Skip to content

Commit ecbac5f

Browse files
committed
docs: use MDX comment syntax in blog placeholder
Fern parses .md pages as MDX, so HTML comments fail to parse and blank the page. No other docs page uses HTML comments.
1 parent ee62cff commit ecbac5f

1 file changed

Lines changed: 7 additions & 7 deletions

File tree

docs/blog/tensor-parallelism-optimization.md

Lines changed: 7 additions & 7 deletions
Original file line numberDiff line numberDiff line change
@@ -4,24 +4,24 @@
44
55
## Background
66

7-
<!-- Why TP matters: model sizes past single-GPU memory, the compute/communication
8-
tradeoff versus FSDP and pipeline parallelism. -->
7+
{/* Why TP matters: model sizes past single-GPU memory, the compute/communication
8+
tradeoff versus FSDP and pipeline parallelism. */}
99

1010
## Where tensor parallelism costs performance
1111

12-
<!-- All-gather / reduce-scatter on the critical path, exposed communication,
13-
small GEMM tiles at high TP degree, NVLink versus inter-node bandwidth. -->
12+
{/* All-gather / reduce-scatter on the critical path, exposed communication,
13+
small GEMM tiles at high TP degree, NVLink versus inter-node bandwidth. */}
1414

1515
## Optimizations
1616

17-
<!-- One subsection per optimization, with the mechanism and how to enable it. -->
17+
{/* One subsection per optimization, with the mechanism and how to enable it. */}
1818

1919
## Results
2020

21-
<!-- Model, hardware, TP degrees swept, throughput and step-time deltas. -->
21+
{/* Model, hardware, TP degrees swept, throughput and step-time deltas. */}
2222

2323
## Reproducing
2424

25-
<!-- Container tag, framework config flags, launch command. -->
25+
{/* Container tag, framework config flags, launch command. */}
2626

2727
## Takeaways

0 commit comments

Comments
 (0)