NVIDIA nccl issues - Githubissues

NVIDIA / nccl

Optimized primitives for collective multi-GPU communication

Other

3.28k stars 831 forks source link

issues

Newest

Newest Most commented Recently updated Oldest Least commented Least recently updated

get rid of some unbounded sprintfs

#1330 madeleineth closed 3 months ago
0
Why can't two GPUs in a virtual machine communicate using P2P？

#1329 qianxiaoliang opened 5 months ago
1
NCCL error "receiving 524288 bytes instead of 65536"

#1328 JingyuQian closed 4 months ago
2
how double binary tree communicate

#1327 ltm920716 closed 4 months ago
4
Why dose theoretical busBw multiply by the ratio 5/6?

#1326 cyqmonkey opened 5 months ago
0
fix stack smash

#1325 madeleineth closed 4 months ago
1
How to locate the hanging node?

#1324 Puzzzle7 opened 5 months ago
1
nccl with specified pkey_index

#1323 wengyao04 closed 5 months ago
1
For channel computing, why nvlinkBw is accumulated, but pciBw is not? Is this a BUG?

#1322 GodHforever opened 5 months ago
2
Has NCCL support inter-node through NVswitch and NVlink?

#1321 shanleo2024 closed 5 months ago
8
Execute all_reduce_perf block

#1320 zhaotyer closed 5 months ago
1
How sendProxyProgress() in net.cc works

#1319 ZhiyiHu1999 opened 5 months ago
2
How are threads in different channels parallelized

#1318 ZhiyiHu1999 opened 5 months ago
0
NCCL2.21 hangs at cudaLaunchKernelExC()

#1317 leiyi666 opened 5 months ago
6
Performance Degradation in Alltoall Operation with NCCL 2.19 and 2.20

#1316 GeofferyGeng opened 5 months ago
5
Understanding LL, LL128, and Simple Protocols

#1315 arkhadem opened 5 months ago
0
Compute time in the reduction operation

#1314 tks2004 opened 5 months ago
0
Some questions about selecting NET when searching channels.

#1313 shanleo2024 closed 5 months ago
12
Local user buffer registration for NVLink SHARP

#1312 zhang662817 opened 6 months ago
1
Added retries to EHOSTUNREACH socket error.

#1311 newellz2 opened 6 months ago
0
Profiling Tools for NCCL collective operations

#1310 Chen-Chang opened 6 months ago
0
Dual 4090 bandwidth slower with PCIe

#1309 YZP17121579 closed 6 months ago
1
link against rdma-core libs when build with RDMA-CORE

#1308 changchengx opened 6 months ago
1
nccl-test can use nvidia sharp, but training job can not use nvidia sharp

#1307 Lzhang-hub closed 6 months ago
0
About NVLS MC/UC buffer

#1306 vvmex opened 6 months ago
0
why two GPU far than PXB under intel cpu use P2P will be slower(without NVLink)

#1305 HuangShiqing opened 6 months ago
2
NCCL fallback to Ring,LL on broadcast perf and NCCL_ALGO=Tree

#1304 arttianezhu opened 6 months ago
1
All Reduce Performance on H100 VMs

#1303 apoorvemohan opened 6 months ago
1
Why duplicate nChannels in connect.cc

#1302 jxh314 opened 6 months ago
1
Why there are two IDs for MNNVL support?

#1301 dearsxx0918 closed 6 months ago
2
How does collective operations call runRing, runTreeUpDown, and runTreeSplit

#1300 ZhiyiHu1999 opened 6 months ago
1
How is the logic for allocating data across different channels?

#1299 jxh314 closed 1 month ago
1
all-reduce slower on v2.20.5 compared to v2.18.5 on AWS g5.48xlarge (8 x A10G)

#1298 abdulfatir opened 6 months ago
15
Internal error when submitting a job to a Ray cluster

#1297 troelsfr opened 6 months ago
3
What's the relationship between nccl protcols and inter-node communication?

#1296 Alex-Wong opened 6 months ago
0
NCCL_NET_GDR_READ's performance impact on a PCIe platform

#1295 cold2stone opened 6 months ago
3
Does ncclBroadcast call return at same time on different ranks？

#1294 Eiji911 opened 6 months ago
1
How to understand "bank" in net.cc?

#1293 dearsxx0918 closed 6 months ago
0
Failed to find ncclNetPlugin_v8 symbol

#1292 wwj-2017-1117 opened 6 months ago
4
nccl-tests with two GH200 over Quantum2 iB stuck

#1291 itzsimpl opened 6 months ago
2
Inquiry about NCCL's Tree Algorithm Performance in Single and Dual Machine Scenarios

#1290 fizzlover opened 6 months ago
0
NCCL stuck when using nccl-test.

#1289 deepzzz123 opened 6 months ago
3
One of the NODE will hang when NCCL_NET_GDR_READ=1

#1288 shanleo2024 closed 4 months ago
1
How can this be ported to Windows?

#1287 eabase opened 6 months ago
4
How can I identify level1 nvswitch and level2 nvswitch in NCCL

#1286 Ryan201802 opened 6 months ago
12
AMD EPYC 7K62 NCCL-test 4090 bandwidth too

#1285 ghoul02015 opened 6 months ago
1
an NCCL timeout and the error "torch.cuda.OutOfMemoryError: CUDA out of memory. Tried to allocate more than 1EB memory"

#1284 jessiewy closed 6 months ago
5
RuntimeError: NCCL Error 1: unhandled cuda error (run with NCCL_DEBUG=INFO for details) when torch._C._broadcast_coalesced

#1283 zhoulei-biubiu opened 6 months ago
2
Why nccl ring all reduce stream duration doesn't scales with theoretical (N-1)/N?

#1282 CraneQinghe opened 6 months ago
1
Why is allgather's busbw a little worse than allreduce/reducescatter for the same nccl environment variables

#1281 pkuleo opened 6 months ago
1

Previous Next