He Jia

@jq Any idea for this?

May you use WeChat? Because there too less people in the discord/email-list, so maybe WeChat Group is much more convenience for me. Here is my WeChat ID: mofheka, add me...

I would like to transmit the sparse tensors, and there is no way for two workers to know the size of the data in advance. I don't know if I...

@yosefe Which callback? **ucp_am_recv_data_nbx_callback_t** from **ucp_request_param_t** or **ucp_am_recv_callback_t** from **ucp_am_handler_param_t**. I'm a little confused about what these two callbacks are used for respectively.

@yosefe Thank you for replying. I read the documentation and found that there is a **ucp_am_callback_t** that seems similar to the **ucp_am_recv_data_nbx_callback_t**. Did **ucp_am_callback_t** has being deprecated? Or it will...

Sorry for another question. What is purpose of the memh in ucp_request_param_t. It says is a pre-registered buffer. But what if I have already a pointer to ucp_am_recv_data_nbx, what does...

> > Sorry for another question. What is purpose of the memh in ucp_request_param_t. It says is a pre-registered buffer. But what if I have already a pointer to ucp_am_recv_data_nbx,...

@jakevdp This is tensorflow internal dynamic_partition source code using XLA , can see what can be referred to? https://github.com/tensorflow/tensorflow/blob/132c98b29c1a42298cd13055d1168c8d67177bc9/tensorflow/compiler/tf2xla/kernels/dynamic_partition_op.cc

I thought that what could be implemented with XLA must be implemented with JAX, is it because JAX does not support dynamic tensor? So how does tensorflow use dynamic_shape in...

@superbobry Sorry my mistake. A simpler description might be, when I use pmap, how do I implement the all2allv operator in transform code? Among them, the data parallel dimension (batch)...