llama.cpp
d9b33fe9 - metal : round up to 16 to fix MTLDebugComputeCommandEncoder assertion (#3938)

Commit

1 year ago

metal : round up to 16 to fix MTLDebugComputeCommandEncoder assertion (#3938)

References

#3938 - ggml-metal: round up to 16 to fix setThreadgroupMemoryLength assertion

Author

psugihara

psugihara

Parents

Loading