Skip to content

preflight: cuda-0-40-2 and mlx-cuda-0-40-2 admit the v0.40.3 lineage #492

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
glennneuber merged 1 commit into from
Oct 13, 2026
Merged

preflight: cuda-0-40-2 and mlx-cuda-0-40-2 admit the v0.40.3 lineage #492

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
10 changes: 8 additions & 2 deletions docs/maxusai/vision-suite/preflight/expectations.toml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3587,7 +3587,11 @@ model = "qwen2.5vl:3b-q4_K_M"
# payload_proof on that regression.
# - 903 here is the tag's. Builds of main after #462 carry #462's 903 (#29953 as merged plus the y-tile guard)
# under the same label; it changes the MMQ src1/ids_dst padding, not image token accounting.
version_pattern = '^0\.40\.2-dynres(\.\d+)?-\d+-g[0-9a-f]{7,40}$'
# Widened 2026-10-14 to 0.40.3 for the v0.40.3 fold (#488), as ADR 0032 says for a Go-only fold: v0.40.3 moves no
# native input (llama.cpp b11351 and MLX a59cc231 stay), so its builds share this payload and this profile. Its head,
# 0.40.2-dynres.1-6-g9f3baaf, passed here on a production-like canary (--gpus all): PASS=20 SKIP=9. 0.40.4 stays
# refused: its rc1 moves llama.cpp to b11521.
version_pattern = '^0\.40\.[23]-dynres(\.\d+)?-\d+-g[0-9a-f]{7,40}$'
patchset = ["001", "002", "004", "005", "801", "802", "805", "903", "908"]
llama_cpp_build = "631109b34"
mlx_build = "a59cc23190f1e26fabcba8693607bb4c34b2ac2b"
Expand Down Expand Up @@ -3780,7 +3784,9 @@ require_nonempty_thinking = true
# 13, 13.
# - main-601b8a65 (0.40.2-dynres-10-g601b8a6, the full build of main 601b8a65c, which v0.40.2-dynres.1 restamps):
# PASS=11 SKIP=18 (runs/preflight-mlx-cuda-0402-main-g601b8a6.json).
version_pattern = '^0\.40\.2-dynres(\.\d+)?-\d+-g[0-9a-f]{7,40}$'
# Widened 2026-10-14 to 0.40.3 with cuda-0-40-2 (#488, Go-only): the v0.40.3 fold's head passed here on the
# production-like canary, PASS=11 SKIP=18, 0 cudaMemcpy errors with both GPUs visible. 0.40.4 stays refused.
version_pattern = '^0\.40\.[23]-dynres(\.\d+)?-\d+-g[0-9a-f]{7,40}$'
mlx_build = "a59cc23190f1e26fabcba8693607bb4c34b2ac2b"
platform = "mlx-cuda"
reference_image = "maxusai/ollama:0.32.14-rc0-dynres-mlxfix"
Expand Down
Loading