def f(self, t, y_aug):
sde, params, n_tensors = self._base_sde, self.params, len(y_aug) // 3
y, adj_y, adj_l = y_aug[:n_tensors], y_aug[
n_tensors:2 * n_tensors], y_aug[2 * n_tensors:3 * n_tensors]
vjp_l = tuple(torch.zeros_like(adj_l_) for adj_l_ in adj_l)
with torch.enable_grad():
y = tuple(y_.detach().requires_grad_(True) for y_ in y)
adj_y = tuple(adj_y_.detach() for adj_y_ in adj_y)
f_eval = sde.f(-t, y)
f_eval = tuple(-f_eval_ for f_eval_ in f_eval)
f_eval = misc.make_seq_requires_grad_y(f_eval, y)
vjp_y_and_params = torch.autograd.grad(
outputs=f_eval,
inputs=y + params,
grad_outputs=tuple(-adj_y_ for adj_y_ in adj_y),
allow_unused=True,
create_graph=True)
vjp_y = vjp_y_and_params[:n_tensors]
vjp_y = misc.convert_none_to_zeros(vjp_y, y)
vjp_params = vjp_y_and_params[n_tensors:]
vjp_params = misc.flatten_convert_none_to_zeros(vjp_params, params)
# Vector field change due to log-ratio term, i.e. ||u||^2 / 2.
g_eval = sde.g(-t, y)
h_eval = sde.h(-t, y)
ginv_eval = tuple(torch.pinverse(g_eval_) for g_eval_ in g_eval)
u_eval = misc.seq_sub(f_eval, h_eval)
u_eval = tuple(
torch.bmm(ginv_eval_, u_eval_)
for ginv_eval_, u_eval_ in zip(ginv_eval, u_eval))
log_ratio_correction = tuple(.5 * torch.sum(u_eval_**2., dim=1)
for u_eval_ in u_eval)
log_ratio_correction = misc.make_seq_requires_grad_y(
log_ratio_correction, y)
corr_vjp_y_and_params = torch.autograd.grad(
outputs=log_ratio_correction,
inputs=y + params,
grad_outputs=adj_l,
allow_unused=True,
)
corr_vjp_y = corr_vjp_y_and_params[:n_tensors]
corr_vjp_y = misc.convert_none_to_zeros(corr_vjp_y, y)
corr_vjp_params = corr_vjp_y_and_params[n_tensors:]
corr_vjp_params = misc.flatten_convert_none_to_zeros(
corr_vjp_params, params)
vjp_y = misc.seq_add(vjp_y, corr_vjp_y)
vjp_params = vjp_params + corr_vjp_params
return (*f_eval, *vjp_y, *vjp_l, vjp_params)
def step_logqp(self, t, y, dt, logqp0):
t1, y1 = self.step(t, y, dt)
if self.sde.noise_type in ("diagonal", "scalar"):
f_eval = self.sde.f(t, y)
g_eval = self.sde.g(t, y)
h_eval = self.sde.h(t, y)
u_eval = misc.seq_sub_div(f_eval, h_eval, g_eval)
logqp1 = tuple(logqp0_i + .5 * torch.sum(u_eval_i**2., dim=1) * dt
for logqp0_i, u_eval_i in zip(logqp0, u_eval))
else:
f_eval = self.sde.f(t, y)
g_eval = self.sde.g(t, y)
h_eval = self.sde.h(t, y)
ginv_eval = tuple(torch.pinverse(g_eval_) for g_eval_ in g_eval)
u_eval = misc.seq_sub(f_eval, h_eval)
u_eval = misc.seq_batch_mvp(ms=ginv_eval, vs=u_eval)
logqp1 = tuple(logqp0_i + .5 * torch.sum(u_eval_i**2., dim=1) * dt
for logqp0_i, u_eval_i in zip(logqp0, u_eval))
return t1, y1, logqp1
def gdg_prod(self, t, y_aug, noise):
sde, params, n_tensors = self._base_sde, self.params, len(y_aug) // 3
y, adj_y, adj_l = y_aug[:n_tensors], y_aug[
n_tensors:2 * n_tensors], y_aug[2 * n_tensors:3 * n_tensors]
vjp_l = tuple(torch.zeros_like(adj_l_) for adj_l_ in adj_l)
with torch.enable_grad():
y = tuple(y_.detach().requires_grad_(True) for y_ in y)
adj_y = tuple(adj_y_.detach().requires_grad_(True)
for adj_y_ in adj_y)
g_eval = sde.g(-t, y)
g_eval = misc.make_seq_requires_grad_y(g_eval, y)
gdg_times_v = torch.autograd.grad(
outputs=g_eval,
inputs=y,
grad_outputs=misc.seq_mul(g_eval, noise),
allow_unused=True,
create_graph=True,
)
gdg_times_v = misc.convert_none_to_zeros(gdg_times_v, y)
dgdy = torch.autograd.grad(
outputs=g_eval,
inputs=y,
grad_outputs=tuple(torch.ones_like(y_) for y_ in y),
allow_unused=True,
create_graph=True,
)
dgdy = misc.convert_none_to_zeros(dgdy, y)
prod_partials_adj_y_and_params = torch.autograd.grad(
outputs=g_eval,
inputs=y + params,
grad_outputs=misc.seq_mul(adj_y, noise, dgdy),
allow_unused=True,
create_graph=True,
)
prod_partials_adj_y = prod_partials_adj_y_and_params[:n_tensors]
prod_partials_adj_y = misc.convert_none_to_zeros(
prod_partials_adj_y, y)
prod_partials_params = prod_partials_adj_y_and_params[n_tensors:]
prod_partials_params = misc.flatten_convert_none_to_zeros(
prod_partials_params, params)
gdg_v = torch.autograd.grad(
outputs=g_eval,
inputs=y,
grad_outputs=tuple(
p.detach() for p in misc.seq_mul(adj_y, noise, g_eval)),
allow_unused=True,
create_graph=True,
)
gdg_v = misc.convert_none_to_zeros(gdg_v, y)
gdg_v = misc.make_seq_requires_grad_y(gdg_v, y)
gdg_v = tuple(gdg_v_.sum() for gdg_v_ in gdg_v)
mixed_partials_adj_y_and_params = torch.autograd.grad(
outputs=gdg_v,
inputs=y + params,
allow_unused=True,
)
mixed_partials_adj_y = mixed_partials_adj_y_and_params[:n_tensors]
mixed_partials_adj_y = misc.convert_none_to_zeros(
mixed_partials_adj_y, y)
mixed_partials_params = mixed_partials_adj_y_and_params[n_tensors:]
mixed_partials_params = misc.flatten_convert_none_to_zeros(
mixed_partials_params, params)
return (*gdg_times_v,
*misc.seq_sub(prod_partials_adj_y, mixed_partials_adj_y),
*vjp_l, prod_partials_params - mixed_partials_params)
def f(self, t, y_aug):
sde, params, n_tensors = self._base_sde, self.params, len(y_aug) // 2
y, adj_y = y_aug[:n_tensors], y_aug[n_tensors:2 * n_tensors]
with torch.enable_grad():
y = tuple(y_.detach().requires_grad_(True) for y_ in y)
adj_y = tuple(adj_y_.detach() for adj_y_ in adj_y)
g_eval = sde.g(-t, y)
g_eval = misc.make_seq_requires_grad_y(g_eval, y)
gdg = torch.autograd.grad(outputs=g_eval,
inputs=y,
grad_outputs=g_eval,
allow_unused=True,
create_graph=True)
gdg = misc.convert_none_to_zeros(gdg, y)
f_eval = sde.f(-t, y)
f_eval_corrected = misc.seq_sub(
gdg, f_eval) # Stratonovich correction for reverse-time.
f_eval_corrected = misc.make_seq_requires_grad_y(
f_eval_corrected, y)
vjp_y_and_params = torch.autograd.grad(
outputs=f_eval_corrected,
inputs=y + params,
grad_outputs=tuple(-adj_y_ for adj_y_ in adj_y),
allow_unused=True,
create_graph=True)
vjp_y = vjp_y_and_params[:n_tensors]
vjp_y = misc.convert_none_to_zeros(vjp_y, y)
vjp_params = vjp_y_and_params[n_tensors:]
vjp_params = misc.flatten_convert_none_to_zeros(vjp_params, params)
adj_times_dgdx = torch.autograd.grad(outputs=g_eval,
inputs=y,
grad_outputs=adj_y,
allow_unused=True,
create_graph=True)
adj_times_dgdx = misc.convert_none_to_zeros(adj_times_dgdx, y)
# This extra term is due to converting the *adjoint* Stratonovich backward SDE to Itô.
extra_vjp_y_and_params = torch.autograd.grad(
outputs=g_eval,
inputs=y + params,
grad_outputs=adj_times_dgdx,
allow_unused=True,
)
extra_vjp_y = extra_vjp_y_and_params[:n_tensors]
extra_vjp_y = misc.convert_none_to_zeros(extra_vjp_y, y)
extra_vjp_params = extra_vjp_y_and_params[n_tensors:]
extra_vjp_params = misc.flatten_convert_none_to_zeros(
extra_vjp_params, params)
vjp_y = misc.seq_add(vjp_y, extra_vjp_y)
vjp_params = vjp_params + extra_vjp_params
return (*f_eval_corrected, *vjp_y, vjp_params)
