dylanyunlong/project_6
1
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

fix: revert max-num-seqs to 1 — segfault during profile_run with n=2

Browse Source 
Profile run with 2 sequences triggers segfault in torch.distributed.gather
during determine_num_available_blocks. t2_n_2 is worth 1 point, not worth
the crash risk. max_completion_tokens fix is the priority.
This commit is contained in:
project6-dev
2026-08-12 08:35:42 +00:00
parent 9ff245071b
commit 05c09a2801
1 changed files with 1 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
computility-run.yaml
View File

@@ -15,7 +15,7 @@ command:
- -tp - -tp
- '4' - '4'
- --max-num-seqs - --max-num-seqs
- '2' - '1'
- --disable-log-requests - --disable-log-requests
- --disable-frontend-multiprocessing - --disable-frontend-multiprocessing
- --max-num-batched-tokens - --max-num-batched-tokens