randint(0, n_tot, n_subbatch)
x_minibatch = {i:x[i][:,idx_minibatch] for i in x}
if convertImgs: x_minibatch = {i:x_minibatch[i]/256. for i in x_minibatch}
# Use z ~ q(z|x) to compute d[LB]/d[gw]
_, z, _ = model_q.gen_xz(v, x_minibatch, {}, n_subbatch)
_, logpz_q = model_q.logpxz(v, x_minibatch, z)
logpx_p, logpz_p, _gw, gz_p = model_p.dlogpxz_dwz(w, x_minibatch, z)
for i in _gw: gw[i] += _gw[i]
# Compute d[LB]/d[gv] where gv = v (variational params)
_, _, gv, _ = model_q.dlogpxz_dwz(v, x_minibatch, z)
weight = np.sum(logpx_p) + np.sum(logpz_p) - np.sum(logpz_q)
for i in v:
f = gv[i] * weight
h = gv[i]
cv_cov[i] = cv_cov[i] + cv_lr * (f * h - cv_cov[i])
cv_var[i] = cv_var[i] + cv_lr * (h**2 - cv_var[i])
grad[i] += f - (cv_cov[i]/(cv_var[i] + 1e-8)) * h
_, gwprior = model_p.dlogpw_dw(w)
for i in gw: gw[i] += float(n_subbatch*n_batch)/n_tot * gwprior[i]
def optimize(_w, _gw, gw_ss, stepsize):
reg=1e-8
for i in _gw:
gw_ss[i] += _gw[i]**2
if nsteps[0] > warmup:
_w[i] += stepsize / np.sqrt(gw_ss[i]+reg) * _gw[i]
optimize(w, gw, gw_ss, ada_stepsize)
optimize(v, grad, gv_ss, ada_stepsize)
nsteps[0] += 1
if ndict.hasNaN(grad):
raise Exception()
if ndict.hasNaN(v):
raise Exception()
return z.copy(), logpx_p + logpz_p - logpz_q
def dfd_dw(self, w, x, z, gz2):
x, z = self.xz_to_theano(x, z)
w, z, x, gz2 = ndict.ordereddicts((w, z, x, gz2))
A = self.get_A(x)
r = self.f_dfd_dw(*(list(w.values()) + list(x.values()) +
list(z.values()) + [A] + list(gz2.values())))
logpx, logpz, fd, gw = r[0], r[1], r[2], dict(
list(zip(list(w.keys()), r[3:3 + len(w)])))
if ndict.hasNaN(gw):
if True:
print('NaN detected in gradients')
raise Exception()
for i in gw:
gw[i][np.isnan(gw[i])] = 0
else:
print('fd: ', fd)
print('Values:')
ndict.p(w)
ndict.p(z)
print('Gradients:')
ndict.p(gw)
raise Exception("dfd_dw(): NaN found in gradients")
gw, _ = self.gwgz_to_numpy(gw, {})
return logpx, logpz, fd, gw
def gen_xz(self, w, x, z, n_batch):
A = np.ones((1, n_batch))
if not x.has_key('x'):
x['x'] = np.random.binomial(1, 0.5, size=(self.n_units[0], n_batch))
def f_sigmoid(x): return 1./(1.+np.exp(-x))
def f_softplus(x): return np.log(np.exp(x) + 1)# - np.log(2)
nonlinear = {'tanh': np.tanh, 'sigmoid': f_sigmoid,'softplus': f_softplus}[self.nonlinear]
hidden = []
hidden.append(x['x'])
for i in range(1, len(self.n_units)-1):
hidden.append(nonlinear(np.dot(w['w_%i'%i], hidden[i-1]) + np.dot(w['b_%i'%i], A)))
mean = np.dot(w['w_mean'], hidden[-1]) + self.logmeanb_factor * np.dot(w['b_mean'], A)
logvar = self.logvar_const + self.logvar_factor * np.dot(w['w_logvar'], hidden[-1]) + np.dot(w['b_logvar'], A)
if not z.has_key('eps0'):
z['eps0'] = np.random.normal(mean, np.exp(logvar/2))
if ndict.hasNaN(z):
raise Exception("NaN detected")
_z = {'mean':mean, 'logvar':logvar}
return x, z, _z
def checknan(self, v, w, gv, gw):
if ndict.hasNaN(gv) or ndict.hasNaN(gw):
raise Exception("dL_dw(): NaN found in gradients")
#print 'logpx: ', logpx
#print 'logpz: ', logpz
#print 'logqz: ', logqz
print 'v:'
ndict.p(v)
print 'w:'
ndict.p(w)
print 'gv:'
ndict.p(gv)
print 'gw:'
ndict.p(gw)
raise Exception("dL_dw(): NaN found in gradients")
def checknan(self, v, w, gv, gw):
if ndict.hasNaN(gv) or ndict.hasNaN(gw):
raise Exception("dL_dw(): NaN found in gradients")
#print 'logpx: ', logpx
#print 'logpz: ', logpz
#print 'logqz: ', logqz
print 'v:'
ndict.p(v)
print 'w:'
ndict.p(w)
print 'gv:'
ndict.p(gv)
print 'gw:'
ndict.p(gw)
raise Exception("dL_dw(): NaN found in gradients")
def dlogpxz_dwz(self, w, x, z):
x, z = self.xz_to_theano(x, z)
w, z, x = ndict.ordereddicts((w, z, x))
A = self.get_A(x)
allvars = list(w.values()) + list(x.values()) + list(z.values()) + [A]
# Check if keys are correct
keys = list(w.keys()) + list(x.keys()) + list(z.keys()) + ['A']
for i in range(len(keys)):
if keys[i] != self.allvars_keys[i]:
"Input values are incorrect!"
print('Input:', keys)
print('Should be:', self.allvars_keys)
raise Exception()
r = self.f_dlogpxz_dwz(*allvars)
logpx, logpz, gw, gz = r[0], r[1], dict(
list(zip(list(w.keys()), r[2:2 + len(w)]))), dict(
list(zip(list(z.keys()), r[2 + len(w):])))
if ndict.hasNaN(gw) or ndict.hasNaN(gz):
if True:
print('NaN detected in gradients')
raise Exception()
for i in gw:
gw[i][np.isnan(gw[i])] = 0
for i in gz:
gz[i][np.isnan(gz[i])] = 0
else:
print('logpx: ', logpx)
print('logpz: ', logpz)
print('Values:')
ndict.p(w)
ndict.p(z)
print('Gradients:')
ndict.p(gw)
ndict.p(gz)
raise Exception("dlogpxz_dwz(): NaN found in gradients")
gw, gz = self.gwgz_to_numpy(gw, gz)
return logpx, logpz, gw, gz
def dlogpxz_dwz(self, w, x, z):
x, z = self.xz_to_theano(x, z)
w, z, x = ndict.ordereddicts((w, z, x))
A = self.get_A(x)
allvars = w.values() + x.values() + z.values() + [A]
# Check if keys are correct
keys = w.keys() + x.keys() + z.keys() + ['A']
for i in range(len(keys)):
if keys[i] != self.allvars_keys[i]:
"Input values are incorrect!"
print 'Input:', keys
print 'Should be:', self.allvars_keys
raise Exception()
r = self.f_dlogpxz_dwz(*allvars)
logpx, logpz, gw, gz = r[0], r[1], dict(zip(w.keys(), r[2:2+len(w)])), dict(zip(z.keys(), r[2+len(w):]))
if ndict.hasNaN(gw) or ndict.hasNaN(gz):
if True:
print 'NaN detected in gradients'
raise Exception()
for i in gw: gw[i][np.isnan(gw[i])] = 0
for i in gz: gz[i][np.isnan(gz[i])] = 0
else:
print 'logpx: ', logpx
print 'logpz: ', logpz
print 'Values:'
ndict.p(w)
ndict.p(z)
print 'Gradients:'
ndict.p(gw)
ndict.p(gz)
raise Exception("dlogpxz_dwz(): NaN found in gradients")
gw, gz = self.gwgz_to_numpy(gw, gz)
return logpx, logpz, gw, gz
def dfd_dw(self, w, x, z, gz2):
x, z = self.xz_to_theano(x, z)
w, z, x, gz2 = ndict.ordereddicts((w, z, x, gz2))
A = self.get_A(x)
r = self.f_dfd_dw(*(w.values() + x.values() + z.values() + [A] + gz2.values()))
logpx, logpz, fd, gw = r[0], r[1], r[2], dict(zip(w.keys(), r[3:3+len(w)]))
if ndict.hasNaN(gw):
if True:
print 'NaN detected in gradients'
raise Exception()
for i in gw: gw[i][np.isnan(gw[i])] = 0
else:
print 'fd: ', fd
print 'Values:'
ndict.p(w)
ndict.p(z)
print 'Gradients:'
ndict.p(gw)
raise Exception("dfd_dw(): NaN found in gradients")
gw, _ = self.gwgz_to_numpy(gw, {})
return logpx, logpz, fd, gw
