@@ -18,20 +18,20 @@ While the container engine sets these automatically when using the NCCL hook, th
1818
1919``` bash
2020export NCCL_NET=" AWS Libfabric" # (1)!
21- export NCCL_CROSS_NIC=" 1 "
22- export NCCL_NET_FORCE_FLUSH=" 1 "
21+ export NCCL_CROSS_NIC=1
22+ export NCCL_NET_FORCE_FLUSH=1
2323export NCCL_NET_GDR_LEVEL=PHB # (2)!
24+ export NCCL_SOCKET_IFNAME=hsn
2425export FI_CXI_DEFAULT_CQ_SIZE=131072 # (3)!
2526export FI_CXI_DEFAULT_TX_SIZE=32768
2627export FI_CXI_DISABLE_HOST_REGISTER=1
27- export FI_CXI_RDZV_EAGER_SIZE=" 0"
28- export FI_CXI_RDZV_GET_MIN=" 0"
29- export FI_CXI_RDZV_THRESHOLD=" 0"
30- export FI_CXI_RX_MATCH_MODE=" hardware"
31- export FI_CXI_RX_MATCH_MODE=software
32- export FI_CXI_SAFE_DEVMEM_COPY_THRESHOLD=" 16777216"
33- export FI_MR_CACHE_MAX_COUNT=" 524288"
34- export FI_MR_CACHE_MAX_SIZE=" -1"
28+ export FI_CXI_RDZV_EAGER_SIZE=0
29+ export FI_CXI_RDZV_GET_MIN=0
30+ export FI_CXI_RDZV_THRESHOLD=0
31+ export FI_CXI_RX_MATCH_MODE=hardware
32+ export FI_CXI_SAFE_DEVMEM_COPY_THRESHOLD=16777216
33+ export FI_MR_CACHE_MAX_COUNT=524288
34+ export FI_MR_CACHE_MAX_SIZE=-1
3535export FI_MR_CACHE_MONITOR=userfaultfd
3636export MPICH_GPU_SUPPORT_ENABLED=0 # (4)!
3737```
0 commit comments