Stuck on an issue?

Lightrun Answers was designed to reduce the constant googling that comes with debugging 3rd party libraries. It collects links to all the places you might be looking at while hunting down a tough bug.

And, if you’re still stuck at the end, we’re happy to hop on a call to see how we can help out.

Facing "RunTimeError" for Usage code

See original GitHub issue

Hello @lucidrains,

I have tried running the below code:

import torch
from imagen_pytorch import Unet, Imagen

unet1 = Unet(
    dim = 32,
    cond_dim = 512,
    dim_mults = (1, 2, 4, 8),
    num_resnet_blocks = 3,
    layer_attns = (False, True, True, True),
    layer_cross_attns = (False, True, True, True)
)

unet2 = Unet(
    dim = 32,
    cond_dim = 512,
    dim_mults = (1, 2, 4, 8),
    num_resnet_blocks = (2, 4, 8, 8),
    layer_attns = (False, False, False, True),
    layer_cross_attns = (False, False, False, True)
)

imagen = Imagen(
    unets = (unet1, unet2),
    image_sizes = (64, 256),
    timesteps = 1000,
    cond_drop_prob = 0.1
).cuda()

text_embeds = torch.randn(4, 256, 768).cuda()
text_masks = torch.ones(4, 256).bool().cuda()
images = torch.randn(4, 3, 256, 256).cuda()

for i in (1, 2):
    loss = imagen(images, text_embeds = text_embeds, text_masks = text_masks, unet_number = i)
    loss.backward()

images = imagen.sample(texts = [
    'a whale breaching from afar',
    'young girl blowing out candles on her birthday cake',
    'fireworks with blue and green sparkles'
], cond_scale = 3.)

images.shape # (3, 3, 256, 256)

After executing the above code I am facing the below error : RuntimeError: Expected weight to be a vector of size equal to the number of channels in input, but got weight of shape [256] and input of shape [4, 128, 8, 8]

I am getting this error for the below line: loss = imagen(images, text_embeds = text_embeds, text_masks = text_masks, unet_number = i)