Update README.md
Browse files
README.md
CHANGED
|
@@ -9,11 +9,20 @@ pinned: false
|
|
| 9 |
|
| 10 |
# Open Reproduction Effort to reproduce MUSE
|
| 11 |
|
| 12 |
-
|
|
|
|
|
|
|
|
|
|
| 13 |
|
| 14 |
## Steps:
|
| 15 |
|
| 16 |
-
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 17 |
|
| 18 |
## Weekly Log
|
| 19 |
|
|
|
|
| 9 |
|
| 10 |
# Open Reproduction Effort to reproduce MUSE
|
| 11 |
|
| 12 |
+
Open-muse is a project to openly reproduce the transformer-based [MUSE](https://muse-model.github.io/) text-to-image model.
|
| 13 |
+
The [open-muse](https://github.com/huggingface/muse) repo contains the model implementation, training code and everything related to training.
|
| 14 |
+
All the artefacts of this project, such as the training checkpoints and final models will be uploaded to this organization.
|
| 15 |
+
|
| 16 |
|
| 17 |
## Steps:
|
| 18 |
|
| 19 |
+
1. Setup the codebase and train a class-conditional model on imagenet.
|
| 20 |
+
2. Conduct text2image experiments on CC12M.
|
| 21 |
+
3. Train improved VQGANs models.
|
| 22 |
+
4. Train the full (base-256) model on LAION + COYO.
|
| 23 |
+
5. Train the full (base-512) model on LAION + COYO.
|
| 24 |
+
|
| 25 |
+
|
| 26 |
|
| 27 |
## Weekly Log
|
| 28 |
|