Spaces:

aavetis

/

ugen-image-captioning

like 47

Running on Zero

great model im a huge fan! any way to make it faster?

anything along the lines of vllm or so for this model arch?

batching? blfloat16? onnx? quantization?