oss/syft
2
0
Fork 0
mirror of https://github.com/anchore/syft.git synced 2026-07-05 02:28:25 +02:00
Code Issues Packages Projects Releases Wiki Activity

chore: document one model per directory grouping

Browse Source 
groupByParentDir treats every .safetensors file in a directory as a shard of
one model. Spell out that assumption and its trade-off

Signed-off-by: Christopher Phillips <32073428+spiffcs@users.noreply.github.com>
This commit is contained in:
Christopher Phillips 2026-06-30 15:03:06 -04:00
parent 4d59bdbb7f
commit 88ef52f617
No known key found for this signature in database
1 changed files with 11 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

13
syft/pkg/cataloger/ai/processor.go
View File

@ -97,8 +97,17 @@ func mergeDirModels(ctx context.Context, resolver file.Resolver, pkgs []pkg.Pack
return out
}
// groupByParentDir buckets filesystem-scanned packages by the directory their
// primary-evidence file lives in (the shards of one model share a directory).
// groupByParentDir buckets filesystem-scanned models by the directory their
// primary-evidence file lives in.
// This encodes a deliberate assumption: a directory holds one logical
// model, so every .safetensors file in a directory is treated as a shard of the
// same modeland merged into one package.
// The trade-off is that if a directory happens to contain several unrelated models,
// they are merged into one package rather than reported separately.
// We accept that because the conventional on-disk layout gives each model
// (with all of its shards) its own directory.
// We have no reliable per-file signal to tell co-located-but-independent models apart.
func groupByParentDir(pkgs []pkg.Package) map[string][]pkg.Package {
out := make(map[string][]pkg.Package)
for _, p := range pkgs {