Interpretations of Terms
TensorFlow
TensorFlow is an open source software library that uses data flow diagrams for numerical computation. It is widely used in the programming of various machine learning algorithms. It is developed and maintained by Google Brain, a Google artificial intelligence team.
PaddlePaddle
PaddlePaddle is an open source deep learning framework launched by Baidu, which supports advanced algorithms such as machine vision, natural language processing and recommendation system, etc. Paddle (Parallel Distributed Deep Learning) has the characteristics of ease of use, flexibility, high efficiency and expandability. See Official Website for details.
BML currently supports Paddle Fluid v1.5 deep learning framework
Baidu Object storage(BOS)
BOS (Baidu Object storage) provides stable, secure, efficient and highly extended storage services, and supports any type of data storage, such as text, multimedia, binary and so on, with a single file of up to 5TB. BML accesses data through BOS, saves training results and logs to the designated BOS address, so when opening BML service, authorization is required to access user’s BOS address.
Notebook
Notebook provides Jupyter, a visual code running environment with built-in algorithms such as TensorFlow, PyTorch, Keras, Caffe, Mxnet, Chainer and PaddlePaddle, for data processing and modeling.
Job Modeling
Provide a high-performance computing environment for large-scale distributed model training and optimization. Including deep learning job, machine learning job, AutoDL job, AutoML job.
Prediction Service
Model prediction service feature module provided by BML platform. According to the demand of the model application, reasonably configure and schedule the service resources, build and deploy the highly available online prediction cluster services.
Prediction Model
Prediction model: it is the information set of model data, deployment image and configuration logic required for deployment of prediction service.
Online Prediction
According to the online API generated by the user model, users can send requests to the API for data prediction, or they can establish multi-version services for small traffic experiments.
Container Image
The software environment required for service operation, including OS, basic library, ML framework, prediction service SDK and user-defined logic, etc.