search

kubeedge / sedna

AI tookit over KubeEdge
https://sedna.readthedocs.io
Apache License 2.0
505 stars 163 forks source link

Life long learning SDK documentation missing #203

Open JensenJian opened 3 years ago

JensenJian commented 3 years ago

What would you like to be documented: Sedna python SDK documentation regarding life long learning-feature example

Why is this needed:

  1. Due to the life-long learning feature. It needs to construct a knowledge base, and I am not sure how to do that in an SDK situation.
  2. For now, the Sedna Python SDK documentation only mentions how to construct an estimator and initialize an incremental learning job.
JoeyHwong-gk commented 3 years ago

As so far, knowledgebase is a closed component of the lifelong learning feature, uses a relational database as a backend to store knowledge (model、sample、task 、relation etc.) . Sure that sedna community will open up the related interfaces, even though not open now, follow here for more details.

By the way, whath interfaces will you want the SDK of the knowledge base to meet?

JoeyHwong-gk commented 3 years ago

/assign @JimmyYang20

JensenJian commented 3 years ago

As so far, knowledgebase is a closed component of the lifelong learning feature, uses a relational database as a backend to store knowledge (model、sample、task 、relation etc.) . Sure that sedna community will open up the related interfaces, even though not open now, follow here for more details.

By the way, whath interfaces will you want the SDK of the knowledge base to meet?

Hi Joey, It is not about the interface in this issue, what I mean is I need a documentation to tell me what specific environment variables is needed (such as MIN_TRAIN_SAMPLE, MODEL_URLS, KB_SERVER etc.), plus it require user run the run_kb script. By haveing those information it supports me use the life long learning feature in SDK case better and more handy.

JoeyHwong-gk commented 3 years ago

As so far, knowledgebase is a closed component of the lifelong learning feature, uses a relational database as a backend to store knowledge (model、sample、task 、relation etc.) . Sure that sedna community will open up the related interfaces, even though not open now, follow here for more details. By the way, whath interfaces will you want the SDK of the knowledge base to meet?

Hi Joey, It is not about the interface in this issue, what I mean is I need a documentation to tell me what specific environment variables is needed (such as MIN_TRAIN_SAMPLE, MODEL_URLS, KB_SERVER etc.), plus it require user run the run_kb script. By haveing those information it supports me use the life long learning feature in SDK case better and more handy.

Sorry, what you described is not related to the interface of knowledge base. Maybe your intention is to use lib independently of CRD that I'm not sure if it goes against the original intent of the community.

JoeyHwong-gk commented 3 years ago

/unassign @JimmyYang20 /assign @jaypume