Skip to content

Commit

Permalink
Fix an invalid reference in the doc (#1362)
Browse files Browse the repository at this point in the history
  • Loading branch information
wujingyue authored Dec 14, 2024
1 parent e7bfc0c commit 1ae8190
Showing 1 changed file with 2 additions and 2 deletions.
4 changes: 2 additions & 2 deletions examples/pytorch/comm_gemm_overlap/README.md
Original file line number Diff line number Diff line change
Expand Up @@ -16,7 +16,7 @@
Forward and backward passes with layer weights distributed over all GPUs in a single node.

```bash
$ torchrun --nnodes=1 --nproc-per-node=$(nvidia-smi -L | wc -l) ln_mlp_with_overlap.py
$ torchrun --nnodes=1 --nproc-per-node=$(nvidia-smi -L | wc -l) te_layer_with_overlap.py

# Sample output on 8x H100s:
# [rank0:node0] |-- Created tensor-parallel group: [0, 1, 2, 3, 4, 5, 6, 7]
Expand Down Expand Up @@ -70,7 +70,7 @@ Uses `torch.nn.parallel.DistributedDataParallel` for replicatin the model across
groups in a single node.

```bash
$ torchrun --nnodes=1 --nproc-per-node=$(nvidia-smi -L | wc -l) ln_mlp_overlap.py --num-replicas 2
$ torchrun --nnodes=1 --nproc-per-node=$(nvidia-smi -L | wc -l) te_layer_with_overlap.py --num-replicas 2

# Sample output on 8x H100s:
# [rank0:node0] |-- Created tensor-parallel group: [0, 1, 2, 3]
Expand Down

0 comments on commit 1ae8190

Please sign in to comment.