From 46e485f32cda551c2b39876bcf05eee2d1a559fb Mon Sep 17 00:00:00 2001
From: Michael Yang <git@mxy.ng>
Date: Tue, 2 Sep 2025 10:42:16 -0700
Subject: [PATCH] runner: disable embedding models in ollama engine

---
 fs/ggml/ggml.go | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/fs/ggml/ggml.go b/fs/ggml/ggml.go
index 3f4374cd00..2eb1dbddeb 100644
--- a/fs/ggml/ggml.go
+++ b/fs/ggml/ggml.go
@@ -175,6 +175,10 @@ func (kv KV) Bools(key string, defaultValue ...[]bool) []bool {
 }
 
 func (kv KV) OllamaEngineRequired() bool {
+	if kv.Uint("pooling_type") > 0 {
+		return false
+	}
+
 	return slices.Contains([]string{
 		"gemma3",
 		"gemma3n",
@@ -769,8 +773,7 @@ func (f GGML) SupportsKVCacheType(cacheType string) bool {
 
 // SupportsFlashAttention checks if the model supports flash attention
 func (f GGML) SupportsFlashAttention() bool {
-	_, isEmbedding := f.KV()[fmt.Sprintf("%s.pooling_type", f.KV().Architecture())]
-	if isEmbedding {
+	if f.KV().Uint("pooling_type") > 0 {
 		return false
 	}