Search results

Eval bug: Crashes when using version 0.3.0, when using two or three cards. Gemma 31b, dflash

Status: Open.

#38 In Anbeeld/beellama.cpp;

· nikitabalakin opened on May 25, 2026

Eval bug: When trying to WRITE a large content file I see the below error in the logs and the write timesout...

Status: Open.

#37 In Anbeeld/beellama.cpp;

· The-Dude-2020 opened on May 25, 2026

Eval bug: MTP output is random garbage

Status: Open.

#36 In Anbeeld/beellama.cpp;

· ethernidee opened on May 25, 2026

Eval bug: Gemma MoE model crashing

Status: Open.

#34 In Anbeeld/beellama.cpp;

· Offset0x opened on May 24, 2026

DFlash segfault (exit 139) with Qwen3.6-27B on RTX PRO 4500 Blackwell (SM120)

Status: Open.

#33 In Anbeeld/beellama.cpp;

· gdeyoung opened on May 24, 2026

Fatal errors on startup in Windows 11

Status: Open.

#31 In Anbeeld/beellama.cpp;

· artyfacialintelagent opened on May 23, 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: Multi GPU with draft errors: - the tokens for sequence 0 in the input batch have a starting position of Y = 1360 it is required that the sequence positions remain consecutive: Y = X + 1

Status: Open.

#28 In Anbeeld/beellama.cpp;

· FortinFred opened on May 22, 2026

invalid reduced-logits token -1

Status: Open.

#27 In Anbeeld/beellama.cpp;

· l3ateman opened on May 22, 2026

DFlash speculative + multi-slot (-np > 1) crashes with CUDA error on first multi-slot decode (sm_120 Blackwell)

Status: Open.

#24 In Anbeeld/beellama.cpp;

· alexop1000 opened on May 21, 2026

No speed gain on 20gb vram with 27b q4

Status: Open.

#22 In Anbeeld/beellama.cpp;

· cryptopsy0 opened on May 19, 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