From fcec04bf4249f91c553ec3d2914a493bf794d105 Mon Sep 17 00:00:00 2001 From: Michael Yang Date: Tue, 5 Aug 2025 15:56:12 -0700 Subject: [PATCH] gptoss: fix memory calc (#11700) --- fs/ggml/ggml.go | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/fs/ggml/ggml.go b/fs/ggml/ggml.go index fb993a28..45d58bc4 100644 --- a/fs/ggml/ggml.go +++ b/fs/ggml/ggml.go @@ -676,7 +676,7 @@ func (f GGML) GraphSize(context, batch uint64, numParallel int, kvCacheType stri } } fullOffload = 4 * f.KV().HeadCountMax() / cmp.Or(f.KV().HeadCountKVMin(), 1) * kvTotal / 6 - partialOffload = 2 * fullOffload + partialOffload = fullOffload } return