Search results

Gemma4 Value Error - Broadcast Shapes

Status: Open.

#923 In Blaizzy/mlx-vlm;

· jayluxferro opened on Apr 4, 2026

Gemma 4: sanitize() duplicates 'model.' prefix, all weights load as zero

Status: Open.

#912 In Blaizzy/mlx-vlm;

· raullenchai opened on Apr 4, 2026

[Bug] sft_trainer crashes on checkpoint save — args.adapter_file is None

Status: Open.

#908 In Blaizzy/mlx-vlm;

· Snail3D opened on Apr 3, 2026

[Bug] Server drops LoRA adapters after every request — adapter_path not preserved in cache

Status: Open.

#907 In Blaizzy/mlx-vlm;

· Snail3D opened on Apr 3, 2026

kv-quant-scheme TurboQuant is not working with MoE

Status: Open.

#904 In Blaizzy/mlx-vlm;

· qwertyczx opened on Apr 3, 2026

20260327 -- Diagnostic Report (transformers 5.4)

Status: Open.

#866 In Blaizzy/mlx-vlm;

· jrp2014 opened on Mar 27, 2026

PagedEviction: Structured Block-wise KV Cache Pruning for Efficient Large Language Model Inference

Status: Open.

#863 In Blaizzy/mlx-vlm;

· asrvastava opened on Mar 26, 2026

Support for siglip_vision_model in llava_next?

Status: Open.

#862 In Blaizzy/mlx-vlm;

· dmolesUC opened on Mar 26, 2026

Qwen3-VL-8B: Significantly degraded visual understanding compared to LM Studio with same weights

Status: Open.

#856 In Blaizzy/mlx-vlm;

· Hassan-A-K opened on Mar 24, 2026

20260322 -- Diagnostic report

Status: Open.

#852 In Blaizzy/mlx-vlm;

· jrp2014 opened on Mar 22, 2026

LoRaLayer uses raw alpha instead of alpha/rank — 8x scaling error with default settings

Status: Open.

#845 In Blaizzy/mlx-vlm;

· kikoncuo opened on Mar 21, 2026

In multi-modal training of qwen models tokenizer-level chat templating can fail for list-based multimodal content

Status: Open.

#840 In Blaizzy/mlx-vlm;

· quist00 opened on Mar 20, 2026