deprecate ggml

- remove ggml runner
- automatically pull gguf models when ggml detected
- tell users to update to gguf in the case automatic pull fails

Co-Authored-By: Jeffrey Morgan <jmorganca@gmail.com>

server/images.go

@@ -418,6 +418,27 @@ func CreateModel(ctx context.Context, name, modelFileDir string, commands []pars
 					return err
 				}
 
+				// if the model is not in gguf format, pull the base model to try and get it in gguf format
+				if fromConfig.ModelFormat != "gguf" {
+					fn(api.ProgressResponse{Status: "updating base model"})
+					if err := PullModel(ctx, c.Args, &RegistryOptions{}, fn); err != nil {
+						log.Printf("error pulling model: %v", err)
+					}
+					// Reset the file pointer to the beginning of the file
+					_, err = fromConfigFile.Seek(0, 0)
+					if err != nil {
+						return fmt.Errorf("update from config after pull: %w", err)
+					}
+					if err := json.NewDecoder(fromConfigFile).Decode(&fromConfig); err != nil {
+						return err
+					}
+				}
+
+				// if the model is still not in gguf format, error out
+				if fromConfig.ModelFormat != "gguf" {
+					return fmt.Errorf("%s is not in gguf format, this base model is not compatible with this version of ollama", c.Args)
+				}
+
 				config.SetModelFormat(fromConfig.ModelFormat)
 				config.SetModelFamily(append(fromConfig.ModelFamilies, fromConfig.ModelFamily)...)
 				config.SetModelType(fromConfig.ModelType)
@@ -456,15 +477,21 @@ func CreateModel(ctx context.Context, name, modelFileDir string, commands []pars
 			defer bin.Close()
 
 			var offset int64
+		CREATE:
 			for {
 				fn(api.ProgressResponse{Status: "creating model layer"})
 
 				bin.Seek(offset, io.SeekStart)
 				ggml, err := llm.DecodeGGML(bin)
-				if errors.Is(err, io.EOF) {
-					break
-				} else if err != nil {
-					return err
+				if err != nil {
+					switch {
+					case errors.Is(err, io.EOF):
+						break CREATE
+					case errors.Is(err, llm.ErrUnsupportedFormat):
+						return fmt.Errorf("model binary specified in FROM field is not a valid gguf format model, %w", err)
+					default:
+						return err
+					}
 				}
 
 				config.SetModelFormat(ggml.Name())

server/routes.go

@@ -114,7 +114,7 @@ func load(c *gin.Context, modelName string, reqOpts map[string]interface{}, sess
 			// some older models are not compatible with newer versions of llama.cpp
 			// show a generalized compatibility error until there is a better way to
 			// check for model compatibility
-			if strings.Contains(err.Error(), "failed to load model") {
+			if errors.Is(llm.ErrUnsupportedFormat, err) || strings.Contains(err.Error(), "failed to load model") {
 				err = fmt.Errorf("%v: this model may be incompatible with your version of Ollama. If you previously pulled this model, try updating it by running `ollama pull %s`", err, model.ShortName)
 			}