Zhang Haitao
Zhang Haitao
We can build python codes on the host computer with `setup.py` without any problem. The only issue we need to solve is PS Go codes in elasticdl building. The host...
For prediction, there can be two kinds of prediction calls. - Take a model input or a batch of inputs, returns the model outputs. ``` elasticdl.predict(trained_model, input_data) ``` This is...
Can you give out more detailed repro steps? How to `sequentially submit multiple training commands`?
Can you still repro this issue with the current ElasticDL version?
`records = [record for r in reader.read_records(task)]` 这样就需要把 task 里面的所有数据都读进来，数据读取 IO 没法和计算并行。 可以我们设计一个 Dataset, 里面有 buffer 在异步的读取数据， `__getitem__` 就从 buffer 拿数据，有就直接返回，数据还没有就等待。 这样 IO 和计算能并行。
https://pytorch.org/docs/stable/data.html#torch.utils.data.IterableDataset IterableDataset 看着就和 tf generator 一样，只需要提供一个 iter() 接口？
* initializer, TensorFlow automatically initializes the variables if Keras functional API is used. For Keras subclass model, the worker sends the variable initial values to PS (`report_variable`). * regularizer, this...
> @skydoorkai > > It seems that the problem mainly focuses on embedding table. > > * initializer, Should we initialize parameters lazily in pserver. Currently, we first calls `lookup`...
Vote for logo 2. The curve represents the elastic resource usage in a cluster, like a curve in our benchmark report.
Voting result - Logo 2 : brightcoder01, QiJune, skydoorkai - Logo 4: wangkuiyi , workingloong