"""
Various utilities for neural networks.
"""
import math
import torch as th
import torch.nn as nn
# PyTorch 1.7 has SiLU, but we support PyTorch 1.5.
[docs]
class SiLU(nn.Module):
def forward(self, x):
return x * th.sigmoid(x)
[docs]
class GroupNorm32(nn.GroupNorm):
def forward(self, x):
return super().forward(x.float()).type(x.dtype)
[docs]
def conv_nd(dims, *args, **kwargs):
"""
Create a 1D, 2D, or 3D convolution module.
"""
if dims == 1:
return nn.Conv1d(*args, **kwargs)
elif dims == 2:
return nn.Conv2d(*args, **kwargs)
elif dims == 3:
return nn.Conv3d(*args, **kwargs)
raise ValueError(f"unsupported dimensions: {dims}")
[docs]
def linear(*args, **kwargs):
"""
Create a linear module.
"""
return nn.Linear(*args, **kwargs)
[docs]
def avg_pool_nd(dims, *args, **kwargs):
"""
Create a 1D, 2D, or 3D average pooling module.
"""
if dims == 1:
return nn.AvgPool1d(*args, **kwargs)
elif dims == 2:
return nn.AvgPool2d(*args, **kwargs)
elif dims == 3:
return nn.AvgPool3d(*args, **kwargs)
raise ValueError(f"unsupported dimensions: {dims}")
[docs]
def update_ema(target_params, source_params, rate=0.99):
"""
Update target parameters to be closer to those of source parameters using
an exponential moving average.
:param target_params: the target parameter sequence.
:param source_params: the source parameter sequence.
:param rate: the EMA rate (closer to 1 means slower).
"""
for targ, src in zip(target_params, source_params):
targ.detach().mul_(rate).add_(src, alpha=1 - rate)
[docs]
def zero_module(module):
"""
Zero out the parameters of a module and return it.
"""
for p in module.parameters():
p.detach().zero_()
return module
[docs]
def scale_module(module, scale):
"""
Scale the parameters of a module and return it.
"""
for p in module.parameters():
p.detach().mul_(scale)
return module
[docs]
def mean_flat(tensor):
"""
Take the mean over all non-batch dimensions.
"""
return tensor.mean(dim=list(range(1, len(tensor.shape))))
[docs]
def normalization(channels):
"""
Make a standard normalization layer.
:param channels: number of input channels.
:return: an nn.Module for normalization.
"""
return GroupNorm32(32, channels)
[docs]
def timestep_embedding(timesteps, dim, max_period=10000):
"""
Create sinusoidal timestep embeddings.
:param timesteps: a 1-D Tensor of N indices, one per batch element.
These may be fractional.
:param dim: the dimension of the output.
:param max_period: controls the minimum frequency of the embeddings.
:return: an [N x dim] Tensor of positional embeddings.
"""
half = dim // 2
freqs = th.exp(
-math.log(max_period) * th.arange(start=0, end=half, dtype=th.float32) / half
).to(device=timesteps.device)
args = timesteps[:, None].float() * freqs[None]
embedding = th.cat([th.cos(args), th.sin(args)], dim=-1)
if dim % 2:
embedding = th.cat([embedding, th.zeros_like(embedding[:, :1])], dim=-1)
return embedding
[docs]
def checkpoint(func, inputs, params, flag):
"""
Evaluate a function without caching intermediate activations, allowing for
reduced memory at the expense of extra compute in the backward pass.
:param func: the function to evaluate.
:param inputs: the argument sequence to pass to `func`.
:param params: a sequence of parameters `func` depends on but does not
explicitly take as arguments.
:param flag: if False, disable gradient checkpointing.
"""
if flag:
args = tuple(inputs) + tuple(params)
return CheckpointFunction.apply(func, len(inputs), *args)
else:
return func(*inputs)
[docs]
class CheckpointFunction(th.autograd.Function):
@staticmethod
def forward(ctx, run_function, length, *args):
ctx.run_function = run_function
ctx.input_tensors = list(args[:length])
ctx.input_params = list(args[length:])
with th.no_grad():
output_tensors = ctx.run_function(*ctx.input_tensors)
return output_tensors
@staticmethod
def backward(ctx, *output_grads):
ctx.input_tensors = [x.detach().requires_grad_(True) for x in ctx.input_tensors]
with th.enable_grad():
# Fixes a bug where the first op in run_function modifies the
# Tensor storage in place, which is not allowed for detach()'d
# Tensors.
shallow_copies = [x.view_as(x) for x in ctx.input_tensors]
output_tensors = ctx.run_function(*shallow_copies)
input_grads = th.autograd.grad(
output_tensors,
ctx.input_tensors + ctx.input_params,
output_grads,
allow_unused=True,
)
del ctx.input_tensors
del ctx.input_params
del output_tensors
return (None, None) + input_grads