highperfocused/ollama
1
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2025-04-07 19:38:08 +02:00
Code Issues Packages Projects Releases Wiki Activity

fix: error if image requested without vision model

Browse Source
This commit is contained in:
Michael Yang 2025-03-13 10:37:31 -07:00
parent 45a13b1dec
commit 5e2e0b46b1
3 changed files with 10 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
model/model.go
View File

@ -22,6 +22,8 @@ import (
"github.com/ollama/ollama/model/input"
)
var ErrNoVisionModel = errors.New("vision model not found")
// Model implements a specific model architecture, defining the forward pass and any model-specific configuration
type Model interface {
Forward(ml.Context, input.Options) (ml.Tensor, error)

4
model/models/gemma3/model.go
View File

@ -84,6 +84,10 @@ func New(c ml.Config) (model.Model, error) {
}
func (m *Model) EncodeMultimodal(ctx ml.Context, multimodalData []byte) (any, error) {
if len(m.VisionModel.Layers) == 0 {
return nil, model.ErrNoVisionModel
}
image, _, err := image.Decode(bytes.NewReader(multimodalData))
if err != nil {
return nil, err

4
model/models/mllama/model.go
View File

@ -63,6 +63,10 @@ func New(c ml.Config) (model.Model, error) {
}
func (m *Model) EncodeMultimodal(ctx ml.Context, multimodalData []byte) (any, error) {
if len(m.VisionModel.Transformer.Layers) == 0 || len(m.GlobalTransformer.Layers) == 0 {
return nil, model.ErrNoVisionModel
}
image, _, err := image.Decode(bytes.NewReader(multimodalData))
if err != nil {
return nil, err