llama: update vendored code to commit 40c6d79f (#7875)

527cc978 · Jeffrey Morgan · GitHub · a37f4a86 · 527cc978 · 527cc978
Unverified Commit 527cc978 authored Dec 10, 2024 by Jeffrey Morgan Committed by GitHub Dec 10, 2024
20 changed files
--- a/llama/ggml-cuda/fattn-common.cuh
+++ b/llama/ggml-cuda/fattn-common.cuh
--- a/llama/ggml-cuda/fattn-tile-f16.cu
+++ b/llama/ggml-cuda/fattn-tile-f16.cu
--- a/llama/ggml-cuda/fattn-tile-f16.cuh
+++ b/llama/ggml-cuda/fattn-tile-f16.cuh
--- a/llama/ggml-cuda/fattn-tile-f32.cu
+++ b/llama/ggml-cuda/fattn-tile-f32.cu
--- a/llama/ggml-cuda/fattn-tile-f32.cuh
+++ b/llama/ggml-cuda/fattn-tile-f32.cuh
--- a/llama/ggml-cuda/fattn-vec-f16.cuh
+++ b/llama/ggml-cuda/fattn-vec-f16.cuh
--- a/llama/ggml-cuda/fattn-vec-f32.cuh
+++ b/llama/ggml-cuda/fattn-vec-f32.cuh
--- a/llama/ggml-cuda/fattn-wmma-f16.cuh
+++ b/llama/ggml-cuda/fattn-wmma-f16.cuh
--- a/llama/ggml-cuda/fattn.cu
+++ b/llama/ggml-cuda/fattn.cu
--- a/llama/ggml-cuda/fattn.cuh
+++ b/llama/ggml-cuda/fattn.cuh
--- a/llama/ggml-cuda/getrows.cu
+++ b/llama/ggml-cuda/getrows.cu
--- a/llama/ggml-cuda/getrows.cuh
+++ b/llama/ggml-cuda/getrows.cuh
--- a/llama/ggml-cuda.cu
+++ b/llama/ggml-cuda.cu
--- a/llama/ggml-cuda/im2col.cu
+++ b/llama/ggml-cuda/im2col.cu
--- a/llama/ggml-cuda/im2col.cuh
+++ b/llama/ggml-cuda/im2col.cuh
--- a/llama/ggml-cuda/mma.cuh
+++ b/llama/ggml-cuda/mma.cuh
--- a/llama/ggml-cuda/mmq.cu
+++ b/llama/ggml-cuda/mmq.cu
--- a/llama/ggml-cuda/mmq.cuh
+++ b/llama/ggml-cuda/mmq.cuh
--- a/llama/ggml-cuda/mmv.cu
+++ b/llama/ggml-cuda/mmv.cu
--- a/llama/ggml-cuda/dmmv.cuh
+++ b/llama/ggml-cuda/dmmv.cuh