EHPC Deep Learning (DL) Cloud
Integrated with deep learning algorithms, it provides GPU and CPU heterogeneous cluster resources, and supports image/video detection, text/picture classification, semantic speech recognition, intelligent recommendation, and other scenarios. It delivers the complete lifecycle of deep learning, with efficient training, sound recommendation, and open algorithms
Contact Us
Benefits
Heterogeneous Resources
Each node with CPU and GPU heterogeneous mode and can be configured with multiple NVIDIA A100 40GB GPU cards with 8 NVLINK interfaces.
Powerful Training
Multiple GPU instances available for computing resources required for model training and data analysis
Excellent Reasoning
NVIDIA A100 delivers industry-leading reasoning capabilities with a full range of precision acceleration, enabling powerful and diverse application.
Architecture
1
Features
Mainstream frameworks such as Tensorflow, PyTorch, Keras, Caffe and MXNet, etc., to quickly install custom applications with a variety of application environments.
Enables heterogeneous CPU/GPU resources to accelerate model training and reduce lead times.
Benefits
Reduce Reasoning Cost: GPU A100 instance could be customized to reduce reasoning cost.
On-demand Provisioning: Decouples CPU from GPU, with ready-to-use and flexible CPU/GPU ratio resources.
Dynamic Scaling: GPU resources for inference acceleration easily scaled to optimize the cost per actual workload.
Use Cases
Speech Recognition Synthesis
Remote Image Classification
Features
Based on the deep neural network DNN model, a architecture with GPU is adopted to accelerate training for more efficient speech recognition
Based on the GPU A100 accelerator, GPU computing resources greatly improve the training efficiency of the speech recognition model
Benefits
Based on the GPUA100 accelerator, the heterogeneous GPU and CPU architecture improves the training efficiency of supervised and unsupervised remote image classification model.
With the shared storage on multiple nodes, it covers parallel processing of shared storage data by multiple nodes and facilitates parallel data processing by region, classification type.
Based on the GPU cluster, it greatly improves the time, accuracy and efficiency of remote sensing images processing.
For more information about the use cases and technical architecture, please contact our customer service team.
Contact Us
© 2009-2022 PetaExpress All Rights Reserved