score_sde_pytorch
score_sde_pytorch copied to clipboard
yang-song
PyTorch implementation for Score-Based Generative Modeling through Stochastic Differential Equations (ICLR 2021, Oral)
Hello, thanks for your amazing work! I am wondering why scale neural network output by `standard deviation and flip sign` in VP score function, and NOT in VE score function?...
https://github.com/yang-song/score_sde_pytorch/blob/cb1f359f4aadf0ff9a5e122fe8fffc9451fd6e44/models/layers.py#L406-L416 The above code would create mismatched shortcut and output shape when dilation is larger than 1.
Thank you for open-sourcing your SDE code. However, when attempting to use the FID computation code in the provided Colab file, I encountered an error message stating that it cannot...
Hi, In the paper, the sampling equation of VP SDE (=: DDPM) differs from the DDPM's form. Specifically, the multiplicand to the score function is defined as $\beta_i$ while DDPM...
The code is [here](https://github.com/yang-song/score_sde_pytorch/blob/main/sampling.py#L254): ```Python class LangevinCorrector(Corrector): def __init__(self, sde, score_fn, snr, n_steps): super().__init__(sde, score_fn, snr, n_steps) if not isinstance(sde, sde_lib.VPSDE) \ and not isinstance(sde, sde_lib.VESDE) \ and not isinstance(sde,...
hello author, when I tried to run the checkpoint, an error occured: KeyError Traceback (most recent call last) Cell In[12], line 34 30 inverse_scaler = datasets.get_data_inverse_scaler(config) 32 # print(config) --->...
Hello Does this repo work on cpu? in particular the part in the OP folder? seems this only works only on Cuda? Are there some settings which can be set...
My question concerns [this line](https://github.com/yang-song/score_sde_pytorch/blob/cb1f359f4aadf0ff9a5e122fe8fffc9451fd6e44/controllable_generation.py#L73): ```python # Initial sample x = data * mask + sde.prior_sampling(data.shape).to(data.device) * (1. - mask) ``` Let's assume `data` is normalised to have approx `std=1`....
Hi, I am trying to run training/evaluation with 4 A100s. However, after some experiments I noticed that the training speed was same compared with process trained with a single GPU....
I try to perform the experiments on 8A100 GPUs. However, as I observed, the utilities of GPUs are quite low (
