Skip to content

Fix Qwen35 VLM crash on text-only inference #149

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

Open
dirvine wants to merge 1 commit into
base: main
Choose a base branch
from
+4 −0
Open

Fix Qwen35 VLM crash on text-only inference #149

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
4 changes: 4 additions & 0 deletions Libraries/MLXVLM/Models/Qwen35.swift
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -921,6 +921,10 @@ enum Qwen35Language {
imageGridTHW: [THW]? = nil,
videoGridTHW: [THW]? = nil
) -> LMOutput {
// Ensure inputs is 2D [batch, seq]. Text-only callers (e.g.
// WiredMemoryUtils, TokenIterator) may pass 1D token arrays.
Copy link
Copy Markdown
Collaborator

@davidkoski davidkoski Mar 19, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I wonder if that is incorrect (as in this is the bug). I would expect the shape to always be [B, S].

So:

  • are these paths buggy in that they allow 1d arrays?
  • if not, perhaps the callers should ensure the correct shape -- otherwise this is a one-off fix for a single model

let inputs = inputs.ndim == 1 ? inputs.expandedDimensions(axis: 0) : inputs

if pixelValues != nil {
precomputedPositionIds = nil
ropeDeltas = nil
Expand Down