llama.cpp
Fix and restrict NVFP4 edge-cases in llama-graph
#24331

Merged

Fix and restrict NVFP4 edge-cases in llama-graph #24331

ORippler merged 2 commits into ggml-org:master from ORippler:osimons/nvfp4_scale_b4_lora_b4_bias

Move post-GEMM MUL required for dequant b4 lora and bias add

21ab2d77

Restrict build_ffn for NVFP4 to supported combinations

592c2c3c

ORippler requested a review from

CISC 70 days ago

CISC approved these changes on 2026-06-09

ORippler requested a review from

ggerganov 68 days ago

ORippler requested a review from

ngxson 68 days ago

ggerganov assigned

ggerganov 64 days ago

ggerganov approved these changes on 2026-06-16

ORippler merged 02810c7a into master 63 days ago

ORippler deleted the osimons/nvfp4_scale_b4_lora_b4_bias branch 63 days ago

Reviewers

ggerganov

CISC

ngxson

Assignees

ggerganov

Labels

None yet

Milestone

No milestone