Repository navigation
Conversation
The shard writer joined rows with "\n" and wrote no trailing newline, while the launcher counts pending rows with `wc -l`, which counts newlines. A shard with exactly one pending row therefore read as 0 and its GPU worker was skipped, leaving that video ungenerated. Terminate every row with a newline so the count matches the rows written. Diagnosis from NVIDIA/cosmos#376. Signed-off-by: Tarang (TJ) Jammalamadaka <tarangjammalamadaka9@gmail.com>
This branch has not been deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Ports the fix for NVIDIA/cosmos#376 (diagnosis by @JIN-strong; their PR NVIDIA/cosmos#377 targets the old repository layout) into
run_paibench_c.sh.Bug. The shard writer joins rows with
"\n"and never writes a trailing newline. The launcher then counts pending rows withwc -l, which counts newlines, so a shard with exactly one pending row reads as 0, logs "all tasks done, skipping", and that video is never generated. Larger shards are undercounted by one but still run.Fix. Write every row with its own terminating newline (one changed line plus a comment). An empty shard still produces an empty file, so the zero case behaves as before.
Verification (CPU only). I ran the old and new write expressions through the same
wc -l < file | tr -d ' 'the launcher uses:Every file parses back to the same rows either way.
bash -npasses. The Python block sits inside a quoted heredoc (<<'PY'), so the backticks in the new comment are not expanded. I have not run a full GPU benchmark.This touches a different region of the script than #315, so the two should not conflict.