Wu Yi

We decide to rename Couler to `flow`. SQLFlow compiles a SQL program into a workflow program, this workflow program should be able to run on different runtime environments like running...

1. 快速开始 / Quick Start 1. 云端体验 / Cloud Playground 1. 本地快速安装 / Local Try Out 1. 部署 / Deployment 1. 在Kubernetes集群上部署 / Deploy on Kubernetes Cluster 1. 高可用部署 /...

> ii 这里是指VM里安装吗？ 是的。 > 感觉 ii 和 iii 目前没有什么内容好写？ 有的。 > Self-defined -> Customized Runnable ? 同意。 > 6,7,8 是否可以放到一个大目录”开发者指南/Developer Guide“下？ 不行。开发者指南，顾名思义是开发SQLFlow的指南，很难直接联想到是贡献模型，需要分开。

> 最佳性能是指什么性能呢？SQLFlow server 还是作业的性能呢？ 包括SQLFlow 部署需要的怎么和database使用可以达到最佳性能，比如和Hive, MySQL 不能跨机房部署，Kubernetes集群必须有至少几个节点，pod之间网络通信带宽，如果是GPU集群，GPU集群之间的通信带宽等。

We may have below runtime entries to submit training/predicting jobs: ```python runtime.pai.train(datasource, ....) runtime.local.train(datasource, ...) runtime.elasticdl.train(datasource, ...) ``` For the PAI runtime API, when we call `runtime.pai.train()`, we should generate...

https://github.com/sql-machine-learning/sqlflow/blob/484753fc49e3974cc2fe2a18dd6cc31c2e3cfc51/go/parser/lexer.go#L185-L202 The `IDENT` can represent characters like `[/_:.]` currently I think.

I see, sounds good. By the way, we may need to update `TO TRAIN` clause also.

Some details about this issue? Why using tuples can confuse the data transform process?

We have https://github.com/sql-machine-learning/sqlflow/blob/develop/doc/model_parameter.md yet the supported model list document is still needed.

Actually you can define any training/predicting procedure with a custom model definition, please refer to https://github.com/sql-machine-learning/sqlflow/blob/develop/doc/contribute_models.md#write-models-other-than-keras