-
When a connection from the spdk initiator (perf) to spdk target (nvmf_tgt) is established and data transfer is occurring, powering off the initiator by pulling the power prevents the initiator from …
-
Je potřeba se napojit na data z Hlídače Státu. V současném kódu je ve `startup.cs` přidaná InMemory databáze, která má nějaká základní data pro test. https://github.com/cesko-digital/nasi-politici/blo…
-
https://drive.google.com/open?id=14eph67uGIQJgEBSmvWAufdRY49sI3JFq9G9_lH8G7h4
-
```
OMPI: 4.0.2rc1
MOFED: MLNX_OFED_LINUX-4.6-1.0.1.1
Module: hpcx-gcc (2019-09-07)
Test module: none
Nodes: orion x24 (ppn=28(x24), nodelist=clx-orion-[031-032,049-054,056-064,075-081])
```
…
-
I tried to run the benchmark on the 100G Ethernet via RoCE, but it failed with this err.
```
wc->status == IBV_WC_SUCCESS. 5 vs 0. Send for slot 0: Work Request Flushed Error
```
I wonder if it …
-
Hi nccl team:
Today I run horovod on two nodes with each nodes has 4 V100 GPU, and I find nccl maybe hang all time time, output is:
```
[1,4]:2019-05-15 06:47:06.677421: I tensorflow/core/platf…
-
### Describe the bug
ucx_info -d
shows various errors (depending on the ucx version) on nodes with connectx-6 hca. I tried several versions of ucx but didn't succeed using it with connectx-6. Befor…
-
Thanks very much for your quick reply on my last issue!
I have been continually trying to build an 1P-1M kernel but it panics saying: "not syncing - no RoCE".
I am quite sure that I am using th…
-
Really good work. Just would like to know if the packets traveling between CPU and FPGA has the metadata in the front? If so, how does the pkt go through ASIC nic between the two?
-
When I trained the bert model with horovod and XLA, although XLA can significantly improve the perf, it decreased the scalability significantly.
Our cluster has 16 nodes, 4 V100 32GB/node. The nod…