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

llama/patches/0006-conditional-fattn.patch

@@ -0,0 +1,25 @@
+From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
+From: Daniel Hiltgen <daniel@ollama.com>
+Date: Wed, 9 Oct 2024 17:26:23 -0700
+Subject: [PATCH] conditional-fattn
+
+---
+ ggml/src/ggml-cuda/ggml-cuda.cu | 2 ++
+ 1 file changed, 2 insertions(+)
+
+diff --git a/ggml/src/ggml-cuda/ggml-cuda.cu b/ggml/src/ggml-cuda/ggml-cuda.cu
+index 52aec229..cbf4fddf 100644
+--- a/ggml/src/ggml-cuda/ggml-cuda.cu
++++ b/ggml/src/ggml-cuda/ggml-cuda.cu
+@@ -2162,9 +2162,11 @@ static bool ggml_cuda_compute_forward(ggml_backend_cuda_context & ctx, struct gg
+         case GGML_OP_ARGSORT:
+             ggml_cuda_op_argsort(ctx, dst);
+             break;
++#if !defined(GGML_DISABLE_FLASH_ATTN)
+         case GGML_OP_FLASH_ATTN_EXT:
+             ggml_cuda_flash_attn_ext(ctx, dst);
+             break;
++#endif
+         case GGML_OP_CROSS_ENTROPY_LOSS:
+             ggml_cuda_cross_entropy_loss(ctx, dst);
+             break;