Qeravio
Canonical AI event

The DFlash/Dspark project now supports vision targets.

The update maps vision targets to their corresponding text models for vocabulary.

24 Sept 20261 verified claims1 sources1 observations
What happened

The official source reports this update: b11151: convert : allow vision target for DFlash/Dspark (#29339). Resolve the target arch with get_model_architecture so vision targets (e.g. Lfm2VlForConditionalGeneration) map to their text model for the vocab. Fix double rope reorder for LFM2/LFM2.5 DSpark drafters

Why it matters

This official update documents a development concerning b11151: convert : allow vision target for DFlash/Dspark (#29339). Its practical significance depends on the scope and evidence stated by the source.

What to watch next

Read the official source update and verify its stated scope, evidence and timing before acting on it.

Connected knowledge

Entities affected by this event

Llama Cppsubject
Continue this topic
infrastructureGemma 4 draft support added to DSpark with GGUF conversion24 Sept 2026infrastructureLlama.cpp server fixes crash on sleep, updates wake-up process24 Sept 2026infrastructureDraft HF model deduplication implemented via dedup-cache-models24 Sept 2026infrastructureCUDA sparse-fa for dsv4 prefill re-enabled, query loop unrolled.24 Sept 2026infrastructurellama.cpp updated to version 0.5.0, now supports CUDA 13.4 libraries.24 Sept 2026infrastructureLlama.cpp adds A8 Q6_K non-MoE dp4a binary kernel24 Sept 2026
Evidence trail

Sources behind the event

llama.cpp official releasesb11151: convert : allow vision target for DFlash/Dspark (#29339)
Editorial presentation

Read the public story

Open story