aljaz/DALLE2-pytorch
1
1
Fork 0
mirror of https://github.com/lucidrains/DALLE2-pytorch.git synced 2025-12-19 09:44:19 +01:00
Code Issues Packages Projects Releases Wiki Activity

revert back to old upsampling, paper does not work

Browse Source
This commit is contained in:
Phil Wang
2022-04-26 07:39:04 -07:00
parent 45262a4bb7
commit 0b28ee0d01
3 changed files with 2 additions and 12 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
README.md
View File

@@ -577,14 +577,4 @@ Once built, images will be saved to the same directory the command is invoked
} }
``` ```
```bibtex
@article{Arar2021LearnedQF,
title = {Learned Queries for Efficient Local Attention},
author = {Moab Arar and Ariel Shamir and Amit H. Bermano},
journal = {ArXiv},
year = {2021},
volume = {abs/2112.11435}
}
```
*Creating noise from data is easy; creating data from noise is generative modeling.* - Yang Song's <a href="https://arxiv.org/abs/2011.13456">paper</a> *Creating noise from data is easy; creating data from noise is generative modeling.* - Yang Song's <a href="https://arxiv.org/abs/2011.13456">paper</a>

2
dalle2_pytorch/dalle2_pytorch.py
View File

@@ -693,7 +693,7 @@ class DiffusionPrior(nn.Module):
# decoder # decoder
def Upsample(dim): def Upsample(dim):
return QueryAttnUpsample(dim) return nn.ConvTranspose2d(dim, dim, 4, 2, 1)
def Downsample(dim): def Downsample(dim):
return nn.Conv2d(dim, dim, 4, 2, 1) return nn.Conv2d(dim, dim, 4, 2, 1)

2
dalle2_pytorch/vqgan_vae.py
View File

@@ -378,7 +378,7 @@ class VQGanVAE(nn.Module):
for layer_index, (dim_in, dim_out), layer_num_resnet_blocks, layer_use_attn in zip(range(layers), dim_pairs, num_resnet_blocks, use_attn): for layer_index, (dim_in, dim_out), layer_num_resnet_blocks, layer_use_attn in zip(range(layers), dim_pairs, num_resnet_blocks, use_attn):
append(self.encoders, nn.Sequential(nn.Conv2d(dim_in, dim_out, 4, stride = 2, padding = 1), leaky_relu())) append(self.encoders, nn.Sequential(nn.Conv2d(dim_in, dim_out, 4, stride = 2, padding = 1), leaky_relu()))
prepend(self.decoders, nn.Sequential(QueryAttnUpsample(dim_out), nn.Conv2d(dim_out, dim_in, 3, padding = 1), leaky_relu())) prepend(self.decoders, nn.Sequential(nn.ConvTranspose2d(dim_out, dim_in, 4, 2, 1), leaky_relu()))
if layer_use_attn: if layer_use_attn:
prepend(self.decoders, VQGanAttention(dim = dim_out, heads = attn_heads, dim_head = attn_dim_head, dropout = attn_dropout)) prepend(self.decoders, VQGanAttention(dim = dim_out, heads = attn_heads, dim_head = attn_dim_head, dropout = attn_dropout))