Search results

Eval bug: dflash: target and drafter vocab are incompatible; DFlash cannot retokenize draft outputs

Status: Open.

#51 In Anbeeld/beellama.cpp;

· ovadmani-sudo opened on Jun 2, 2026

Misc. bug: xmfp6 and turbo3 sharing same ID:42 this preventing from loading xmfp6 models

Status: Open.

#49 In Anbeeld/beellama.cpp;

· ovadmani-sudo opened on Jun 2, 2026

Eval bug: dflash: invalid reduced-logits token -1 in draft

Status: Open.

#47 In Anbeeld/beellama.cpp;

· Riconec opened on May 31, 2026

Eval bug: suboptimal DFlash multi-slot performance

Status: Open.

#46 In Anbeeld/beellama.cpp;

· Anbeeld opened on May 31, 2026

Eval bug: llama.cpp (AMD GPU): partial layers run on CPU when loading dflash model, cannot fully offload to GPU

Status: Open.

#45 In Anbeeld/beellama.cpp;

· rykerzhou opened on May 29, 2026

Misc. bug: can not stop output

Status: Open.

#44 In Anbeeld/beellama.cpp;

· callMeMakerRen opened on May 29, 2026

Eval bug: DFlash keeps outputting / on my multi-GPU system

Status: Open.

#42 In Anbeeld/beellama.cpp;

· IndigoFloyd opened on May 27, 2026

Eval bug: Segfault at slot initialization with CUDA on SM75 (Turing) — ngl > 0

Status: Open.

#41 In Anbeeld/beellama.cpp;

· Bino5150 opened on May 26, 2026

profiling radeon 7900xt (20gb vram)

Status: Open.

#40 In Anbeeld/beellama.cpp;

· cryptopsy0 opened on May 26, 2026

Eval bug: low DFlash tps/AR for multi-GPU setups

Status: Open.

#39 In Anbeeld/beellama.cpp;

· Anbeeld opened on May 26, 2026

Compile bug: Metal library fails to compile — block_turbo4_0 missing signs / rnorm fields referenced by quantize_turbo4_0

Status: Open.

#30 In Anbeeld/beellama.cpp;

· Hrsh-Venket opened on May 23, 2026

Eval bug: Qwen3.6 35B A3B Q4 crash with "longer" context using turbo3 + dflash, cpu-offloading | Win10

Status: Open.

#15 In Anbeeld/beellama.cpp;

· dani-nagy opened on May 15, 2026