Llava

Informations

• Module : Llava (gguf)
• Function : Interrogate a chatbot about an input image using llama-cpp-python, Llava 1.5 and BakLLaVA
• Input(s) : Input image, Input text
• Output(s) : Output text
• HF models pages :
  • xtuner/llava-llama-3-8b-v1_1-gguf
  • xtuner/llava-phi-3-mini-gguf
  • mys/ggml_bakllava-1
  • cmp-nct/llava-1.6-gguf
  • mys/ggml_llava-v1.5-7b
  • mys/ggml_llava-v1.5-13b

Help

• Usage :
  • Upload or import an Input image
  • Type your request in the Input textbox field
  • (optional) modify settings to use another model, change context size or modify maximum number of tokens generated.
  • Click the Generate button to generate a response to your input, using the chatbot history to keep a context.
  • Click the Continue button to complete the last reply.
• Models :
  • You could place llama-cpp compatible .gguf models in the directory ./biniou/models/llava. Restart Biniou to see them in the models list.

Examples :

• Input image :

• Input : Describe this image.
• Output : The image features a stuffed animal with big, round eyes and floppy ears resembling a rabbit or a squirrel. It is wearing a red sweater that makes it appear cuddly and warm. In the background, there are trees and sticks giving an impression of being in nature. The stuffed toy's outfit and appearance evoke feelings of comfort and coziness, making it an appealing decorative piece for a living space or a child's bedroom.