Location>code7788 >text

ComfyUI plugin: efficiency-nodes-comfyui nodes

Popularity:60 ℃/2024-08-06 12:04:20

Preface:

Learning ComfyUI is a long battle, efficiency-nodes-comfyui is to improve the efficiency of workflow creation tools, including efficiency nodes to integrate the basic functions of the workflow, such as Efficient Loader nodes equivalent to Load Checkpoint + Clip set layer + Load VAE and so on, and the plug-in provides a simpler and quicker X/Y comparison charts, which can further improve the efficiency of the assessment work. I wish you all good luck in your studies and become a master of ComfyUI as soon as possible!

catalogs

I. Installation method

II. Efficient Loader nodes

III. KSampler Adv. (Efficient) node

IV. Lora stack/Controlnet Stacker node

V. XY nodes

VI. XY Plot nodes

 

I. Installation method

Inside the ComfyUI home directory type CMD and enter.

1

Type git clone xxx in the CMD command line that pops up to start the download.

2

Enter the following line of code in the terminal to start the download

git clone /jags111/

 

II. Efficient Loader nodes

This node is a node for loading efficient deep learning models. This node is designed to provide fast and accurate image processing capabilities by loading pre-trained efficient models.

3

Important parameters:

lora_stack → connectable lora model loading stack ** e.g. CR libraries and own libraries own nodes

cnet_stack → Connectable ControlNet Model Loading Stack

token normalization → token normalization, i.e., the way to set text encoding

weight interpretation → weight initialization, base setup parameters for the model

DEPENDENCIES → useful for subsequent X/Y comparison tests

Note: The following figure shows the comparison of four different token normalizations under the same parameters, and there is almost no effect on the results. Five different weight interpretations were tried, and the results were also unaffected.

4

5

Usage Scenarios:

- Fast image processing: fast image processing tasks such as denoising, restoration, enhancement, etc. using efficient models.

- Image Recognition and Classification: image recognition and classification tasks using efficient models that provide accurate results.

- Automated Processing: Efficient and accurate image processing using efficient models in an automated image processing process.

By using the Efficient Loader node, you can achieve efficient model loading and application in the image processing workflow, improving the speed and effectiveness of image processing.

 

III. KSampler Adv. (Efficient) node

The node focuses on efficient image sampling and generation with advanced sampling techniques and optimized algorithms for fast and high quality image processing.

6

Important parameters:

script → related to X/Y testing

add_noise → whether or not to add noise during graph generation **This option is only relevant for the ancestral sampler.

Randomize/last Queued Seed → click on the left to randomize a noise, on the right to use the noise from the last subgenerated map

return_with_leftover_noise → whether or not to perform a complete denoising process, said to affect the retention of detail in the picture

preview method → Set the preview method for the denoising process, the same as the manager's preview method.

vae_decode → when we pass in optional_vae, selecting false will not output the image, selecting true will do so

Usage Scenarios:

- Efficient Image Generation: In scenes where high quality images need to be generated quickly, image generation is realized using efficient sampling techniques.

- Image Enhancement: Enhance and optimize images through advanced sampling techniques to improve image quality.

- Automated Processing: Efficient and accurate image processing through efficient sampling algorithms in an automated image processing process.

By using the KSampler Adv. (Efficient) node, efficient image sampling and generation can be realized in the image processing workflow, improving the speed and quality of image processing.

 

IV. Lora stack/Controlnet Stacker node

The Lora Stack node focuses on image generation and processing by stacking multiple Lora models, a pre-trained model used to enhance image generation capabilities, and the Lora Stack node can load and stack multiple Lora models to achieve more sophisticated and high-quality image processing results.

The Controlnet Stacker node focuses on stacking multiple ControlNet models for image generation and processing.ControlNet is a neural network that controls the generation of images, and by stacking multiple ControlNet models, more complex image control and generation effects can be achieved.

7

Important parameters:

lora_stack → tandem can load multiple lora models

Lora_count → Changing this value increases the number of loadable lora synchronously

Control_net → multiple ControlNet models can be loaded in series

input_mode → select simple to simply set lora weights, select advanced to enable large model weights.

When the advanced option is turned on, you can change the mod weights.

8

Example 1: The following figure shows a series of ControlNet, one of which is a tile to control the elements of the picture, and the second one is an openpose to control the skeleton of the character, which affects the final effect of the picture through the series.

9

Example 2: The following figure shows a multiple Lora loading example workflow, where the number of lora loaded on this node is increased by changing lora_count.

10

Usage Scenarios:

- Complex Image Generation: Enhancements are achieved by overlaying multiple Lora/ControlNet models in tasks that require complex and high-quality image generation.

- Image processing optimization: taking advantage of multiple Lora/ControlNet models for image optimization and enhancement.

By using Lora Stack and Controlnet Stacker nodes, efficient model stacking and application can be realized in the image processing workflow to enhance the complexity and quality of image processing and meet various complex image processing needs.

 

V. XY nodes

The XY node focuses on parameter sweeping during image processing and generation. By setting different parameter values on the X and Y axes respectively to generate a series of images, it is convenient for users to observe and compare the effects of different parameter combinations on the image effect.

11

Important parameters:

first_xxx → start parameter selection

last_xxx → parameter selection for output ** will automatically populate the overprocess based on batch_count

Example: Compare the performance of three large models, under three different CFG values, through the comparison can be more clearly found the advantages and disadvantages of the model.

12

Usage Scenarios:

- Parameter Optimization: Optimize the image processing parameters by comparing the images generated by different parameter combinations to obtain the best results.

- Experiments and tests: experiments and tests are conducted during image processing to observe the effect of parameter changes on the results.

- Image generation: in the image generation task, diverse generation results are obtained by parameter sweeping.

By using XY nodes, efficient parameter sweeping and optimization can be performed during image processing and generation to enhance the effectiveness and quality of image processing.

 

VI. XY Plot nodes

The XY Plot node focuses on generating and presenting a series of images through 2D parameter sweeps. By setting different parameter values on the X-axis and Y-axis, you can visually compare and analyze the effects of parameter changes on image results, thus optimizing image processing parameters.

13

Important parameters:

grip_spacing → output the size of the seams between the comparison images

XY_flip → flip X,Y

Y_label_orientation → set whether Y-axis labels are displayed vertically or vertically.

ksampler_output_image → Select image to output as an image and set the Plot output to be a large image synthesized from the comparison plot.

Example: When XY_flip is turned on and the Y-axis label is set to display vertically, the final output is shown below.

14

Usage Scenarios:

- Parameter Optimization: Find the best image processing parameter settings by comparing images generated by different parameter combinations.

- Experiments and tests: Parameter experiments and tests are conducted during image processing to observe the effect of parameter changes on the results.

- Image generation: In the image generation task, diverse generation results are obtained through parameter sweeps to find the optimal combination of generation parameters.

By using the XY Plot node, efficient parameter sweeping and optimization can be performed during image processing and generation to enhance the effect and quality of image processing and meet various complex image processing needs.

**To go beyond oneself is to strive for excellence. Perseverance is the key to success. **