ggml
bcdb75e3 - CUDA: faster q8_0 -> f16 dequantization (llama/4895)

Commit

2 years ago

CUDA: faster q8_0 -> f16 dequantization (llama/4895)

Author

JohannesGaessler

JohannesGaessler

Committer

ggerganov

ggerganov

Parents

Loading