Details on the code used to produce and use this model are available at:
https://github.com/schrum2/MarioDiffusion
That repo has instructions to check out this model and apply it to the generation of Super Mario Bros. level scenes. There is also an interactive GUI for constructing complete levels out of model-generated scenes.
This model makes use of https://huggingface.co/Alibaba-NLP/gte-large-en-v1.5 as a text embedding model for use with diffusion to generate Mario levels. It makes use of negative guidance during diffusion training, and encodes each of several phrases within each caption as a distinct text embedding. Our results indicate that it has relatively poor performance, so it is made available only to allow full scrutiny of our results.
To see a model using Alibaba-NLP/gte-large-en-v1.5 that performs better with multiple text embeddings without using negative guidance, see https://huggingface.co/schrum2/MarioDiffusion-GTE-multiple-regular0. To see a model that uses a simple token-based transformer model for text embedding and makes use of negative guidance, see https://huggingface.co/schrum2/MarioDiffusion-MLM-negative0.
- Downloads last month
- 2