commit my own version of vq, with a fix for cosine similarity and support for masking

This commit is contained in:
James Betker 2022-07-18 10:12:17 -06:00
parent 4ce5c31705
commit 7a10c3fed8

View File

@ -0,0 +1,444 @@
import functools
import torch
from torch import nn, einsum
import torch.nn.functional as F
import torch.distributed as distributed
from torch.cuda.amp import autocast
from einops import rearrange, repeat
from contextlib import contextmanager
def par(t, nm):
print(f'grad report {nm}: {t}')
return t
def reg(t, nm):
l = torch.tensor([0], requires_grad=True, device=t.device, dtype=torch.float)
l.register_hook(functools.partial(par, nm=nm))
t = t + l
return t
def exists(val):
return val is not None
def default(val, d):
return val if exists(val) else d
def noop(*args, **kwargs):
pass
def l2norm(t):
return F.normalize(t, p = 2, dim = -1)
def log(t, eps = 1e-20):
return torch.log(t.clamp(min = eps))
def gumbel_noise(t):
noise = torch.zeros_like(t).uniform_(0, 1)
return -log(-log(noise))
def gumbel_sample(t, temperature = 1., dim = -1):
if temperature == 0:
return t.argmax(dim = dim)
return ((t / temperature) + gumbel_noise(t)).argmax(dim = dim)
def ema_inplace(moving_avg, new, decay):
moving_avg.data.mul_(decay).add_(new, alpha = (1 - decay))
def laplace_smoothing(x, n_categories, eps = 1e-5):
return (x + eps) / (x.sum() + n_categories * eps)
def sample_vectors(samples, num):
num_samples, device = samples.shape[0], samples.device
if num_samples >= num:
indices = torch.randperm(num_samples, device = device)[:num]
else:
indices = torch.randint(0, num_samples, (num,), device = device)
return samples[indices]
def kmeans(samples, num_clusters, num_iters = 10, use_cosine_sim = False):
dim, dtype, device = samples.shape[-1], samples.dtype, samples.device
means = sample_vectors(samples, num_clusters)
for _ in range(num_iters):
if use_cosine_sim:
dists = samples @ means.t()
else:
diffs = rearrange(samples, 'n d -> n () d') \
- rearrange(means, 'c d -> () c d')
dists = -(diffs ** 2).sum(dim = -1)
buckets = dists.max(dim = -1).indices
bins = torch.bincount(buckets, minlength = num_clusters)
zero_mask = bins == 0
bins_min_clamped = bins.masked_fill(zero_mask, 1)
new_means = buckets.new_zeros(num_clusters, dim, dtype = dtype)
new_means.scatter_add_(0, repeat(buckets, 'n -> n d', d = dim), samples)
new_means = new_means / bins_min_clamped[..., None]
if use_cosine_sim:
new_means = l2norm(new_means)
means = torch.where(zero_mask[..., None], means, new_means)
return means, bins
# regularization losses
def orthgonal_loss_fn(t):
# eq (2) from https://arxiv.org/abs/2112.00384
n = t.shape[0]
normed_codes = l2norm(t)
identity = torch.eye(n, device = t.device)
cosine_sim = einsum('i d, j d -> i j', normed_codes, normed_codes)
return ((cosine_sim - identity) ** 2).sum() / (n ** 2)
# distance types
class EuclideanCodebook(nn.Module):
def __init__(
self,
dim,
codebook_size,
kmeans_init = False,
kmeans_iters = 10,
decay = 0.8,
eps = 1e-5,
threshold_ema_dead_code = 2,
use_ddp = False,
learnable_codebook = False,
sample_codebook_temp = 0
):
super().__init__()
self.decay = decay
init_fn = torch.randn if not kmeans_init else torch.zeros
embed = init_fn(codebook_size, dim)
self.codebook_size = codebook_size
self.kmeans_iters = kmeans_iters
self.eps = eps
self.threshold_ema_dead_code = threshold_ema_dead_code
self.sample_codebook_temp = sample_codebook_temp
self.all_reduce_fn = distributed.all_reduce if use_ddp else noop
self.register_buffer('initted', torch.Tensor([not kmeans_init]))
self.register_buffer('cluster_size', torch.zeros(codebook_size))
self.register_buffer('embed_avg', embed.clone())
self.learnable_codebook = learnable_codebook
if learnable_codebook:
self.embed = nn.Parameter(embed)
else:
self.register_buffer('embed', embed)
@torch.jit.ignore
def init_embed_(self, data):
if self.initted:
return
embed, cluster_size = kmeans(data, self.codebook_size, self.kmeans_iters)
self.embed.data.copy_(embed)
self.embed_avg.data.copy_(embed.clone())
self.cluster_size.data.copy_(cluster_size)
self.initted.data.copy_(torch.Tensor([True]))
def replace(self, samples, mask):
modified_codebook = torch.where(
mask[..., None],
sample_vectors(samples, self.codebook_size),
self.embed
)
self.embed.data.copy_(modified_codebook)
def expire_codes_(self, batch_samples):
if self.threshold_ema_dead_code == 0:
return
expired_codes = self.cluster_size < self.threshold_ema_dead_code
if not torch.any(expired_codes):
return
batch_samples = rearrange(batch_samples, '... d -> (...) d')
self.replace(batch_samples, mask = expired_codes)
@autocast(enabled = False)
def forward(self, x, used_codes=[]):
shape, dtype = x.shape, x.dtype
flatten = rearrange(x, '... d -> (...) d')
self.init_embed_(flatten)
embed = self.embed if not self.learnable_codebook else self.embed.detach()
embed = embed.t()
dist = -(
flatten.pow(2).sum(1, keepdim=True)
- 2 * flatten @ embed
+ embed.pow(2).sum(0, keepdim=True)
)
for uc in used_codes:
mask = torch.arange(0, self.codebook_size, device=x.device).unsqueeze(0).repeat(x.shape[0],1) == uc.unsqueeze(1)
dist[mask] = -torch.inf
embed_ind = gumbel_sample(dist, dim = -1, temperature = self.sample_codebook_temp)
embed_onehot = F.one_hot(embed_ind, self.codebook_size).type(dtype)
embed_ind = embed_ind.view(*shape[:-1])
quantize = F.embedding(embed_ind, self.embed)
# Perform the gumbel trick on the end result (during training)
if self.training:
quantize = flatten + (quantize - flatten).detach()
if self.training:
cluster_size = embed_onehot.sum(0)
self.all_reduce_fn(cluster_size)
ema_inplace(self.cluster_size, cluster_size, self.decay)
embed_sum = flatten.t() @ embed_onehot
self.all_reduce_fn(embed_sum)
ema_inplace(self.embed_avg, embed_sum.t(), self.decay)
cluster_size = laplace_smoothing(self.cluster_size, self.codebook_size, self.eps) * self.cluster_size.sum()
embed_normalized = self.embed_avg / cluster_size.unsqueeze(1)
self.embed.data.copy_(embed_normalized)
self.expire_codes_(x)
return quantize, embed_ind
class CosineSimCodebook(nn.Module):
def __init__(
self,
dim,
codebook_size,
kmeans_init = False,
kmeans_iters = 10,
decay = 0.8,
eps = 1e-5,
threshold_ema_dead_code = 2,
use_ddp = False,
learnable_codebook = False,
sample_codebook_temp = 0.
):
super().__init__()
self.decay = decay
if not kmeans_init:
embed = l2norm(torch.randn(codebook_size, dim))
else:
embed = torch.zeros(codebook_size, dim)
self.codebook_size = codebook_size
self.kmeans_iters = kmeans_iters
self.eps = eps
self.threshold_ema_dead_code = threshold_ema_dead_code
self.sample_codebook_temp = sample_codebook_temp
self.all_reduce_fn = distributed.all_reduce if use_ddp else noop
self.register_buffer('initted', torch.Tensor([not kmeans_init]))
self.register_buffer('cluster_size', torch.zeros(codebook_size))
self.learnable_codebook = learnable_codebook
if learnable_codebook:
self.embed = nn.Parameter(embed)
else:
self.register_buffer('embed', embed)
@torch.jit.ignore
def init_embed_(self, data):
if self.initted:
return
embed, cluster_size = kmeans(data, self.codebook_size, self.kmeans_iters,
use_cosine_sim = True)
self.embed.data.copy_(embed)
self.cluster_size.data.copy_(cluster_size)
self.initted.data.copy_(torch.Tensor([True]))
def replace(self, samples, mask):
samples = l2norm(samples)
modified_codebook = torch.where(
mask[..., None],
sample_vectors(samples, self.codebook_size),
self.embed
)
self.embed.data.copy_(modified_codebook)
def expire_codes_(self, batch_samples):
if self.threshold_ema_dead_code == 0:
return
expired_codes = self.cluster_size < self.threshold_ema_dead_code
if not torch.any(expired_codes):
return
batch_samples = rearrange(batch_samples, '... d -> (...) d')
self.replace(batch_samples, mask = expired_codes)
@autocast(enabled = False)
def forward(self, x, used_codes=[]):
shape, dtype = x.shape, x.dtype
flatten = rearrange(x, '... d -> (...) d')
flatten = l2norm(flatten)
self.init_embed_(flatten)
embed = self.embed if not self.learnable_codebook else self.embed.detach()
embed = l2norm(embed)
dist = flatten @ embed.t()
for uc in used_codes:
mask = torch.arange(0, self.codebook_size, device=x.device).unsqueeze(0).repeat(x.shape[0],1) == uc.unsqueeze(1)
dist[mask] = -torch.inf
embed_ind = gumbel_sample(dist, dim = -1, temperature = self.sample_codebook_temp)
embed_onehot = F.one_hot(embed_ind, self.codebook_size).type(dtype)
embed_ind = embed_ind.view(*shape[:-1])
quantize = F.embedding(embed_ind, self.embed)
# Perform the gumbel trick on the end result (during training)
if self.training:
quantize = flatten + (quantize - flatten).detach()
if self.training:
bins = embed_onehot.sum(0)
self.all_reduce_fn(bins)
ema_inplace(self.cluster_size, bins, self.decay)
zero_mask = (bins == 0)
bins = bins.masked_fill(zero_mask, 1.)
embed_sum = flatten.t() @ embed_onehot
self.all_reduce_fn(embed_sum)
embed_normalized = (embed_sum / bins.unsqueeze(0)).t()
embed_normalized = l2norm(embed_normalized)
embed_normalized = torch.where(zero_mask[..., None], embed,
embed_normalized)
ema_inplace(self.embed, embed_normalized, self.decay)
self.expire_codes_(x)
return quantize, embed_ind
# main class
class VectorQuantize(nn.Module):
def __init__(
self,
dim,
codebook_size,
n_embed = None,
codebook_dim = None,
decay = 0.8,
eps = 1e-5,
kmeans_init = False,
kmeans_iters = 10,
use_cosine_sim = False,
threshold_ema_dead_code = 0,
channel_last = True,
accept_image_fmap = False,
commitment_weight = None,
commitment = 1., # deprecate in next version, turn off by default
orthogonal_reg_weight = 0.,
orthogonal_reg_active_codes_only = False,
orthogonal_reg_max_codes = None,
sample_codebook_temp = 0.,
sync_codebook = False
):
super().__init__()
n_embed = default(n_embed, codebook_size)
codebook_dim = default(codebook_dim, dim)
requires_projection = codebook_dim != dim
self.project_in = nn.Linear(dim, codebook_dim) if requires_projection \
else nn.Identity()
self.project_out = nn.Linear(codebook_dim, dim) if requires_projection \
else nn.Identity()
self.eps = eps
self.commitment_weight = default(commitment_weight, commitment)
has_codebook_orthogonal_loss = orthogonal_reg_weight > 0
self.orthogonal_reg_weight = orthogonal_reg_weight
self.orthogonal_reg_active_codes_only = orthogonal_reg_active_codes_only
self.orthogonal_reg_max_codes = orthogonal_reg_max_codes
codebook_class = EuclideanCodebook if not use_cosine_sim \
else CosineSimCodebook
self._codebook = codebook_class(
dim = codebook_dim,
codebook_size = n_embed,
kmeans_init = kmeans_init,
kmeans_iters = kmeans_iters,
decay = decay,
eps = eps,
threshold_ema_dead_code = threshold_ema_dead_code,
use_ddp = sync_codebook,
learnable_codebook = has_codebook_orthogonal_loss,
sample_codebook_temp = sample_codebook_temp
)
self.codebook_size = codebook_size
self.accept_image_fmap = accept_image_fmap
self.channel_last = channel_last
@property
def codebook(self):
return self._codebook.embed
def forward(self, x, used_codes=None):
shape, device, codebook_size = x.shape, x.device, self.codebook_size
need_transpose = not self.channel_last and not self.accept_image_fmap
if self.accept_image_fmap:
height, width = x.shape[-2:]
x = rearrange(x, 'b c h w -> b (h w) c')
if need_transpose:
x = rearrange(x, 'b d n -> b n d')
x = self.project_in(x)
quantize, embed_ind = self._codebook(x, used_codes)
loss = torch.tensor([0.], device = device, requires_grad = self.training)
if self.training:
if self.commitment_weight > 0:
commit_loss = F.mse_loss(quantize.detach(), x)
loss = loss + commit_loss * self.commitment_weight
if self.orthogonal_reg_weight > 0:
codebook = self.codebook
if self.orthogonal_reg_active_codes_only:
# only calculate orthogonal loss for the activated codes for this batch
unique_code_ids = torch.unique(embed_ind)
codebook = codebook[unique_code_ids]
num_codes = codebook.shape[0]
if exists(self.orthogonal_reg_max_codes) and num_codes > self.orthogonal_reg_max_codes:
rand_ids = torch.randperm(num_codes, device = device)[:self.orthogonal_reg_max_codes]
codebook = codebook[rand_ids]
orthogonal_reg_loss = orthgonal_loss_fn(codebook)
loss = loss + orthogonal_reg_loss * self.orthogonal_reg_weight
quantize = self.project_out(quantize)
if need_transpose:
quantize = rearrange(quantize, 'b n d -> b d n')
if self.accept_image_fmap:
quantize = rearrange(quantize, 'b (h w) c -> b c h w', h = height, w = width)
embed_ind = rearrange(embed_ind, 'b (h w) -> b h w', h = height, w = width)
return quantize, embed_ind, loss