Chenyu Zhiyun

What is Chenyu Zhiyun?

Chenyu Zhiyun is a GPU-based cloud computing platform designed for AI learning, creation, training, inference, and enterprise applications.

Users can create dedicated GPU instances as needed, and they can quickly launch environments such as ComfyUI, Stable Diffusion, and Jupyter using pre-installed images.

Main functions

• Rent exclusive GPU cloud instances.
• Purchase computing power on a pay-as-you-go or periodic basis.
• Select a pre-installed image from the AI app store.
• Run the Stable Diffusion WebUI.
• Run ComfyUI and manage workflows.
• Manage code and models through Jupyter.
• Upload models, LoRA, and project files.
• Training images and videos for LoRA.
• Quickly generate images and videos.
• Create reusable AI workflows.
• Manage instances through open APIs.
• Provide solutions for the privatized deployment of enterprises.

Which users are it suitable for

• AI painting users: No need for a high-end local graphics card.
• ComfyUI creator: Quickly use the pre-installed environment.
• Model trainer: Executes LoRA and fine-tuning tasks.
• Students and teachers: Conducting AI experiments and teaching.
• Universities and research institutions: Manage research computing resources.
• Research institutions: Perform training and reasoning tasks.
• AI companies: Deploy models and application services.
• Developer: Automatically manage computing power via API.

Morning Feather Smart Cloud Usage Guide

1. Register for a Chenyu Zhiyun account.
2. Complete the real-name verification as required.
3. Top up the balance or prepare a computing power card.
4. Choose pay-as-you-go or subscription-based billing.
5. View the available GPU models in real time.
6. Compare memory, performance, and price.
7. Select the mirror from the app store.
8. Set up the system drive and data drive.
9. Create and start a GPU instance.
10. Open WebUI, ComfyUI, or Jupyter.
11. Upload model and project files.
12. Run small tasks to verify the environment.
13. Monitor video memory, operation status, and balance.
14. Download the results and save the important data.
15. Shut down the system or release resources in a timely manner after the task is completed.

AI App Store

The AI app store provides pre-installed POD images, reducing the time required for users to set up drivers, frameworks, and creation tools from scratch.

ComfyUI cloud tutorial

1. Choose a GPU with appropriate video memory.
2. Create an image instance with ComfyUI.
3. Start the instance and open the quick access portal.
4. Import the workflow file.
5. Check for missing custom nodes.
6. Upload Checkpoints, VAEs, and LoRAs.
7. Adjust the model path and node parameters.
8. Use a low-resolution testing workflow.
9. Observe the video memory and runtime.
10. Increase the output specifications after confirming the results.
11. Download the output image or video.
12. Save the model and workflow before shutting down.

Location of Stable Diffusion files

The official tutorial specifies the common output directories, but the actual paths may still change depending on the version of the image.

Preparation for LoRA training

• Ensure that the training materials are legally authorized.
• Remove duplicate and low-quality images.
• Standardize sizes, labels, and naming.
• Select the VRAM specification based on the base template.
• First, test with small amounts of data and a low number of steps.
• Save the training parameters and model version.
• Export intermediate checkpoints regularly.
• Overfitting, as well as risks related to characters and brands, have been verified.

How to choose a GPU

Chenyu Zhiyun prices and billing

GPU models, inventory levels, and real-time prices may change; the official price page indicates the prices as displayed in the console at that time.

Notes on pay-as-you-go billing

• Charging starts as soon as the device is turned on.
• Charging does not occur only when the GPU is under full load.
• After shutdown, GPU resources are usually not reserved on a pay-as-you-go basis.
• Restarting may result in insufficient stock again.
• The lifting configuration and disk expansion will affect the costs.
• An insufficient balance may affect the execution of tasks.
• For long-term continuous use, the package cycle should be considered.

Precautions for periodic billing by package

• Pay in advance on a daily, weekly, monthly, or annual basis.
• The purchased GPU remains even after shutdown.
• The instance will be shut down upon expiration.
• Whether data is retained on a long-term basis is determined in accordance with the current rules.
• Check the end time of the pre-renewal inspection items.
• It still consumes its periodic value even when it is not in use.

Cost control methods

1. First, use small tasks to measure the actual runtime.
2. Record the time taken to load and generate the model.
3. Select the lowest suitable specification that meets the requirements for video memory.
4. Set automatic shutdown after the task is completed.
5. Avoid leaving instances idle while keeping them powered on.
6. Place the model on the persistent data disk.
7. Long-term tasks involve comparing package cycle prices.
8. Monitor costs through the billing API.
9. Set a budget and assign a responsible person for the team.
10. Clean up unused instances and data on a weekly basis.

Open API

The Chenyu Zhiyun v2 open API enables management of GPU instances, account bills, and available resources, making it suitable for automated scheduling of computing resources.

API Integration Tutorial

1. Register and complete the real-name verification.
2. Go to the API management page on the console.
3. Create a separate API Key.
4. Store the key on the server side.
5. Query available Pods and GPUs.
6. Search the image market and select an image.
7. Create a test instance.
8. Poll the instance startup status.
9. Connect the instance to execute tasks.
10. Call the stop interface after completion.
11. Check the bill to verify the costs.
12. Add timeout, rate limiting, and retry.
13. Record the operator and task number.
14. Regularly rotate and revoke old keys.

API security

• Do not expose keys on the web frontend.
• Different keys are used for testing and production.
• Hide the Authorization content in the logs.
• Restrict the servers that can call the API.
• Additional approval is required for creating high-priced instances.
• Set alerts for balance and abnormal bills.
• Perform rotation immediately in case of a suspected leak.

Data management

Cloud instances may contain models, training data, customer images, and generation results; it is necessary to distinguish between the system disk and persistent data before using them.

• Important documents should be stored on the data disk that can retain data.
• Check the rules before shutting down and releasing the instance.
• Download models and training results regularly.
• Use project directories to isolate different customers.
• Remove personal information from the material.
• Save the file checksum and version.
• Evaluate private deployment for sensitive data.

Environmental and network restrictions

• Officially, desktops or laptops are recommended.
• The interfaces of smartphones and tablets may not be displayed in their entirety.
• It is recommended to use the latest version of Chrome.
• Officials advise not to use a proxy when launching the app.
• Agents can cause instability in the WebUI connection.
• Open ports should have authentication and access control in place.

Compliance and Prohibitions

The official documentation explicitly prohibits mining on the platform, and accounts may be banned if such activity is detected.

• Comply with the service agreement and export control requirements.
• Models and plugins are used in accordance with their respective licenses.
• Do not train or generate illegal content.
• Real-person material requires portrait authorization.
• Corporate data must not be uploaded to public platforms without authorization.
• Keep invoices, bills, and project records.

Is it open source?

The Chenyu Zhiyun platform is a commercial cloud service, and to date no official open-source code repository for the entire platform has been found.

The ComfyUI, Stable Diffusion-related components, and Krita plugins contained within the image may be open source; it is necessary to check the respective repositories and licenses.

Product advantages

• There is no need to purchase high-end local GPUs.
• Offers exclusive graphics cards and various billing periods.
• Pre-installed with common AI creation and development images.
• Supports model, workflow, and LoRA training.
• APIs for managing open instances and bills.
• It covers personal, research, and corporate use cases.
• Private deployment is available.

Usage restrictions

• The price per GPU and inventory levels change in real time.
• Pay-as-you-go instances incur charges as soon as they are started up.
• After shutdown, the GPU may no longer be reserved.
• The mobile experience may be incomplete.
• Proxy networks can cause connection issues.
• There are compatibility risks between mirrors and nodes.
• Sensitive data requires additional protection.
• Mining is strictly prohibited.

Frequently Asked Questions

What does Chenyu Zhiyun do?

It offers cloud GPU computing power, AI images, tools for image and video creation, ComfyUI, Stable Diffusion, Jupyter, LoRA training, and open APIs.

How much does Chenyu Zhiyun charge?

GPUs support billing on a per-use, per-day, per-week, per-month, and per-year basis; the price per unit varies in real time and should be based on the information displayed in the console.

Is there still a charge after shutting it down?

When a pay-as-you-go instance is shut down, billing stops but the GPU is no longer reserved; for subscription-based instances, the resources remain available even after shutdown.

Does Chenyu Zhiyun support ComfyUI?

Supported: It is possible to choose an image with ComfyUI pre-installed, upload models and workflows, and run image or video processing tasks using cloud-based GPUs.

Does Chenyu Zhiyun provide APIs?

The v2 API is available to manage instances, check balances and bills, as well as to access resources such as Pods, GPUs, and images.

Can Chenyu Zhiyun be used for mining?

No, the official documentation explicitly prohibits mining; violating this rule may result in the account being banned.

Is Chenyu Zhiyun an open-source platform?

No, the platform itself is a commercial cloud service; the open-source components contained in the images must be used in accordance with their respective licenses.