Please enable JavaScript.
Coggle requires JavaScript to display documents.
DL Open-Source Framework, Server, Nvidia series - Coggle Diagram
DL Open-Source Framework
TensorFlow
-
Static computation graph
The code is complex, and the static graph needs to be built multiple times
Caffe
-
R & D costs are relatively high, not suitable for rapid model development and iteration
-
-
-
Server
NVIDIA DGX
The overall solution is more mature, and technical support will be better.
-
-
DELL
Mature solutions, stable quality
-
NVIDIA GPU + Intel CPU
Large user base, mature community, easily find solutions to problems
Nvidia series
Support the deployment of TensorFlow, PyTorch based DL frameworks
TensorRT, tool optimised for DL
CUDA well encapsulates some API interfaces for graphics card hardware resource scheduling and use, which allows R&D personnel to use the computing resources of the graphics card conveniently and efficiently
CUDA is the only one option of the industry; Many open source frameworks are built on CUDA.