|
--- |
|
license: apache-2.0 |
|
--- |
|
|
|
# Lite-Sora-v1-Experimental |
|
|
|
## Introduction |
|
|
|
Lite-Sora is a Sora replication project jointly initiated by East China Normal University and the ModelScope community, dedicated to constructing a lightweight and open-source video DiT model, with the aim of exploring further potential in generative models. Please refer to our [GitHub repo](https://github.com/modelscope/lite-sora). |
|
|
|
## About This Model |
|
|
|
We trained a denoising model using a small dataset [Pixabay100](https://github.com/ECNU-CILAB/Pixabay100). This model serves to demonstrate that our training code is capable of fitting the training data properly, with a resolution of 64*64. **Obviously this model is overfitting due to the limited amount of training data, and thus it lacks generalization capability at this stage. Its purpose is solely for verifying the correctness of the training algorithm.** |
|
|
|
|airport, people, crowd, busy|beach, ocean, waves, water, sand|bee, honey, insect, beehive, nature|coffee, beans, caffeine, coffee, shop| |
|
|-|-|-|-| |
|
||||| |
|
|fish, underwater, aquarium, swim|forest, woods, mystical, morning|ocean, beach, sunset, sea, atmosphere|hair, wind, girl, woman, people| |
|
||||| |
|
|reeds, grass, wind, golden, sunshine|sea, ocean, seagulls, birds, sunset|woman, flowers, plants, field, garden|wood, anemones, wildflower, flower| |
|
||||| |
|
|
