ZeroCool94
commited on
Commit
•
7699903
1
Parent(s):
23b65e4
Update README.md
Browse files
README.md
CHANGED
@@ -39,11 +39,13 @@ This model is still in its infancy and it's meant to be constantly updated and t
|
|
39 |
|
40 |
## Available Checkpoints:
|
41 |
- #### Stable:
|
|
|
|
|
42 |
- [vae.sygil_muse_v0.1.pt](https://huggingface.co/Sygil/Sygil-Muse/blob/main/vae.sygil_muse_v0.1.pt): Trained from scratch for 3.0M steps with **dim: 128** and **vq_codebook_size: 256**.
|
43 |
- [maskgit.sygil_muse_v0.1.pt](https://huggingface.co/Sygil/Sygil-Muse/blob/main/maskgit.sygil_muse_v0.1.pt): Maskgit trained from the VAE for 3.46M steps
|
44 |
- [vae.sygil_muse_v0.5.pt](https://huggingface.co/Sygil/Sygil-Muse/blob/main/vae.sygil_muse_v0.5.pt): Trained from scratch for 1.99M steps with **dim: 128** and **vq_codebook_size: 8192**.
|
45 |
- #### Beta:
|
46 |
-
- [vae.
|
47 |
- [maskgit.39000.pt](https://huggingface.co/Sygil/Sygil-Muse/blob/main/maskgit.39000.pt): Maskgit trained from the VAE for 39K steps using the hyperparameters `heads 16` and `depth 22` for testing, these values have huge performance effects, the vram usage was also increased so it is just for testing, the quality on this checkpoint did increase a lot and requires a lot less training which is something we want but we need to find a balance between quality and performance.
|
48 |
|
49 |
Note: Checkpoints under the Beta section are updated daily or at least 3-4 times a week. While the beta checkpoints can be used as they are only the latest version is kept on the repo and the older checkpoints are removed when a new one
|
@@ -73,7 +75,7 @@ The model was trained on the following dataset:
|
|
73 |
- **heads:** 8
|
74 |
- **depth:** 4
|
75 |
- **Random Crop:** True
|
76 |
-
- **Total Training Steps:**
|
77 |
|
78 |
Note: On Muse we can change the image_size or resolution at any time without having to train the model from scratch again, this allows us to first train the model at low resolution using the same `dim` and `vq_codebook_size` to train faster and then we can increase the `image_size` and use a higher resolution once the model has trained enough.
|
79 |
|
|
|
39 |
|
40 |
## Available Checkpoints:
|
41 |
- #### Stable:
|
42 |
+
- None
|
43 |
+
- #### Old Checkpoints:
|
44 |
- [vae.sygil_muse_v0.1.pt](https://huggingface.co/Sygil/Sygil-Muse/blob/main/vae.sygil_muse_v0.1.pt): Trained from scratch for 3.0M steps with **dim: 128** and **vq_codebook_size: 256**.
|
45 |
- [maskgit.sygil_muse_v0.1.pt](https://huggingface.co/Sygil/Sygil-Muse/blob/main/maskgit.sygil_muse_v0.1.pt): Maskgit trained from the VAE for 3.46M steps
|
46 |
- [vae.sygil_muse_v0.5.pt](https://huggingface.co/Sygil/Sygil-Muse/blob/main/vae.sygil_muse_v0.5.pt): Trained from scratch for 1.99M steps with **dim: 128** and **vq_codebook_size: 8192**.
|
47 |
- #### Beta:
|
48 |
+
- [vae.195500.pt](https://huggingface.co/Sygil/Sygil-Muse/blob/main/vae.195500.pt): Trained from scratch for 195K steps and higher **vq_codebook_dim** and **vq_codebook_size** than before.
|
49 |
- [maskgit.39000.pt](https://huggingface.co/Sygil/Sygil-Muse/blob/main/maskgit.39000.pt): Maskgit trained from the VAE for 39K steps using the hyperparameters `heads 16` and `depth 22` for testing, these values have huge performance effects, the vram usage was also increased so it is just for testing, the quality on this checkpoint did increase a lot and requires a lot less training which is something we want but we need to find a balance between quality and performance.
|
50 |
|
51 |
Note: Checkpoints under the Beta section are updated daily or at least 3-4 times a week. While the beta checkpoints can be used as they are only the latest version is kept on the repo and the older checkpoints are removed when a new one
|
|
|
75 |
- **heads:** 8
|
76 |
- **depth:** 4
|
77 |
- **Random Crop:** True
|
78 |
+
- **Total Training Steps:** 195,500
|
79 |
|
80 |
Note: On Muse we can change the image_size or resolution at any time without having to train the model from scratch again, this allows us to first train the model at low resolution using the same `dim` and `vq_codebook_size` to train faster and then we can increase the `image_size` and use a higher resolution once the model has trained enough.
|
81 |
|