Use quantizer from rosinality/vqvae with openai dvae

This commit is contained in:
James Betker 2021-08-06 14:06:26 -06:00
parent d3ace153af
commit 0799d95af5

View File

@ -7,6 +7,7 @@ import torch.nn.functional as F
from einops import rearrange from einops import rearrange
from torch import einsum from torch import einsum
from models.vqvae.vqvae import Quantize
from trainer.networks import register_model from trainer.networks import register_model
from utils.util import opt_get from utils.util import opt_get
@ -51,9 +52,6 @@ class DiscreteVAE(nn.Module):
hidden_dim = 64, hidden_dim = 64,
channels = 3, channels = 3,
smooth_l1_loss = False, smooth_l1_loss = False,
starting_temperature = 0.5,
temperature_annealing_rate = 0,
min_temperature = .5,
straight_through = False, straight_through = False,
normalization = None, # ((0.5,) * 3, (0.5,) * 3), normalization = None, # ((0.5,) * 3, (0.5,) * 3),
record_codes = False, record_codes = False,
@ -64,13 +62,9 @@ class DiscreteVAE(nn.Module):
self.num_tokens = num_tokens self.num_tokens = num_tokens
self.num_layers = num_layers self.num_layers = num_layers
self.starting_temperature = starting_temperature
self.current_temperature = starting_temperature
self.straight_through = straight_through self.straight_through = straight_through
self.codebook = nn.Embedding(num_tokens, codebook_dim) self.codebook = Quantize(num_tokens, codebook_dim)
self.positional_dims = positional_dims self.positional_dims = positional_dims
self.temperature_annealing_rate = temperature_annealing_rate
self.min_temperature = min_temperature
assert positional_dims > 0 and positional_dims < 3 # This VAE only supports 1d and 2d inputs for now. assert positional_dims > 0 and positional_dims < 3 # This VAE only supports 1d and 2d inputs for now.
if positional_dims == 2: if positional_dims == 2:
@ -130,14 +124,9 @@ class DiscreteVAE(nn.Module):
images.sub_(means).div_(stds) images.sub_(means).div_(stds)
return images return images
def update_for_step(self, step, __):
# Run the annealing schedule
if self.temperature_annealing_rate != 0:
self.current_temperature = max(self.starting_temperature * math.exp(-self.temperature_annealing_rate * step), self.min_temperature)
def get_debug_values(self, step, __): def get_debug_values(self, step, __):
# Report annealing schedule # Report annealing schedule
return {'current_annealing_temperature': self.current_temperature, 'histogram_codes': self.codes} return {'histogram_codes': self.codes}
@torch.no_grad() @torch.no_grad()
@eval_decorator @eval_decorator
@ -150,7 +139,7 @@ class DiscreteVAE(nn.Module):
self, self,
img_seq img_seq
): ):
image_embeds = self.codebook(img_seq) image_embeds = self.codebook.embed_code(img_seq)
b, n, d = image_embeds.shape b, n, d = image_embeds.shape
kwargs = {} kwargs = {}
@ -168,31 +157,18 @@ class DiscreteVAE(nn.Module):
self, self,
img img
): ):
device, num_tokens = img.device, self.num_tokens
img = self.norm(img) img = self.norm(img)
logits = self.encoder(img) logits = self.encoder(img).permute((0,2,3,1) if len(img.shape) == 4 else (0,2,1))
soft_one_hot = F.gumbel_softmax(logits, tau = self.current_temperature, dim = 1, hard = self.straight_through) sampled, commitment_loss, codes = self.codebook(logits)
sampled = sampled.permute((0,3,1,2) if len(img.shape) == 4 else (0,2,1))
if self.positional_dims == 1:
arrange = 'b n s, n d -> b d s'
else:
arrange = 'b n h w, n d -> b d h w'
sampled = einsum(arrange, soft_one_hot, self.codebook.weight)
out = self.decoder(sampled) out = self.decoder(sampled)
# reconstruction loss # reconstruction loss
recon_loss = self.loss_fn(img, out) recon_loss = self.loss_fn(img, out)
# kl divergence
arrange = 'b n h w -> b (h w) n' if self.positional_dims == 2 else 'b n s -> b s n'
logits = rearrange(logits, arrange)
log_qy = F.log_softmax(logits, dim = -1)
log_uniform = torch.log(torch.tensor([1. / num_tokens], device = device))
kl_div = F.kl_div(log_uniform, log_qy, None, None, 'batchmean', log_target = True)
# This is so we can debug the distribution of codes being learned. # This is so we can debug the distribution of codes being learned.
if self.record_codes: if self.record_codes:
codes = logits.argmax(dim = 2).flatten() codes = codes.flatten()
l = codes.shape[0] l = codes.shape[0]
i = self.code_ind if (self.codes.shape[0] - self.code_ind) > l else self.codes.shape[0] - l i = self.code_ind if (self.codes.shape[0] - self.code_ind) > l else self.codes.shape[0] - l
self.codes[i:i+l] = codes.cpu() self.codes[i:i+l] = codes.cpu()
@ -200,7 +176,7 @@ class DiscreteVAE(nn.Module):
if self.code_ind >= self.codes.shape[0]: if self.code_ind >= self.codes.shape[0]:
self.code_ind = 0 self.code_ind = 0
return recon_loss, kl_div, out return recon_loss, commitment_loss, out
@register_model @register_model
@ -214,4 +190,4 @@ if __name__ == '__main__':
#print(o.shape) #print(o.shape)
v = DiscreteVAE(channels=1, normalization=None, positional_dims=1) v = DiscreteVAE(channels=1, normalization=None, positional_dims=1)
o=v(torch.randn(1,1,256)) o=v(torch.randn(1,1,256))
print(o.shape) print(o[-1].shape)