feely
feely
> It could be resolved by upgrading to v12.19.0, if you must use Node12. Is this a bug from node?
> Another case for segfault: > https://github.com/alibaba/pipcook/runs/1340629270 This case looks like a memory access problem.
Where did you get this URL? I think we cleaned the OSS once. If we confirm that this dataset is useful, we will restore it to the new URL and...
> from https://juejin.cn/post/6968665578491871269 @ManfredHu PipCook 2.0 has been released, you can read [this](https://github.com/alibaba/pipcook/tree/main/example/pipelines) to train and predict. It's much faster than v1.x.
@ManfredHu This is an introduction to Pipcook 2.0 from **juejin**: https://juejin.cn/post/7003297331411943460
Can you please post the full error message?
Maybe we can use office CUDA docker image directly: https://hub.docker.com/r/nvidia/cuda
So the Pipcook image includes CUDA and Pipcook client? Looks not so necessary.
I don't think pipcook will access the 127.0.0.1:80, Is there any hosts setting in your server?
I traced the stack, the pointer [internals_ptr_ptr](https://github.com/pybind/pybind11/blob/master/include/pybind11/embed.h#L165) is invalid. And it's from [builtins[id]](https://github.com/pybind/pybind11/blob/master/include/pybind11/embed.h#L160), I think it is assigned by cpython. we need to confirm this.