61 lines
2.4 KiB
Markdown
61 lines
2.4 KiB
Markdown
# Documentation
|
|
- Class name: ImageUpscaleWithModel
|
|
- Category: image/upscaling
|
|
- Output node: False
|
|
- Repo Ref: https://github.com/comfyanonymous/ComfyUI
|
|
|
|
ImageUpscaleWithModel节点旨在使用指定的放大模型增强输入图像的分辨率。它智能地管理系统内存，以防止内存不足错误，确保平滑的放大过程。节点的功能集中在提高图像质量，同时不牺牲性能。
|
|
|
|
# Input types
|
|
## Required
|
|
- upscale_model
|
|
- upscale_model参数对于节点至关重要，因为它定义了用于放大图像的模型。它直接影响上采样过程和生成的图像质量。
|
|
- Comfy dtype: torch.nn.Module
|
|
- Python dtype: torch.nn.Module
|
|
- image
|
|
- image参数是必需的，因为它代表了节点将处理的输入图像数据。其结构和内容是上采样操作成功的关键。
|
|
- Comfy dtype: torch.Tensor
|
|
- Python dtype: torch.Tensor
|
|
|
|
# Output types
|
|
- IMAGE
|
|
- 输出参数IMAGE代表了节点生成的放大图像。它标志着上采样过程的成功完成，是节点功能的主要成果。
|
|
- Comfy dtype: torch.Tensor
|
|
- Python dtype: torch.Tensor
|
|
|
|
# Usage tips
|
|
- Infra type: GPU
|
|
|
|
# Source code
|
|
```
|
|
class ImageUpscaleWithModel:
|
|
|
|
@classmethod
|
|
def INPUT_TYPES(s):
|
|
return {'required': {'upscale_model': ('UPSCALE_MODEL',), 'image': ('IMAGE',)}}
|
|
RETURN_TYPES = ('IMAGE',)
|
|
FUNCTION = 'upscale'
|
|
CATEGORY = 'image/upscaling'
|
|
|
|
def upscale(self, upscale_model, image):
|
|
device = model_management.get_torch_device()
|
|
upscale_model.to(device)
|
|
in_img = image.movedim(-1, -3).to(device)
|
|
free_memory = model_management.get_free_memory(device)
|
|
tile = 512
|
|
overlap = 32
|
|
oom = True
|
|
while oom:
|
|
try:
|
|
steps = in_img.shape[0] * comfy.utils.get_tiled_scale_steps(in_img.shape[3], in_img.shape[2], tile_x=tile, tile_y=tile, overlap=overlap)
|
|
pbar = comfy.utils.ProgressBar(steps)
|
|
s = comfy.utils.tiled_scale(in_img, lambda a: upscale_model(a), tile_x=tile, tile_y=tile, overlap=overlap, upscale_amount=upscale_model.scale, pbar=pbar)
|
|
oom = False
|
|
except model_management.OOM_EXCEPTION as e:
|
|
tile //= 2
|
|
if tile < 128:
|
|
raise e
|
|
upscale_model.cpu()
|
|
s = torch.clamp(s.movedim(-3, -1), min=0, max=1.0)
|
|
return (s,)
|
|
``` |