[NCHC Exploration] #2 Getting Started with TWCC

The cover Image was captured on April 11, 2025, from the Taiwan AI Cloud.

Introduction

The Taiwan Computing Cloud (TWCC) is a platform under the National Center for High-Performance Computing (NCHC), offering AI technology development and cloud computing services. TWCC features rapid deployment, efficient orchestration, intelligent computing, and centralized data integration. By leveraging advanced container and GPU technologies, it enables the swift creation of secure and flexible computing environments. These capabilities support large-scale parallel processing and efficient data management, accelerating development workflows and meeting diverse application demands.

Services Offered

Container Compute Service (CCS)

TWCC provides two types of container-based computing services:

• Interactive Containers
• Scheduled Containers

Interactive Containers allow users to write, debug, and run code online using Jupyter Notebook, with a user experience similar to Google Colaboratory. However, users must be mindful of resource usage, and be sure to delete containers after use to avoid continuously occupying computing resources and incurring additional costs.

Scheduled Containers are ideal for computational tasks that can be planned in advance. Users can specify the start time, duration, and required resources for each task. Once the task is completed, the system automatically releases the resources, helping to reduce development costs effectively.

High-Performance Computing (HPC)

TWCC’s High-Performance Computing (HPC) jobs are executed using the Taiwania 2 supercomputer, which is designed to handle various AI and high-performance computing tasks.

Taiwania 2 is composed of 252 computing nodes, each equipped with 2 CPUs and 8 GPUs. In total, it features 9,072 Intel Xeon Gold CPU cores and 2,016 NVIDIA Tesla V100 32GB GPUs. Its overall computational power reaches up to 9 petaflops (PFLOPS), making it one of the top-tier high-performance computing systems in the world.

Image source

Taiwania 2 uses the international open-source software Slurm as its resource scheduling tool. Slurm is a highly fault-tolerant and scalable cluster management system that supports both large and small Linux cluster environments. It is widely adopted by supercomputing systems around the world.

Through Slurm, users can obtain either exclusive or shared computing resources within a specified time frame. It also provides a management framework for resource allocation, execution, and monitoring across compute nodes. Additionally, it enables resource arbitration through scheduling management.

Slurm can be integrated with Singularity container technology, allowing users to execute various AI computational tasks using Docker images provided by the HPC service.

The HPC service utilizes a 100 Gb/s high-speed InfiniBand network to connect GPU hosts, offering extremely high throughput and ultra-low latency. Data transmission speeds are over 14 times faster compared to traditional 10GbE networks.

Image source

Thanks to the 32 GB memory capacity of the GPUs used in the HPC service, training data can be fully loaded into the workflow. This significantly enhances model training efficiency and prediction accuracy, resulting in an overall performance improvement of approximately 15%.

Image source

Virtual Compute Service (VCS)

TWCC’s Virtual Compute Service covers the following three main categories:

• Computation

  • Virtual Compute Service (VCS)

• Storage

  • Virtual Disk Service (VDS)

• Networking and Security

  • Virtual Network (VNW)
  • Load Balancer Service (LBS)
  • Auto Scaling
  • Security Groups
  • Basic Virtual Firewall (BVF)

Users can create various operating systems through an online interface and freely select hardware resources such as CPU and memory capacity based on their needs, allowing for the flexible construction of custom computing environments. Additionally, virtual instances can be accessed via SSH (Secure Shell), ensuring that data transmission is encrypted to enhance information security and effectively defend against malicious attacks.

TWCC also provides an image backup system, enabling users to save system configurations and critical data from their virtual instances to the cloud. In case of an emergency, this allows for quick restoration of the system and data, significantly reducing the risk of loss.

Furthermore, the Virtual Compute Service includes automatic load balancing and scaling mechanisms that can automatically detect the system’s resource usage and elastically expand or reduce processing environments and resources. This prevents service interruptions due to system overload and ensures stable operation of computational tasks.

Cloud Object Storage (COS)

TWCC offers a variety of cloud storage services, including:

• High-speed File System (HFS)
• Cloud Object Storage (COS)

These storage systems are designed to integrate seamlessly with Container Compute Services (CCS), High-Performance Computing Jobs (HPC Jobs), and the Taiwania 2 supercomputer (TWNIA2), providing fast, secure, and stable data access capabilities. Users can manage and access their storage via terminal commands or a graphical web portal, with optional notification settings to stay informed about computing and storage status.

TWCC’s cloud storage services are equipped with high-speed read/write capabilities, supporting the real-time processing of large volumes of data and ensuring uninterrupted large-scale AI parallel computing operations. Each account is granted 100 GB + 100 GB of free HFS storage upon registration. All storage is flexibly billed on a per GB per day basis, allowing users to scale according to actual needs.

Additionally, the cloud storage service includes an automatic backup mechanism, with data simultaneously stored on tape devices to provide an extra layer of protection for important files, helping to reduce the risk of data loss in the event of a disaster.

Pricing

TWCC adopts a pay-as-you-go billing model, calculated per second. This flexible approach allows users to start and stop resource usage at any time based on their budget and program requirements with no long-term contracts are required. As of April 11, 2025, the pricing for various services is as follows:

Container Compute Service

The following container types and prices apply to both Interactive Containers and Scheduled Containers.

High-performance Computing Service

The following table lists the HPC job types and their respective pricing:

Virtual Compute Service

The Virtual Compute Service (VCS) is categorized into CPU instances and GPU instances.

CPU Instance Types

GPU Instance Types

Bandwidth & Traffic Charges

TWCC’s VCS only charges for outbound traffic. All other types of traffic are free of charge:

The total outbound traffic generated by all virtual instances is calculated and billed monthly.

Cloud Object Storage

Cloud Object Storage (COS) can be used as a standalone general-purpose storage space, or as a data storage and transmission medium for TWCC’s Container Compute Service and High-Performance Computing service.

Virtual Disk Service

The Virtual Disk Service (VDS) provides storage space for TWCC’s VCS, offering high availability, consistency, and low latency.

Hyper File System

The Hyper File System (HFS) can be used as independent general-purpose storage or as a medium for storing and transferring computing data within TWCC’s Container Compute Service and High-Performance Computing services.

Upon account registration, users receive the following free storage:

HFS allows users to flexibly increase or decrease storage capacity. Adjustments take effect immediately upon request and are billed against the project’s available quota.

Preparation

Before using TWCC’s computing services, you should prepare the following:

1. An iService account with NCHC (National Center for High-performance Computing). If you don’t have one, refer to [NCHC Exploration] #1 Account Creation to register.

2. A budget allocated for your project.

3. An SSH client. Linux and macOS users can use the built-in SSH terminal. Windows 10/11 includes the built-in OpenSSH client. If you’re using another OS version or encounter firewall-related issues preventing SSH usage, please seek an alternative solution.

The following section will demonstrate how to connect to NCHC using container computing services with Windows 11 as the client system.

Log in Account

If you haven’t created any projects yet, the first step is to log in to your account and create a project on the iService Computing Resource Service Network.

Step 1: Click 登入 (Login).

Step 2: Enter your previously registered iService account.

Step 3: Click 下一步 (Next).

If your account is linked to Google, you will be redirected to log in through your Google account.

If you are not logged into Google, you will be prompted to log in to your linked Google account.

If you are already logged into Google, simply select your previously linked Google account.

Creating a Project

Next, to use TWCC’s computing resources, you need to create a project for your upcoming tasks. Start by returning to the iService Computing Resource Service Network homepage.

Step 4: Click 會員中心 (Member Center).

Step 5: Click 計畫管理 (Project Management).

Step 6: Click 我的計畫 (My Projects).

Step 7: Click 新增 (Add).

Choose the project type that best fits your needs. Click 說明 (Details) to learn more about the eligibility and how to apply.

Below is an overview of the available plans for individuals and enterprises.

After selecting your desired project, click 申請 (Apply) to view the TWCC Taiwan AI Cloud Service Level Agreement. Please read it carefully.

Step 8: After reading the agreement, click 我同意 (I Agree).

Next, follow the steps to fill out the project creation form.

Step 9: Fill in the project information.

Step 10: Select the services you wish to apply for.

Step 11: Click 下一步 (Next).

Next, follow the steps to complete the project information.

After submission, the newly created project will appear in your project list. If a principal investigator (PI) of another project has added you to their team, you will also see that project listed here.

The following steps use an 學術計畫 (Academic Project) as an example, the process is the same for all types of projects.

Step 12: Click the project you want to activate.

Once inside the project, you’ll see all relevant information. This is where you can manage project components such as members, wallet, and other settings.

Before using any TWCC resources, you must add budget to your project wallet. Only when your project wallet has sufficient quota can you proceed with resource usage.

Step 13: Click 錢包管理 (Wallet Management).

Step 14: Click 購買額度 (Purchase Quota).

Next, follow the on-screen steps to complete the purchase. Afterward, you can verify the Remaining Quota under 錢包管理 (Wallet Management) to confirm the quota has been successfully added.

As a side note, clicking 用量統計 (Usage Statistics) will show each project member’s usage of the project wallet.

Introduction to TWCC

After creating a project, you can proceed to use the TWCC resources.

Step 1: From the iService Computing Resource Service Network homepage, click 服務介紹 (Service Introduction).

Step 2: Click 計算服務 (Compute Services).

Step 3: Click 臺灣 AI 雲服務 (Taiwan AI Cloud Service).

Step 4: Click 登入 (Login).

Step 5: Enter your member account.

Step 6: Click 下一步 (Next).

The first time you enter TWCC, there will be a guided tour where you can follow the prompts to explore the various features.

Below is the TWCC Quick Tour after being organized in the 計畫資訊 (Project Information) page.

Step 7: Click the 選取計畫或專案 (Select Plan or Project) block at the top. Here, you can choose the plan or project you wish to change.

Step 8: Select the plan/project you wish to change.

Step 9: Click 確認 (Confirm).

Next, you will return to the 計畫資訊 (Project Information) page. Click 查看服務概況 (View Service Overview) to check the current storage and resource usage of the member account.

Cloud Storage

Before starting cloud development, you always need to have your data ready or upload your model training scripts. First, we’ll introduce how to use Cloud Object Storage on TWCC.

Using Cloud Storage via Browser

Step 1: Go to the 計畫資訊 (Project Information) page and scroll down to 所有服務 (All Services), then click on 雲端物件儲存 (Cloud Object Storage).

Step 2: Click on 建立 (Create).

Step 3: Enter a name for your storage. You can also use the default name.

Step 4: Click 確定 (Confirm).

At this point, you can view the storage bucket you just created.

Step 5: Click on the newly created storage bucket.

Since the bucket is brand new, it will be completely empty. You can now click 上傳 (Upload) or 開始上傳 (Start Upload) to upload the training data or scripts to TWCC’s cloud. These files will be available for use with the container computing service.

If you want to download or delete files from your storage bucket, simply select the desired file, click the   $\vdots$   icon on the right, and choose 下載 (Download) or 刪除 (Delete).

To delete multiple files at once, check the boxes in front of the files, then click 下載 (Download) or 刪除 (Delete) next to the 上傳 (Upload) button.

Likewise, to delete a storage bucket, select the one you want to remove, click the   $\vdots$   icon on the right, and choose 刪除 (Delete).

Using Third-Party Software to Manage Cloud Storage

You can also manage your cloud storage using third-party software. In this example, we’ll use S3 Browser, which is recommended by TWCC.
S3 Browser is a free Windows client for Amazon S3 and Amazon CloudFront. It provides a simple web service interface for easily browsing, retrieving, and storing data.

Step 6: Click on 連線資訊 (Connection Info).

You can choose to connect to either a 公共空間 (Public Space) or a 私有空間 (Private Space). Once selected:

Step 7: Click the icon on the right to view and copy the 服務端點 (Service Endpoint), Access Key, and Secret Key.

Step 8: Click on 下載 S3 Browser (Download S3 Browser).

Step 9: Click on Download S3 Browser Freeware XX.X.X to begin the download.

Step 10: Follow the installation instructions to complete the setup of S3 Browser.

Once installation is complete, launch S3 Browser. You should see a screen like this:

Next, you’ll need to add a new remote connection account.

Step 11: Enter a name for the account. You can customize this name.

Step 12: Select the account type. Choose S3 Compatible Storage.

Step 13: Enter (or paste) the 服務端點 (Service Endpoint) you noted earlier, usually cos.twcc.ai.

Step 14: Enter (or paste) your Access Key.

Step 15: Enter (or paste) your Secret Key.

Step 16: Click Add new account to complete the setup.

You’ll now be taken to the S3 Browser main interface. Any operations you perform here will be reflected in real time on the remote cloud storage.

You can create a new storage bucket from here by clicking on Buckets and then selecting Create New Bucket...

You can also manage the files within a bucket. Click on the desired bucket to open it. From there, you can drag and drop files into the window to upload them. You can also use the toolbar at the bottom to upload, download, delete, and perform other operations.

Just like the web version, you can perform actions on individual files. Right-click on the file you want to manage to access options like Download, Delete, and more.

Remember that all actions are synced in real time to the remote cloud storage.

Likewise, you can delete a bucket or perform other actions by right-clicking the desired bucket and selecting Delete Bucket or another option from the menu.

Interactive Container

After setting up cloud storage, the next step is to begin Interactive To do this, we need to create a container, which will later use TWCC’s computing resources. Here, we’ll be using a Interactive Container.

Step 1: Click 服務列表 (Service List).

Step 2: Select 開發型容器 (Interactive Container).

Step 3: Click 建立 (Create).

Step 4: Choose the image you wish to use.

After selecting an image, the next step is to fill in the basic information for the container.

Step 5: Enter a name for the container.

Step 6: Select the image version according to your computational needs.

Step 7: Choose the computing service you wish to use.

Step 8: Click 檢閱 + 建立 (Review + Create).

Step 9: After confirming all the details are correct, click 建立 (Create).

That’s all, your container is now created. Once its status shows Ready, the setup is complete. You can click 重新整理 (Refresh) multiple times to check the container’s status.

After the interactive container is created, billing starts immediately, charged by the second from your project’s wallet. You can view detailed information about the container.

Step 10: Double-click on the container you want to inspect.

From this interface, you can view both the basic information and connection details of the container.

Next, let’s go over how to delete a container.

Step 11: Go to the 開發型容器管理 (Interactive Container Management) interface, locate the container you wish to delete, and click the   $\vdots$   icon.

Step 12: Click 刪除 (Delete).

Step 13: Click 確認 (Confirm).

Step 14: Click 確認 (Confirm) again.

And that’s it, the container has been deleted. You can click 重新整理 (Refresh) a few times to confirm the status update.

Development Environment

We can begin container-based development, TWCC provides an online Jupyter notebook for use.

Step 1: Go to the Interactive Container details page, find Jupyter, and click 開啟 (Open).

Once the Jupyter notebook is launched, you can begin using the container.

Connect to Cloud Storage

To use COS in the container, we need to open the terminal first.

Step 2: Click File.

Step 3: Click New.

Step 4: Click Terminal to open a terminal.

You can now operate via the Jupyter terminal.

Step 5: Optionally, run

1

pwd

to check the current terminal path.

Step 6: Enter

1

pip install s3cmd

and press Enter to install the s3cmd tool. s3cmd is a command-line tool used to interact with Amazon S3 (Simple Storage Service).

Step 7: Enter

1

s3cmd --configure

and press Enter to configure s3cmd.

Step 8: Enter the Access Key remembered from Cloud Storage, then press Enter.

Step 9: Enter the Secret Key remembered from Cloud Storage, then press Enter.

Step 10: Set the Default Region, just press Enter.

Step 11: Enter the 服務端點 (Endpoint) remembered from Cloud Storage, typically cos.twcc.ai, then press Enter.

Step 12: Set the DNS-style, enter the same 服務端點 (Endpoint) again, typically cos.twcc.ai, then press Enter.

Step 13: Set the Encryption password. If you don’t need to set one, just press Enter.

Step 14: Set the Path to GPG program. This specifies the location of the GPG (GNU Privacy Guard) program. If you don’t use encryption, just press Enter.

Step 15: Set Use HTTPS protocol. This determines whether to use the HTTPS protocol to communicate with S3. Enter Y and press Enter.

Step 16: Set the HTTP Proxy server name. This specifies the hostname or IP address of the HTTP proxy server. If not needed, just press Enter.

Step 17: For Test access with supplied credentials, type n and press Enter.

Step 18: To save the configuration, type Y and press Enter.

Step 19: Remember the configuration file path for s3cmd. As shown in the image, it is typically /home/your-host-account/.s3cfg.

Now you can access COS contents in the current container. Use the following command

1

s3cmd ls s3://your-bucket-name/

to list all files in the bucket. Or use

1

s3cmd ls s3://your-bucket-name/folder-name/

to list all files under a specific folder in the bucket.

Here are the commands you can use to interact with COS.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31

# `~/folder-name/` is the folder under `/home/`

# --recursive      | Recursively process subdirectories, useful for folder upload or download.
# --skip-existing  | Skip existing files to avoid redundant uploads or downloads.
# --verbose        | Show detailed output for each file's status.
# --encoding=utf-8 | Use this when filenames contain Chinese characters to prevent garbled text.

# Use the specified configuration file to run s3cmd
s3cmd -c /home/your-host-account/.s3cfg

# List the files in the storage bucket
s3cmd ls s3://your-bucket-name/

# Download all files from the bucket to a folder under `/home/`
s3cmd get --recursive --verbose s3://your-bucket-name/ ~/folder-name/

# Upload all files from a folder under `/home/` to the bucket
s3cmd put --recursive --verbose ~/folder-name/ s3://your-bucket-name/

# Sync folders
s3cmd sync ~/folder-name/ s3://your-bucket-name/folder-name/

# To upload or download a single file
# Download a file from the storage bucket to a folder under `/home/`
s3cmd get s3://your-bucket-name/file-name ~/folder-name/

# Upload a file from a folder under `/home/` to the storage bucket
s3cmd put ~/folder-name/file-name s3://your-bucket-name/

# You can also use wildcards `*`, e.g.,
s3cmd get s3://your-bucket-name/*.csv ~/folder-name/

Remote Connection

Using Windows PowerShell

We can also use SSH to connect to the container for development. Here’s how:

Step 1: Go to the Interactive container details, find ssh, and click the copy icon next to it to copy the command.

You’ll notice that the SSH connection command includes the host account you registered earlier.

Step 2: Open PowerShell, paste the copied connection command, and press Enter.

The first time you connect, a fingerprint will need to be established, and you won’t need to do this again after that.

Step 3: Type yes and press Enter.

Next time you connect via SSH, you can directly access the remote terminal.

Step 4: Open PowerShell, paste the copied connection command, and press Enter.

Step 5: Enter the host password you set when registering your account, and press Enter.

Step 6: Once the environment is initialized, and the TWCC Ubuntu system command prompt $ appears, you can start using the TWCC container resources.

Using VSCode

Using SSH to connect is not the most convenient editing environment. We can use VSCode as a code editor and take advantage of its extensions to facilitate development.

Before using VSCode for connection, as mentioned in Using Windows PowerShell, we need to first establish the fingerprint.

Next, you can start working in VSCode.

Step 1: Click the Extensions page on the activity bar.

Step 2: Enter Remote in the search bar and press Enter.

Step 3: Install the Remote Explorer extension.

Step 4: Install the Remote - SSH extension.

After installing the above extensions, you should see the Remote Explorer extension in the activity bar. If not, please follow the steps below.

Step 5: Right-click on the activity bar, open the menu, and click Remote Explorer to display it in the activity bar.

Step 6: Click the Remote Explorer page.

Step 7: Click the + next to SSH.

Step 8: Go to the detailed page of the development container, find ssh, and click the copy icon next to it to copy the connection command. Paste the copied connection command into VSCode and press Enter.

The result will appear as shown below.

Step 9: The first time you connect, you will be prompted to select the SSH configuration file. Select the first one.

Step 10: Click the newly created connection configuration.

Step 11: Click the $\mathbf{\rightarrow}$ next to the connection profile you wish to connect to.

Step 12: Select the type of remote host. Since TWCC’s host type is a Linux-based Ubuntu, select Linux.

Step 13: Enter the host password you provided during registration, and press Enter.

You can also click (details) at the bottom right to enter the password.

Clicking (details) at the bottom right allows you to enter the password in the terminal.

After entering the password, wait for a moment. Once initialization is complete and you see end, the connection is successful. You can now use VSCode to operate the remote container.

After completing the tasks, you can check the Remote Explorer page to verify the connection status or check the status at the bottom left corner.

At this point, we may find that there is no terminal available for input.

Step 14: Click + to add a terminal for operating the remote container.

At this point, you can operate the container in the terminal just like in PowerShell.

However, we will find that this does not fully leverage the advantages of VSCode.

Next, we will demonstrate how to link the remote container directory.

Step 15: Click the Explorer tab.

Step 16: Click Open Folder.

Step 17: Click OK.

Step 18: Enter the host password again and press Enter.

Step 19: Check Trust all files in the parent folder ‘home’ from the author.

Step 20: Click Yes, I trust the author.

At this point, we can read and edit files in the container.

Finally, we can exit the connection using the following steps:

Step 21: Click SSH connection info in the bottom-left corner.

Step 22: Click Close Remote Connection.

To reconnect, click $\rightarrow$.

Create an Image File

After using a container, it should be deleted to save costs. However, reconfiguring the environment every time can be tedious and prone to errors. In this case, we can choose to create an image file of the container, which allows us to continue from the last configured environment the next time we use it.

Step 1: Go to the details page of the container for which you want to create an image file, and click 映像檔 (Image).

Step 2: Enter a 標籤 (label) for the image file, to help distinguish it from other image files.

Step 3: Click 確定 (Confirm).

Step 4: Click 映像檔申請紀錄 (Image Application Records) to check the status of the newly created image file. After applying for the image file, you need to wait about 3 to 5 minutes before it appears in 映像檔 (Images).

Step 5: Click 映像檔 (Images). After the application is approved, the newly created image file will be visible here, indicating it is ready for use.

Step 6: To delete an image file, find the image file you want to delete and click the   $\vdots$   button on the right.

Step 7: Click 刪除 (Delete).

Step 8: Click 確定 (Confirm) to delete the image file.

Using Image Files

Restoring the environment configured last time with an image file is also very simple.

Step 1: On the Interactive Container page, click 建立 (Create).

Step 2: Enter custom.

Step 3: Select Custom Image.

Step 4: Click the 映像檔 (Image File) dropdown menu.

Step 5: Select the image file.

The following steps are the same as those described in Interactive Container.

Conclusion

The Taiwan Computing Cloud (TWCC) provided by the National Center for High-performance Computing (NCHC) plays a pivotal role as an infrastructure backbone in academic research, industrial applications, and governmental policy implementation. As the use cases for artificial intelligence continue to expand, from natural language processing to medical image analysis, from smart manufacturing to urban governance, the demand for a flexible, efficient, and locally deployed computing environment grows stronger. TWCC offers a technically mature and policy-reliable cloud platform, serving as a vital bridge between innovative ideas and real-world applications.

Therefore, mastering the use of TWCC’s resources can accelerate the realization of innovative applications, enhance our development and competitiveness in the AI field, and prepare us to meet the challenges of the coming AI-driven era.

See Also

• [NCHC Exploration] #1 Account Creation

References

• HPC 高速運算任務. (2024). 台灣智慧雲端服務股份有限公司. Retrieved April 11, 2025, from https://docs.twcc.ai/docs/user-guides/twcc/hpc-job/

• InfiniBand. (November 6, 2024). Wikipedia, the free encyclopedia. Retrieved April 11, 2025, from https://en.wikipedia.org/wiki/InfiniBand

• NTHU SCC Team. (July 23, 2024). 2024 HPCAI Summer Camp 國網機器介紹(音訊增強) [Video]. YouTube. Retrieved April 15, 2025, from https://youtu.be/trNHSnGq36c

• S3 Browser - Free Windows Client for Amazon S3 and Amazon CloudFrontS3 Browser Freeware. (n.d.). S3 Browser. Retrieved April 14, 2025, from https://s3browser.com/

• TWCC 定價. (n.d.). HackMD. Retrieved April 11, 2025, from https://man.twcc.ai/@twccdocs/SJWlN3YDr

• TWCC FAQs | 容器運算服務 (CCS). HackMD. Retrieved April 12, 2025, from https://man.twcc.ai/@twccdocs/faq-zh/https%3A%2F%2Fman.twcc.ai%2F%40twccdocs%2Ffaq-ccs-zh

• 台灣杉二號. (August 26, 2024). Wikipedia, the free encyclopedia. Retrieved April 11, 2025, from https://zh.wikipedia.org/wiki/台灣杉二號

• 服務概觀. (n.d.). HackMD. Retrieved April 11, 2025, from https://man.twcc.ai/@twccdocs/twnia2-overview-zh

• 超級電腦. (2025). National Center for High-performance Computing. Retrieved April 11, 2025, from https://www.nchc.org.tw/Page?itemid=2&mid=4

• 開發型容器映像檔. (n.d.). HackMD. Retrieved April 18, 2025, from https://man.twcc.ai/@twccdocs/guide-ccs-image-zh

• 臺灣 AI 雲. (2025). Taiwan AI Cloud. Retrieved April 11, 2025, from https://www.twcc.ai/