Don't expose model information in /api/generate

This commit is contained in:
Jeffrey Morgan 2023-12-09 02:05:43 -08:00
parent b74580c913
commit 9e1406e4ed
3 changed files with 13 additions and 21 deletions

View file

@ -203,22 +203,12 @@ type GenerateResponse struct {
CreatedAt time.Time `json:"created_at"`
Response string `json:"response"`
ModelConfiguration ModelConfiguration `json:"model_configuration"`
Done bool `json:"done"`
Context []int `json:"context,omitempty"`
Metrics
}
type ModelConfiguration struct {
ModelFormat string `json:"model_format"`
ModelFamily string `json:"model_family"`
ModelFamilies []string `json:"model_families"`
ModelType string `json:"model_type"`
FileType string `json:"file_type"`
}
func (m *Metrics) Summary() {
if m.TotalDuration > 0 {
fmt.Fprintf(os.Stderr, "total duration: %v\n", m.TotalDuration)

View file

@ -146,12 +146,16 @@ type ManifestV2 struct {
}
type ConfigV2 struct {
ModelFormat string `json:"model_format"`
ModelFamily string `json:"model_family"`
ModelFamilies []string `json:"model_families"`
ModelType string `json:"model_type"`
FileType string `json:"file_type"`
// required by spec
Architecture string `json:"architecture"`
OS string `json:"os"`
RootFS RootFS `json:"rootfs"`
api.ModelConfiguration
}
func (c *ConfigV2) SetModelFormat(format string) {

View file

@ -199,10 +199,9 @@ func GenerateHandler(c *gin.Context) {
// an empty request loads the model
if req.Prompt == "" && req.Template == "" && req.System == "" {
c.JSON(http.StatusOK, api.GenerateResponse{
CreatedAt: time.Now().UTC(),
Model: req.Model,
ModelConfiguration: model.Config.ModelConfiguration,
Done: true})
CreatedAt: time.Now().UTC(),
Model: req.Model,
Done: true})
return
}
@ -261,11 +260,10 @@ func GenerateHandler(c *gin.Context) {
}
resp := api.GenerateResponse{
Model: r.Model,
ModelConfiguration: model.Config.ModelConfiguration,
CreatedAt: r.CreatedAt,
Done: r.Done,
Response: r.Content,
Model: r.Model,
CreatedAt: r.CreatedAt,
Done: r.Done,
Response: r.Content,
Metrics: api.Metrics{
TotalDuration: r.TotalDuration,
LoadDuration: r.LoadDuration,