Spaces:
Runtime error
Apply for community grant: Company project
Within Cmotions, a Dutch company in consulting and data science, a couple of colleagues decided to upgrade their NLP knowledge. We've fine-tuned multiple language models on lyrics from The Beatles to generate Beatles-like text. We demo the output of this model in a HF Space. After lyrics generation a title is generated and an album cover is suggested using Stable Diffusion 1.4.
We’ve trained a GPT-medium, GPT-Neo and a BLOOM 560m model for the lyrics generation and are currently evaluating the model using HF metrics and external reviewers. In this space we require the use of a GPU, mainly for the album cover suggestion.
In this space we require the use of a GPU, mainly for the album cover suggestion.
Hey there! Instead of loading and running Stable Diffusion yourself, Gradio has a mechanism that allows you to use the main official space, with a special accelerated backend.
stable_diffusion = gr.Blocks.load(name="spaces/stabilityai/stable-diffusion")
Here is an example space using this: https://huggingface.co/spaces/fffiloni/whisper-to-stable-diffusion/blob/main/app.py
That way, you have even faster inference! I hope this helps
Hi @osanseviero you pointed me to the use of the accelerated backend with Gradio. Since the release of a new stable diffusion model this is no longer working. It still works in a Colab environment (without GPU support), but it fails with a runtime error on HF Space with the following error: click() got an unexpected keyword argument 'batch'
Are things still working in your spaces, or do you have a work-around for this issue?
