lint

* Check exists projtype

* Maintain Ordering

api/types.go

@@ -327,7 +327,6 @@ type ProcessResponse struct {
 // ListModelResponse is a single model description in [ListResponse].
 type ListModelResponse struct {
 	Name       string       `json:"name"`
-	Model      string       `json:"model"`
 	ModifiedAt time.Time    `json:"modified_at"`
 	Size       int64        `json:"size"`
 	Digest     string       `json:"digest"`
@@ -336,7 +335,6 @@ type ListModelResponse struct {
 
 // ProcessModelResponse is a single model description in [ProcessResponse].
 type ProcessModelResponse struct {
-	Name      string       `json:"name"`
 	Model     string       `json:"model"`
 	Size      int64        `json:"size"`
 	Digest    string       `json:"digest"`

cmd/cmd.go

@@ -526,7 +526,7 @@ func ListRunningHandler(cmd *cobra.Command, args []string) error {
 	var data [][]string
 
 	for _, m := range models.Models {
-		if len(args) == 0 || strings.HasPrefix(m.Name, args[0]) {
+		if len(args) == 0 || strings.HasPrefix(m.Model, args[0]) {
 			var procStr string
 			switch {
 			case m.SizeVRAM == 0:
@@ -540,7 +540,7 @@ func ListRunningHandler(cmd *cobra.Command, args []string) error {
 				cpuPercent := math.Round(float64(sizeCPU) / float64(m.Size) * 100)
 				procStr = fmt.Sprintf("%d%%/%d%% CPU/GPU", int(cpuPercent), int(100-cpuPercent))
 			}
-			data = append(data, []string{m.Name, m.Digest[:12], format.HumanBytes(m.Size), procStr, format.HumanTime(m.ExpiresAt, "Never")})
+			data = append(data, []string{m.Model, m.Digest[:12], format.HumanBytes(m.Size), procStr, format.HumanTime(m.ExpiresAt, "Never")})
 		}
 	}
 
@@ -672,11 +672,17 @@ func ShowHandler(cmd *cobra.Command, args []string) error {
 		projectorData := [][]string{
 			{"arch", "clip"},
 			{"parameters", format.HumanNumber(uint64(resp.ProjectorInfo["general.parameter_count"].(float64)))},
-			{"projector type", resp.ProjectorInfo["clip.projector_type"].(string)},
-			{"embedding length", fmt.Sprintf("%v", resp.ProjectorInfo["clip.vision.embedding_length"].(float64))},
-			{"projection dimensionality", fmt.Sprintf("%v", resp.ProjectorInfo["clip.vision.projection_dim"].(float64))},
 		}
 
+		if projectorType, ok := resp.ProjectorInfo["clip.projector_type"]; ok {
+			projectorData = append(projectorData, []string{"projector type", projectorType.(string)})
+		}
+
+		projectorData = append(projectorData,
+			[]string{"embedding length", fmt.Sprintf("%v", resp.ProjectorInfo["clip.vision.embedding_length"].(float64))},
+			[]string{"projection dimensionality", fmt.Sprintf("%v", resp.ProjectorInfo["clip.vision.projection_dim"].(float64))},
+		)
+
 		mainTableData = append(mainTableData,
 			[]string{"Projector"},
 			[]string{renderSubTable(projectorData, false)},

docs/api.md

@@ -1086,7 +1086,6 @@ A single JSON object will be returned.
 {
   "models": [
     {
-      "name": "mistral:latest",
       "model": "mistral:latest",
       "size": 5137025024,
       "digest": "2ae6f6dd7a3dd734790bbbf58b8909a606e0e7e97e94b7604e0aa7ae4490e6d8",

docs/openai.md

@@ -104,7 +104,6 @@ curl http://localhost:11434/v1/chat/completions \
 
 #### Notes
 
-- `finish_reason` will always be `stop`
 - `usage.prompt_tokens` will be 0 for completions where prompt evaluation is cached
 
 ## Models

llm/ggml.go

@@ -366,9 +366,18 @@ func (llm GGML) GraphSize(context, batch uint64) (partialOffload, fullOffload ui
 				4*batch*(1+2*embedding+context*(1+heads))+embedding*(6*context*headsKV/heads+embedding*9/16),
 			)
 		}
-	case "gemma":
-		fullOffload = 4 * batch * (embedding + vocab)
-		partialOffload = 4*batch*(2*embedding+vocab+1) + embedding*vocab*105/128
+	case "gemma", "gemma2":
+		fullOffload = max(
+			4*batch*(embedding+vocab),
+			4*batch*(2+context+context*heads+2*embedding+2*embeddingHeadsK*heads),
+		)
+
+		partialOffload = max(
+			4*embedding*batch+embedding*vocab*105/128+4*vocab*batch,
+			4*batch*(2*embedding+1+2*embeddingHeadsK*heads+context+context*heads)+
+				4*embeddingHeadsK*context*8+
+				embedding*embeddingHeadsK*heads*9/16,
+		)
 	case "command-r":
 		fullOffload = max(
 			4*batch*(embedding+vocab),

server/routes.go

@@ -800,7 +800,6 @@ func (s *Server) ListModelsHandler(c *gin.Context) {
 
 		// tag should never be masked
 		models = append(models, api.ListModelResponse{
-			Model:      n.DisplayShortest(),
 			Name:       n.DisplayShortest(),
 			Size:       m.Size(),
 			Digest:     m.digest,
@@ -1219,7 +1218,6 @@ func (s *Server) ProcessHandler(c *gin.Context) {
 
 		mr := api.ProcessModelResponse{
 			Model:     model.ShortName,
-			Name:      model.ShortName,
 			Size:      int64(v.estimatedTotal),
 			SizeVRAM:  int64(v.estimatedVRAM),
 			Digest:    model.Digest,