chainneverliu / pytorch-ddpm

基于pytorch框架从零实现DDPM算法

Pytorch-DDPM

基于pytorch框架从零实现DDPM算法

详细的介绍可参考本人博客：https://zhuanlan.zhihu.com/p/617895786

pytorch-ddpm's Issues

Could you kindly provide the trained BestModel.pth file?

Thank you for sharing your insights on DDPM developed in PyTorch. Would it be possible for you to upload a trained BestModel.pth file for users who are testing this model for the first time? I would greatly appreciate it if you could do so.

没有ｃｏｎｄｉｃｔｉｏｎ如何确保生成那个类别

没有ｃｏｎｄｉｃｔｉｏｎ如何确保生成那个类别

请问如何训练自己的数据集

在多个GPU上采样失败。

在1个GPU上可以顺利执行image_test.py，最后生成了一张图片。

但是在4个GPU上执行时，报错：

Traceback (most recent call last):
File "/home/ubuntu/dev/Pytorch-DDPM/DiffusionModels/image_test.py", line 103, in
samples = DDPM(mode="generate", image_size=image_size, batch_size=64, channels=channels)
File "/data/miniconda3/envs/diffusion/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl
return forward_call(*input, **kwargs)
File "/data/miniconda3/envs/diffusion/lib/python3.9/site-packages/torch/nn/parallel/data_parallel.py", line 168, in forward
outputs = self.parallel_apply(replicas, inputs, kwargs)
File "/data/miniconda3/envs/diffusion/lib/python3.9/site-packages/torch/nn/parallel/data_parallel.py", line 178, in parallel_apply
return parallel_apply(replicas, inputs, kwargs, self.device_ids[:len(replicas)])
File "/data/miniconda3/envs/diffusion/lib/python3.9/site-packages/torch/nn/parallel/parallel_apply.py", line 86, in parallel_apply
output.reraise()
File "/data/miniconda3/envs/diffusion/lib/python3.9/site-packages/torch/_utils.py", line 434, in reraise
raise exception
AssertionError: Caught AssertionError in replica 0 on device 0.
Original Traceback (most recent call last):
File "/data/miniconda3/envs/diffusion/lib/python3.9/site-packages/torch/nn/parallel/parallel_apply.py", line 61, in _worker
output = module(*input, **kwargs)
File "/data/miniconda3/envs/diffusion/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl
return forward_call(*input, **kwargs)
File "/home/ubuntu/dev/Pytorch-DDPM/DiffusionModels/diffusionModels/simpleDiffusion/simpleDiffusion.py", line 127, in forward
return self.sample(image_size=kwargs["image_size"],
File "/data/miniconda3/envs/diffusion/lib/python3.9/site-packages/torch/autograd/grad_mode.py", line 28, in decorate_context
return func(*args, **kwargs)
File "/home/ubuntu/dev/Pytorch-DDPM/DiffusionModels/diffusionModels/simpleDiffusion/simpleDiffusion.py", line 105, in sample
return self.p_sample_loop(shape=(batch_size, channels, image_size, image_size))
File "/data/miniconda3/envs/diffusion/lib/python3.9/site-packages/torch/autograd/grad_mode.py", line 28, in decorate_context
return func(*args, **kwargs)
File "/home/ubuntu/dev/Pytorch-DDPM/DiffusionModels/diffusionModels/simpleDiffusion/simpleDiffusion.py", line 90, in p_sample_loop
assert list(self.denoise_model.parameters()), "model.parameters() is empty"
AssertionError: model.parameters() is empty

除了在DDPM.to(device)后面加了一行DDPM = nn.DataParallel(DDPM)以外，代码基本没变。