Search results

Unsupported gpu architecture 'compute_120a'

Status: Open.

#58 In TheTom/turboquant_plus;

· Ezzz-dev opened on Mar 31, 2026

turbo - cant compile in Windows

Status: Open.

#57 In TheTom/turboquant_plus;

· Ezzz-dev opened on Mar 31, 2026

turbo3/turbo4 cache produces garbled output on NVIDIA Blackwell GPU (RTX 5070 Laptop, compute capability 12.0)

Status: Open.

#56 In TheTom/turboquant_plus;

· arifkurnaz opened on Mar 31, 2026

Norm correction validated on vLLM + Nemotron hybrid

Status: Open.

#55 In TheTom/turboquant_plus;

· MidasMining opened on Mar 31, 2026

Diagnostic: Mac Studio M2 Ultra

Status: Open.

#50 In TheTom/turboquant_plus;

· dylanzr opened on Mar 31, 2026

feat: GPU as async dequant coprocessor for CPU-primary inference

Status: Open.

#49 In TheTom/turboquant_plus;

· homeofe opened on Mar 30, 2026

No Difference in tokens/sec - Ministral3 8B Q5_K_M

Status: Open.

#48 In TheTom/turboquant_plus;

· MrMuhannadObeidat opened on Mar 29, 2026

CUDA doesn't work

Status: Open.

#47 In TheTom/turboquant_plus;

· DragonRollDev opened on Mar 29, 2026

Don't work on vulkan device

Status: Open.

#46 In TheTom/turboquant_plus;

· vincalkr opened on Mar 28, 2026

Engineering findings: K/V norm disparity + MSE > Prod + outlier mixed precision

Status: Open.

#45 In TheTom/turboquant_plus;

· scos-lab opened on Mar 28, 2026

Enable sparse V dequant on pre-M5 hardware after verification

Status: Open.

#42 In TheTom/turboquant_plus;

· TheTom opened on Mar 27, 2026

Review spiritbuun's CUDA fork for portable optimizations

Status: Open.

#40 In TheTom/turboquant_plus;

· TheTom opened on Mar 27, 2026