Wu Yi

@m3ngyang Can you please take a look at this?

This is a bug when parsing optimization configs. Is there core files generated? Can you find the full call stack using `gdb` and the core file?

One way is to download the core file and then use the core file locally in a docker container.

Stack trace looks like in the core file: ``` #0 0x00007f6f556427fb in runtime.sched_getaffinity () at /usr/local/go/src/runtime/sys_linux_amd64.s:519 #1 0x00007f6f55a0bc82 in encoding/gob.(*Encoder).encodeArray (enc=0x7f6f55a0c518 , b=0xc4202207e0, value=..., op= {void (struct encoding/gob.encInstr *, struct...

可以详细描述下你的需求。 如果是文本类的任务，可以参考 http://paddlepaddle.org/docs/develop/documentation/en/api/v2/config/layer.html#context-projection 做sequence的1d conv。

@Yancey1989 Why we need `https://gonum.org/v1/plot `? Is it for drawing the edl experiment graph?

Yes, it's possible, but requires a lot of online cluster data.

报错信息可以上下部分截取全一些么，最好直接用粘贴code block。

看起来`path`参数被python端改错了

Thought to put logs on CephFS so it will be persisted. Then add commands like `tail` or `head` to let users to check out the contents of the log files.