LocalAI/core/http/endpoints/openai/inference.go
Ettore Di Giacinto fbb9facda4
feat(api): allow to pass videos to backends (#3601)
This prepares the API to receive videos as well for video understanding.

It works similarly to images, where the request should be in the form:

{
 "type": "video_url",
 "video_url": { "url": "url or base64 data" }
}

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
2024-09-19 11:21:59 +02:00

60 lines
1.5 KiB
Go

package openai
import (
"github.com/mudler/LocalAI/core/backend"
"github.com/mudler/LocalAI/core/config"
"github.com/mudler/LocalAI/core/schema"
model "github.com/mudler/LocalAI/pkg/model"
)
func ComputeChoices(
req *schema.OpenAIRequest,
predInput string,
config *config.BackendConfig,
o *config.ApplicationConfig,
loader *model.ModelLoader,
cb func(string, *[]schema.Choice),
tokenCallback func(string, backend.TokenUsage) bool) ([]schema.Choice, backend.TokenUsage, error) {
n := req.N // number of completions to return
result := []schema.Choice{}
if n == 0 {
n = 1
}
images := []string{}
for _, m := range req.Messages {
images = append(images, m.StringImages...)
}
videos := []string{}
for _, m := range req.Messages {
videos = append(videos, m.StringVideos...)
}
// get the model function to call for the result
predFunc, err := backend.ModelInference(req.Context, predInput, req.Messages, images, videos, loader, *config, o, tokenCallback)
if err != nil {
return result, backend.TokenUsage{}, err
}
tokenUsage := backend.TokenUsage{}
for i := 0; i < n; i++ {
prediction, err := predFunc()
if err != nil {
return result, backend.TokenUsage{}, err
}
tokenUsage.Prompt += prediction.Usage.Prompt
tokenUsage.Completion += prediction.Usage.Completion
finetunedResponse := backend.Finetune(*config, predInput, prediction.Response)
cb(finetunedResponse, &result)
//result = append(result, Choice{Text: prediction})
}
return result, tokenUsage, err
}