Calculate the X4 kernel from the X2 kernel (for proof see appendix in paper)

Kai Zhang (github: https://github.com/cszn)

Kai Zhang (github: https://github.com/cszn)

Create a beta schedule that discretizes the given alpha_t_bar function, which defines the cumulative product of (1-beta) over time from t = [

x: image, NxcxHxW k: kernel, Nx1xhxw

x: image, NxcxHxW k: kernel, Nx1xhxw

calculate SSIM the same outputs as MATLAB's img1, img2: [0, 255]

blur + downsampling Args: x: HxWxC image, [0, 1]/[0, 255] k: hxw, double sf: down-scale factor Return: downsa

blur + downsampling Args: x: HxWxC image, [0, 1]/[0, 255] k: hxw, double sf: down-scale factor Return: downsa

Convert primitive modules to float16.

Convert primitive modules to float16.

Convert primitive modules to float32, undoing convert_module_to_f16().

Convert primitive modules to float32, undoing convert_module_to_f16().

Convert the torso of the model to float16.

Convert the torso of the model to float16.

Convert the torso of the model to float16.

Convert the torso of the model to float32.

Convert the torso of the model to float32.

Convert the torso of the model to float32.

Convert the torso of the model to float32.

Convert the torso of the model to float32.

A counter for the `thop` package to count the operations in an attention operation. Meant to be used like: macs

Create a noise-conditional image classifier. Args: prng_key: A JAX random state. batch_size: The batch size of input data. ckpt_path: T

Create the score model.

1x1 convolution with DDPM initialization.

1x1 convolution with DDPM initialization.

This is the degradation model of BSRGAN from the paper "Designing a Practical Degradation Model for Deep Blind Image Super-Resolution" --

This is the degradation model of BSRGAN from the paper "Designing a Practical Degradation Model for Deep Blind Image Super-Resolution" --

This is an extended degradation model by combining the degradation models of BSRGAN and Real-ESRGAN ---------- img: HXWXC, [0, 1], it

This is the degradation model of BSRGAN from the paper "Designing a Practical Degradation Model for Deep Blind Image Super-Resolution" --

This is the degradation model of BSRGAN from the paper "Designing a Practical Degradation Model for Deep Blind Image Super-Resolution" --

Initializer for the final dense layer. Args: key: PRNG key to use to sample the weights. shape: Shape of the tensor to initialize. dtyp

Overwrite model.train with this function to make sure train/eval mode does not change anymore.

Overwrite model.train with this function to make sure train/eval mode does not change anymore.

Create discretized iteration rules for the reverse diffusion sampler.

DDPM discretization.

SMLD(NCSN) discretization.

Create discretized iteration rules for the reverse diffusion sampler.

DDPM discretization.

SMLD(NCSN) discretization.

r"""Downsample a batch of 2D images with the given filter. Accepts a batch of 2D images of the shape `[N, C, H, W]` or `[N, H, W, C]` and dow

r"""Downsample a batch of 2D images with the given filter. Accepts a batch of 2D images of the shape `[N, C, H, W]` or `[N, H, W, C]` and dow

The DPM-Solver sampler funciton. Args: model: A score model. Returns: Samples, number of function evaluations.

The DPM-Solver sampler funciton. Args: rng: A JAX random state state: A `flax.struct.dataclass` object that represents the training

bicubic downsampling + blur Args: x: HxWxC image, [0, 1] k: hxw, double sf: down-scale factor Return: downsam

bicubic downsampling + blur Args: x: HxWxC image, [0, 1] k: hxw, double sf: down-scale factor Return: downsam

Write all of the diagnostics from the current iteration

The dynamic thresholding method.

The dynamic thresholding method.

The dynamic thresholding method.

The dynamic thresholding method.