highperfocused/ollama
1
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2025-11-10 20:17:59 +01:00
Code Issues Packages Projects Releases Wiki Activity

create: inherit FROM model's renderer/parser

Browse Source 
On main, the `RENDERER` and `PARSER` fields from the `Modelfile` don't
get propagated to a new model created with a `req.From` parameter. This
is easily triggered via `ollama run qwen3-coder`, then running some save
command like `/save qwen3-coder-custom`.

Added a regression test for this, and then open the config for the
"from" model in order to use its renderer/parser as a default for the
new model. This will fix the CLI and also API-based creates.

Fixes: https://github.com/ollama/ollama/issues/12792
This commit is contained in:
Devon Rifkin
2025-10-27 15:14:19 -07:00
parent ad6f6a1d29
commit 1bdd816910
2 changed files with 87 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

21
server/create.go
View File

@@ -119,6 +119,27 @@ func (s *Server) CreateHandler(c *gin.Context) {
if err != nil {
ch <- gin.H{"error": err.Error()}
}
if err == nil && !remote && (config.Renderer == "" || config.Parser == "") {
manifest, mErr := ParseNamedManifest(fromName)
if mErr == nil && manifest.Config.Digest != "" {
configPath, pErr := GetBlobsPath(manifest.Config.Digest)
if pErr == nil {
if cfgFile, fErr := os.Open(configPath); fErr == nil {
var baseConfig ConfigV2
if decErr := json.NewDecoder(cfgFile).Decode(&baseConfig); decErr == nil {
if config.Renderer == "" {
config.Renderer = baseConfig.Renderer
}
if config.Parser == "" {
config.Parser = baseConfig.Parser
}
}
cfgFile.Close()
}
}
}
}
}
} else if r.Files != nil {
baseLayers, err = convertModelFromFiles(r.Files, baseLayers, false, fn)