Sold by: Hanwei Software Technology
Deployed on AWS
AWS Free Tier
This product has charges associated with it for Hanwei's integration of TensorFlow 2.21, PyTorch 2.5.1, CUDA 12.6, cuDNN 9.x, NVIDIA driver 595, Python 3.12, Jupyter, and scientific libraries on Ubuntu 24.04 LTS, plus GPU configuration and testing for NVIDIA T4 on g4dn.
Overview
This product provides TensorFlow 2.21 and PyTorch 2.5.1 integrated with CUDA 12.6, cuDNN 9.x, NVIDIA driver 595, Python 3.12, Jupyter Notebook, and scientific-computing libraries on Ubuntu 24.04 LTS. The Python frameworks and libraries are installed in an isolated virtual environment to provide a defined dependency boundary for the delivered stack.
Marketplace software charges cover Hanwei's integration of the GPU driver, CUDA, cuDNN, Python runtime, deep-learning frameworks, notebook environment, and scientific libraries; preparation of the isolated Python environment; GPU configuration for NVIDIA T4 hardware; image construction; and final AMI testing on a g4dn instance. Support is supplementary and is not the sole basis for the software charges.
What Hanwei Adds to the Upstream Software
- Integrated GPU software stack: TensorFlow 2.21 and PyTorch 2.5.1 are installed with CUDA 12.6, cuDNN 9.x, and NVIDIA driver 595 as a coordinated image configuration rather than as separate upstream packages.
- Isolated Python environment: The frameworks and scientific libraries are installed in an isolated Python 3.12 virtual environment to separate the delivered Python dependencies from the operating-system Python environment.
- Notebook integration: Jupyter Notebook is installed and configured to start as a managed service on port 8888. Jupyter authentication requires a unique, unpredictable secret created for the instance or a customer-configured first-launch credential. Customers must restrict network access and configure TLS or another protected access path before exposing the notebook service.
- Scientific-computing environment: NumPy, SciPy, scikit-learn, and Matplotlib are included for numerical computing, machine-learning utilities, and visualization workflows.
- T4 and g4dn configuration: The software stack is configured for NVIDIA T4 GPUs on g4dn instances. Final-image verification covers driver loading, CUDA visibility, framework GPU detection, representative tensor operations, Jupyter startup, and reboot behavior on the stated instance type.
- Build-date package baseline: Ubuntu and Python packages reflect the configured repositories and package indexes at image-build time. Customers remain responsible for reviewing compatibility and applying updates after launch.
Included Components
- Ubuntu 24.04 LTS
- Python 3.12 virtual environment
- TensorFlow 2.21 with GPU integration
- PyTorch 2.5.1 with GPU integration
- CUDA 12.6
- cuDNN 9.x
- NVIDIA driver 595
- Jupyter Notebook
- NumPy
- SciPy
- scikit-learn
- Matplotlib
Jupyter Access and Security
- Jupyter Notebook uses port 8888 in the delivered configuration.
- Before allowing inbound access to port 8888, customers must configure a unique authentication secret and restrict the security-group source range.
- TLS termination, identity-aware access, VPN access, SSH tunneling, or AWS Systems Manager port forwarding should be used according to the customer's security architecture.
- Port 8888 should not be exposed to 0.0.0.0/0 without an independently reviewed authentication and encryption layer.
- Notebook code can access instance data and IAM permissions; customers are responsible for least-privilege IAM roles, secrets, network controls, user access, and notebook content.
Highlights
- TensorFlow 2.21 and PyTorch 2.5.1 are integrated with CUDA 12.6, cuDNN 9.x, NVIDIA driver 595, and Python 3.12 on Ubuntu 24.04 LTS.
- Jupyter Notebook and NumPy, SciPy, scikit-learn, and Matplotlib are prepared in an isolated Python environment.
- The GPU stack is configured and tested for NVIDIA T4 on g4dn, with explicit boundaries for other GPUs, performance, and distributed workloads.
Details
Sold by
Delivery method
Delivery option
64-bit (x86) Amazon Machine Image (AMI)
Latest version
Operating system
Ubuntu 24.04
Deployed on AWS
New
Introducing multi-product solutions
You can now purchase comprehensive solutions tailored to use cases and industries.
Features and programs
Financing for AWS Marketplace purchases
AWS Marketplace now accepts line of credit payments through the PNC Vendor Finance program. This program is available to select AWS customers in the US, excluding NV, NC, ND, TN, & VT.
Pricing
Pricing is based on actual usage, with charges varying according to how much you consume. Subscriptions have no end date and may be canceled any time.
Additional AWS infrastructure costs may apply. Use the AWS Pricing Calculator to estimate your infrastructure costs.
If you are an AWS Free Tier customer with a free plan, you are eligible to subscribe to this offer. You can use free credits to cover the cost of eligible AWS infrastructure. See AWS Free Tier for more details. If you created an AWS account before July 15th, 2025, and qualify for the Legacy AWS Free Tier, Amazon EC2 charges for Micro instances are free for up to 750 hours per month. See Legacy AWS Free Tier for more details.
Dimension | Cost/hour |
|---|---|
g4dn.4xlarge Recommended | $0.25 |
t2.micro | $0.00 |
t3.micro | $0.00 |
g4ad.16xlarge | $0.12 |
t3.2xlarge | $0.18 |
p3dn.24xlarge | $0.20 |
g5.48xlarge | $0.12 |
g5.12xlarge | $0.12 |
g4dn.8xlarge | $0.25 |
t3.medium | $0.08 |
Dimensions summary
You pay by the hour for this hardened deep learning software, based on the AWS EC2 instance type you run it on. The price is charged only while the instance runs, added to your AWS compute cost. Options span general-purpose burstable instances (the t2 and t3 families) and GPU-accelerated instances (the g2, g3, g4, g5, and p3/p4 families) sized from micro to metal. Larger and GPU-heavy instances carry higher hourly rates. You choose the instance that fits your workload, and pricing scales with the size and hardware you select.
Top-of-mind questions for buyers
Each hour covers the hardened deep learning software running on one EC2 instance of the type you select. The rate reflects that instance's size and hardware. GPU families like g4, g5, and p3/p4 add graphics processors for training. The t2 and t3 families use burstable general-purpose CPUs.
The software fee meters running hours only. A stopped instance stops accruing the hourly software charge. You may still pay separate AWS fees for attached storage while the instance is stopped. Charges resume when you restart the instance.
You pick one instance type at launch, and its hourly rate applies for as long as that instance runs. To use a different size, you launch a new instance at that type's rate. Charges do not blend across types; each running instance bills at its own rate.
www.proimage.cloud
Helpful?
Vendor refund policy
no refund
Custom pricing options
Request a private offer to receive a custom quote.
Legal
Vendor terms and conditions
Upon subscribing to this product, you must acknowledge and agree to the terms and conditions outlined in the vendor's End User License Agreement (EULA) .
Content disclaimer
Vendors are responsible for their product descriptions and other product content. AWS does not warrant that vendors' product descriptions or other product content are accurate, complete, reliable, current, or error-free.
Delivery details
64-bit (x86) Amazon Machine Image (AMI)
Amazon Machine Image (AMI)
An AMI is a virtual image that provides the information required to launch an instance. Amazon EC2 (Elastic Compute Cloud) instances are virtual servers on which you can run your applications and workloads, offering varying combinations of CPU, memory, storage, and networking resources. You can launch as many instances from as many different AMIs as you need.
Version release notes
Version 2.0.0 - 2026-04-30
Major Changes
- Upgraded OS from Ubuntu 20.04 LTS (EOL) to Ubuntu 24.04 LTS (supported until 2034)
- Upgraded NVIDIA Driver from 525 to 595
- Upgraded CUDA Toolkit from 11.7 to 12.6
- Upgraded cuDNN from 8.9 to 9.x
- Upgraded TensorFlow from 2.11 to 2.21 (GPU enabled)
- Upgraded PyTorch from 1.13.1 to 2.5.1 (GPU enabled)
- Upgraded Python from 3.8 to 3.12
New Features
- Out-of-the-box Jupyter access: Default password is the EC2 Instance ID, no SSH setup required
- Python venv isolation: All frameworks installed in /home/ubuntu/dl-env, avoiding system package conflicts
- Version pinning: CUDA and NVIDIA driver locked to prevent unintended upgrades during apt upgrade
Security Fixes
- Resolved all known vulnerabilities from Ubuntu 20.04 EOL by migrating to Ubuntu 24.04 LTS
- IMDSv2 support for secure instance metadata access
Known Issues
- TensorFlow and PyTorch ship their own CUDA runtime libraries inside the venv. System CUDA 12.6 is used for nvcc compilation only.
- First page load of Jupyter may take a few seconds after instance boot.
Upgrade Notes
- This is a full OS migration, not an in-place upgrade.
- Existing models and code using TensorFlow or PyTorch should be tested for compatibility with the new framework versions.
- Users who pinned specific package versions should verify compatibility with Python 3.12.
Additional details
Usage instructions
Getting Started
This AMI is ready to use out of the box. No SSH setup required.
Step 1: Launch an EC2 Instance Launch an instance using this AMI. Make sure your security group allows inbound traffic on port 22 (SSH) and port 8888 (Jupyter Notebook).
Step 2: Access Jupyter Notebook Open your browser and go to: http://(your-instance-public-ip):8888
Default password: your EC2 Instance ID (example: i-0a1b2c3d4e5f6g7h8) You can find the Instance ID in the AWS Console under EC2 - Instances.
Step 3: Start Working You now have a fully configured deep learning environment with TensorFlow, PyTorch, and GPU acceleration. Create a new notebook and start coding.
Recommended Instance Types
g4dn.xlarge (Best Value) GPU: 1x NVIDIA T4, 16 GB GPU Memory vCPU: 4, RAM: 16 GB Best for: Development, inference, light training On-Demand: approximately USD 0.526/hr
g4dn.2xlarge GPU: 1x NVIDIA T4, 16 GB GPU Memory vCPU: 8, RAM: 32 GB Best for: Larger datasets, CPU-intensive preprocessing On-Demand: approximately USD 0.752/hr
g5.xlarge GPU: 1x NVIDIA A10G, 24 GB GPU Memory vCPU: 4, RAM: 16 GB Best for: Larger models, faster training On-Demand: approximately USD 1.006/hr
g5.2xlarge GPU: 1x NVIDIA A10G, 24 GB GPU Memory vCPU: 8, RAM: 32 GB Best for: Production inference, medium-scale training On-Demand: approximately USD 1.212/hr
p3.2xlarge GPU: 1x NVIDIA V100, 16 GB GPU Memory vCPU: 8, RAM: 61 GB Best for: Serious training workloads On-Demand: approximately USD 3.06/hr
Prices are approximate and may vary. Please check AWS pricing for the latest rates. Use Spot Instances to save up to 70%.
Cost Saving Tips
- Use Spot Instances for development and non-critical workloads
- Stop the instance when not in use, you only pay for EBS storage when stopped
- Start with g4dn.xlarge and scale up only if needed
SSH Access (Optional)
ssh -i your-key.pem ubuntu@(your-instance-public-ip) source ~/dl-env/bin/activate
Changing the Jupyter Password
source ~/dl-env/bin/activate jupyter notebook password sudo systemctl restart jupyter.service
Pre-installed Frameworks
TensorFlow 2.21 (GPU enabled) PyTorch 2.5.1 (GPU enabled) NumPy, SciPy, scikit-learn Matplotlib, Pillow, h5py Jupyter Notebook
Support
For framework-specific questions, refer to the official documentation: TensorFlow: https://www.tensorflow.org/ PyTorch: https://pytorch.org/
Resources
Vendor resources
Support
Vendor support
prosupport@hanweie.com If you encounter problems in the process of using the system, please feel free to contact us by email, thank you!
AWS infrastructure support
AWS Support is a one-on-one, fast-response support channel that is staffed 24x7x365 with experienced and technical support engineers. The service helps customers of all sizes and technical abilities to successfully utilize the products and features provided by Amazon Web Services.
Updated weekly
By Hanwei Software Technology
By Sigmodata
By Elm Computing
AI generated from product descriptions
Pre-installed Deep Learning Frameworks
TensorFlow 2.21 (GPU), PyTorch 2.5.1 (GPU), CUDA 12.6, cuDNN 9.x, and NVIDIA Driver 595
GPU Optimization
Optimized for NVIDIA T4 GPUs on g4dn instances with GPU-accelerated framework support
Development Environment
Jupyter Notebook with auto-start on port 8888, Python 3.12 isolated virtual environment for dependency management
Scientific Computing Libraries
Bundled with NumPy, SciPy, scikit-learn, and Matplotlib for data processing and visualization
Operating System and Support
Ubuntu 24.04 LTS with long-term security support until 2034
Deep Learning Framework Support
Includes latest TensorFlow and PyTorch versions with Python 3.12 runtime environment
GPU Acceleration
Automatic GPU support with NVIDIA CUDA 12.9 and cuDNN, optimized for instances such as g6.xlarge (NVIDIA L4), g4dn.xlarge (T4), g6.2xlarge, and g6e.xlarge (L40S)
Authentication and Access Control
Supports multiple authentication methods including local accounts, OIDC (Okta, Microsoft Entra ID, Google Workspace, Auth0), and LDAP/Active Directory integration
Secure Communication
Browser-based setup wizard over HTTPS with optional trusted TLS certificate from Let's Encrypt
Pre-installed Scientific Libraries
Includes Scikit Learn, Matplotlib, and Numpy as built-in dependencies for data processing and visualization
Multi-User Notebook Server
Pre-configured JupyterHub server capable of serving Jupyter notebooks to multiple concurrent users
Multiple Python Versions
Support for multiple Python versions installed and available within the environment
Pre-installed Data Science Libraries
Popular data science Python packages and SageMath pre-installed and configured
Extensive Python Package Ecosystem
Hundreds of Python packages available for data science, scientific computing, and development workflows
Standard contract
No
Customer reviews
No customer reviews yet
Be the first to review this product . We've partnered with PeerSpot to gather customer feedback. You can share your experience by writing or recording a review, or scheduling a call with a PeerSpot analyst.