llama.cpp releasesTools
b12207
ggml-cuda : add d_state=64 fast path to SSM_SCAN kernel ( #31818 ) Zamba2 and similar architectures use a state size of 64 in their recurrent-scan step, which fell outside the GPU kernel's existing…
Read at llama.cpp releases ↗Related

SiliconNvidia released its 8800 GTX GPU 20 years ago today, celebrates with Steam game homage Tom's Hardware

The brain-as-computer metaphor shaped modern AI but may be misleading the consciousness debate The Decoder

SiliconOpen-source GPU task manager reveals exactly what's eating your VRAM Tom's Hardware

SiliconHundreds arrested after Microsoft data center protest turns violent Tom's Hardware

SiliconTwo Chinese companies cleared to buy banned Nvidia AI chips have empty offices and unknown addresses Tom's Hardware
