]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commit
CUDA: faster softmax via shared memory + fp16 math (#4742)
authorJohannes Gäßler <redacted>
Tue, 9 Jan 2024 07:58:55 +0000 (08:58 +0100)
committerGitHub <redacted>
Tue, 9 Jan 2024 07:58:55 +0000 (08:58 +0100)
commit8f900abfc09851e281bc9027e0ab2f16bf079b29
tree08f27d1c3a182663c28b7f36aee767d27cadbab6
parent1fc2f265ff9377a37fd2c61eae9cd813a3491bea
CUDA: faster softmax via shared memory + fp16 math (#4742)
ggml-cuda.cu
tests/test-backend-ops.cpp