QI JUN

@xushiwei Good suggestions. Actually, there are two built-in functions in Python2, `range` and `xrange`. `range` returns a sequence of numbers, while `xrange` returns an iterable object. In Python3, there is...

@terrytangyuan I am trying to optimize the speed of odps data reader now. I will take this into consideration in next step refactoring work.

Please refer to #1252 , I think we might have two kinds of evaluation.

@brightcoder01 这个没有。Transform是读数据的时候 on-the-fly 去做的。

@linkerzhang As @Kelang-Tian writes, the `feed()` function transforms a training sample in a `RecordIO` format to a `tf.Tensor` or `torch.Tensor`. Users could also add their preprocessing logic. For example, adding...

Please refer to #1253 . Maybe we could just let it failed. Task status is also no need to keep.

@skydoorkai It seems that the problem mainly focuses on embedding table. - initializer, Should we initialize parameters lazily in pserver. Currently, we first calls `lookup` --> not found, then `initialize`in...

@backyes Thank you for your interest! ElasticDL supports large embedding tables and also supports sparse SGD updating. An embedding table will be sharded to several PS instances. In forward pass,...

vote for logo 2. 主要是考虑elastic这个单词，希望logo能够体现弹性寓意，目前看起来logo 2的曲线更接近一些。

Yes, we could remove the parameter server implementation with Python. The parameter server implementation with Go has been used in production already. It's safe to remove Python codes.