highperfocused/ollama
1
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2025-11-11 11:27:25 +01:00
Code Issues Packages Projects Releases Wiki Activity

Merge pull request #5051 from ollama/mxyng/capabilities

Browse Source 
add model capabilities
This commit is contained in:
Michael Yang
2024-07-02 14:26:07 -07:00
committed by GitHub
parent 400056e154 da8e2a0447
commit dddb58a38b
31 changed files with 354 additions and 192 deletions
Download Patch File Download Diff File Expand all files Collapse all files

34
server/routes.go
View File

@@ -31,6 +31,7 @@ import (
"github.com/ollama/ollama/llm"
"github.com/ollama/ollama/openai"
"github.com/ollama/ollama/parser"
"github.com/ollama/ollama/template"
"github.com/ollama/ollama/types/errtypes"
"github.com/ollama/ollama/types/model"
"github.com/ollama/ollama/version"
@@ -121,8 +122,8 @@ func (s *Server) GenerateHandler(c *gin.Context) {
return
}
if model.IsEmbedding() {
c.AbortWithStatusJSON(http.StatusBadRequest, gin.H{"error": "embedding models do not support generate"})
if !model.Has(CapabilityCompletion) {
c.JSON(http.StatusBadRequest, gin.H{"error": fmt.Sprintf("%s does not support generate", req.Model)})
return
}
@@ -161,6 +162,12 @@ func (s *Server) GenerateHandler(c *gin.Context) {
return
}
tmpl, err := template.Parse(req.Template)
if err != nil {
c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
return
}
checkpointLoaded := time.Now()
var prompt string
@@ -169,7 +176,11 @@ func (s *Server) GenerateHandler(c *gin.Context) {
prompt = req.Prompt
case req.Prompt != "":
if req.Template == "" {
req.Template = model.Template
model.Template, err = template.Parse(req.Template)
if err != nil {
c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
return
}
}
if req.System == "" {
@@ -187,7 +198,7 @@ func (s *Server) GenerateHandler(c *gin.Context) {
sb.WriteString(req.Prompt)
p, err := Prompt(req.Template, req.System, sb.String(), "", true)
p, err := Prompt(tmpl, req.System, sb.String(), "", true)
if err != nil {
c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
return
@@ -242,7 +253,7 @@ func (s *Server) GenerateHandler(c *gin.Context) {
resp.LoadDuration = checkpointLoaded.Sub(checkpointStart)
if !req.Raw {
p, err := Prompt(req.Template, req.System, req.Prompt, generated.String(), false)
p, err := Prompt(tmpl, req.System, req.Prompt, generated.String(), false)
if err != nil {
c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
return
@@ -680,7 +691,10 @@ func GetModelInfo(req api.ShowRequest) (*api.ShowResponse, error) {
}
if req.Template != "" {
m.Template = req.Template
m.Template, err = template.Parse(req.Template)
if err != nil {
return nil, err
}
}
msgs := make([]api.Message, 0)
@@ -701,7 +715,7 @@ func GetModelInfo(req api.ShowRequest) (*api.ShowResponse, error) {
resp := &api.ShowResponse{
License: strings.Join(m.License, "\n"),
System: m.System,
Template: m.Template,
Template: m.Template.String(),
Details: modelDetails,
Messages: msgs,
ModifiedAt: manifest.fi.ModTime(),
@@ -1248,7 +1262,7 @@ func (s *Server) ProcessHandler(c *gin.Context) {
}
// ChatPrompt builds up a prompt from a series of messages for the currently `loaded` model
func chatPrompt(ctx context.Context, runner *runnerRef, template string, messages []api.Message, numCtx int) (string, error) {
func chatPrompt(ctx context.Context, runner *runnerRef, template *template.Template, messages []api.Message, numCtx int) (string, error) {
encode := func(s string) ([]int, error) {
return runner.llama.Tokenize(ctx, s)
}
@@ -1296,8 +1310,8 @@ func (s *Server) ChatHandler(c *gin.Context) {
return
}
if model.IsEmbedding() {
c.AbortWithStatusJSON(http.StatusBadRequest, gin.H{"error": "embedding models do not support chat"})
if !model.Has(CapabilityCompletion) {
c.JSON(http.StatusBadRequest, gin.H{"error": fmt.Sprintf("%s does not support chat", req.Model)})
return
}