media-server / src /getVideoPrompt.mts
jbilcke-hf's picture
jbilcke-hf HF staff
initial commit
07ae658
raw
history blame
718 Bytes
/*
import { HfInference } from '@huggingface/inference'
const hf = new HfInference(process.env.WEBTV_HF_API_TOKEN)
const model = 'OpenAssistant/oasst-sft-4-pythia-12b-epoch-3.5'
export const getVideoPrompt = async (seed: string) => {
const inputs = `<|prompter|>Can you propose a caption for a new video, that we could display after the first one? Here's the caption for the first video: "${seed}"<|endoftext|><|assistant|>The best caption would be:`
const { generated_text } = await hf.textGeneration({
model,
inputs,
parameters: {
max_new_tokens: 50,
return_full_text: false,
// temperature: 0.3,
}
})
return generated_text.split(/[#\[\]]/)[0].replace('"', '')
}
*/