Adapt tests

Signed-off-by: Ettore Di Giacinto <[email protected]>

Author: mudler

pkg/templates/multimodal.go

@@ -22,7 +22,8 @@ type MultimodalContent struct {
 }
 
 // https://github.com/ggml-org/llama.cpp/blob/be1d4a13db26750fac702ceb3af88ae4f39dc9f4/tools/mtmd/mtmd.h#L42
-const DefaultMultiModalTemplate = "{{ range .Audio }}[audio-{{.ID}}]{{end}}{{ range .Images }}<__image__>{{end}}{{ range .Video }}[vid-{{.ID}}]{{end}}{{.Text}}"
+// from <__image__> to <__media__> https://github.com/ggml-org/llama.cpp/blob/79c137f77677b3c8ee3c60a7da033721b938399a/tools/mtmd/mtmd.cpp#L83
+const DefaultMultiModalTemplate = "{{ range .Audio }}<__media__>{{end}}{{ range .Images }}<__media__>{{end}}{{ range .Video }}[vid-{{.ID}}]{{end}}{{.Text}}"
 
 func TemplateMultiModal(templateString string, opts MultiModalOptions, text string) (string, error) {
 	if templateString == "" {

pkg/templates/multimodal_test.go

@@ -20,7 +20,7 @@ var _ = Describe("EvaluateTemplate", func() {
 				VideosInMessage: 0,
 			}, "bar")
 			Expect(err).NotTo(HaveOccurred())
-			Expect(result).To(Equal("<__image__>bar"))
+			Expect(result).To(Equal("<__media__>bar"))
 		})
 
 		It("should handle messages with more images correctly", func() {
@@ -33,7 +33,7 @@ var _ = Describe("EvaluateTemplate", func() {
 				VideosInMessage: 0,
 			}, "bar")
 			Expect(err).NotTo(HaveOccurred())
-			Expect(result).To(Equal("<__image__><__image__>bar"))
+			Expect(result).To(Equal("<__media__><__media__>bar"))
 		})
 		It("should handle messages with more images correctly", func() {
 			result, err := TemplateMultiModal("", MultiModalOptions{
@@ -45,7 +45,7 @@ var _ = Describe("EvaluateTemplate", func() {
 				VideosInMessage: 0,
 			}, "bar")
 			Expect(err).NotTo(HaveOccurred())
-			Expect(result).To(Equal("[audio-0]<__image__><__image__>bar"))
+			Expect(result).To(Equal("<__media__><__media__><__media__>bar"))
 		})
 		It("should handle messages with more images correctly", func() {
 			result, err := TemplateMultiModal("", MultiModalOptions{
@@ -57,7 +57,7 @@ var _ = Describe("EvaluateTemplate", func() {
 				VideosInMessage: 0,
 			}, "bar")
 			Expect(err).NotTo(HaveOccurred())
-			Expect(result).To(Equal("[audio-0]<__image__>bar"))
+			Expect(result).To(Equal("<__media__><__media__>bar"))
 		})
 		It("should handle messages with more images correctly", func() {
 			result, err := TemplateMultiModal("", MultiModalOptions{