youtu/ youtu-vl:latest

133 1 month ago

Youtu-VL: a lightweight 4B VLM built on Youtu-LLM, pioneering VLUAS to improve visual perception and multimodal understanding.Not yet runnable. Requires Ollama with the latest llama.cpp changes integrated.

vision
... /
template
4175e009c870 · 771B
{{- range $i, $_ := .Messages }}
{{- if and (eq $i 0) (ne .Role "system") }}<|begin_of_text|>system
You are a helpful assistant.<|end_of_text|>
{{ end -}}
<|begin_of_text|>{{ .Role }}
{{- if .Content }}
{{- if eq (printf "%T" .Content) "string" }}{{ .Content }}<|end_of_text|>
{{ else }}
{{- range .Content }}
{{- if eq .Type "image" }}<|vision_start|><|image_pad|><|vision_end|>{{ end }}
{{- end }}
{{- range .Content }}
{{- if eq .Type "video" }}<|vision_start|><|video_pad|><|vision_end|>{{ end }}
{{- end }}
{{- range .Content }}
{{- if .Text }}{{ .Text }}{{ end }}
{{- end }}<|end_of_text|>
{{ end }}
{{- else }}{{ .Content }}<|end_of_text|>
{{ end }}
{{ end }}
{{- if .Prompt }}<|begin_of_text|>user
{{ .Prompt }}<|end_of_text|>
{{ end }}<|begin_of_text|>assistant