Search results

Crash on mlx 0.31.2: 'There is no Stream(gpu, N) in current thread' when generate() runs in a worker thread

Status: Open.

#1049 In Blaizzy/mlx-vlm;

· csheaff opened on Apr 22, 2026

/responses accepts previous_response_id but does not replay previous response context

Status: Open.

#1046 In Blaizzy/mlx-vlm;

· eloe opened on Apr 21, 2026

OpenAI-compatible server accepts tool_choice but does not enforce it on /chat/completions or /responses

Status: Open.

#1045 In Blaizzy/mlx-vlm;

· eloe opened on Apr 21, 2026

OpenAI-compatible server accepts stop but silently ignores it on /chat/completions and /responses

Status: Open.

#1044 In Blaizzy/mlx-vlm;

· eloe opened on Apr 21, 2026

20260419 -- Diagnostics Report — 2 failure(s), 11 harness issue(s) (mlx-vlm 0.4.4)

Status: Open.

#1038 In Blaizzy/mlx-vlm;

· jrp2014 opened on Apr 19, 2026

Batch gen TQB kernel support

Status: Open.

#1036 In Blaizzy/mlx-vlm;

· Blaizzy opened on Apr 19, 2026

Gemma 4 chat template: assistant tool-call turn not closed after tool_response when content is empty leads to infinite tool-call loop

Status: Open.

#1034 In Blaizzy/mlx-vlm;

· reza-yousefi opened on Apr 17, 2026

Gemma 4 chat template: assistant tool-call turn not closed after tool_response when content is empty → infinite tool-call loop

Status: Open.

#1033 In Blaizzy/mlx-vlm;

· reza-yousefi opened on Apr 17, 2026

Qwen2-VL-2B models have started to fail: shapes cannot be broadcast

Status: Open.

#1031 In Blaizzy/mlx-vlm;

· jrp2014 opened on Apr 17, 2026

Gemma-4 mxfp8 missing language_model.model.per_layer_model_projection.scales

Status: Open.

#1022 In Blaizzy/mlx-vlm;

· nightmedia opened on Apr 14, 2026

GLM-OCR: /chat/completions produces repetition loops on prose; repetition_penalty does not prevent them

Status: Open.

#1021 In Blaizzy/mlx-vlm;

· esaruoho opened on Apr 14, 2026

Fix deepstack_visual_embeds being silently set to None in Qwen3-VL

Status: Open.

#1020 In Blaizzy/mlx-vlm;

· tiandunx opened on Apr 14, 2026