mirror of
https://github.com/dogkeeper886/ollama37.git
synced 2025-12-12 08:47:01 +00:00
ccb7eb8135
When ggml_backend_buffer_free() is called, the device memory is released but not all backends consistently release the actual ggml_backend_buffer_t in system RAM, causing a memory leak. Bug #10040
26 lines
919 B
Diff
26 lines
919 B
Diff
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 dfff21a2..1b0d074b 100644
|
|
--- a/ggml/src/ggml-cuda/ggml-cuda.cu
|
|
+++ b/ggml/src/ggml-cuda/ggml-cuda.cu
|
|
@@ -2284,9 +2284,11 @@ 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;
|