Update README.md
Browse files
README.md
CHANGED
|
@@ -7,17 +7,17 @@ This repository contains the dataset of model weights and utility files for [Int
|
|
| 7 |
The `files/` folder contains the files needed in our code [here](https://snap-research.github.io/weights2weights/) in order to conduct sampling, inversion, and editing in *weights2weights* space.
|
| 8 |
- `files/V.pt`
|
| 9 |
- 99648x10000 dimensional tensor used to project or unproject LoRA weights onto a principal component representation in *w2w* space or to unproject back into the LoRA space.
|
| 10 |
-
-
|
| 11 |
- 99648 dimensional tensor of the mean for each parameter in the original LoRA space. Used for projection/unprojection.
|
| 12 |
-
-
|
| 13 |
- 99648 dimensional tensor of the standard deviation for each parameter in the original LoRA space. Used for projection/unprojection.
|
| 14 |
-
-
|
| 15 |
- 64974x1000 dimensional tensor where each row is a 1000-dimensional principal component projection for each identity-encoding model in our dataset of model weights.
|
| 16 |
-
-
|
| 17 |
- Precomputed pseudoinverse of 'proj.pt', used for obtaining the classifier weight space directions given labels.
|
| 18 |
-
-
|
| 19 |
- A pandas dataframe with each identity model in our dataset of weights and their associated binary labels. Used for getting labels for training linear classifiers.
|
| 20 |
-
-
|
| 21 |
- A dictionary of the dimensionality for each LoRA module in the diffusion UNet. Used to save models in Diffusers pipeline format.
|
| 22 |
|
| 23 |
**Drag-Based Manipulation**
|
|
|
|
| 7 |
The `files/` folder contains the files needed in our code [here](https://snap-research.github.io/weights2weights/) in order to conduct sampling, inversion, and editing in *weights2weights* space.
|
| 8 |
- `files/V.pt`
|
| 9 |
- 99648x10000 dimensional tensor used to project or unproject LoRA weights onto a principal component representation in *w2w* space or to unproject back into the LoRA space.
|
| 10 |
+
- `files/mean.pt`
|
| 11 |
- 99648 dimensional tensor of the mean for each parameter in the original LoRA space. Used for projection/unprojection.
|
| 12 |
+
- `files/std.pt`
|
| 13 |
- 99648 dimensional tensor of the standard deviation for each parameter in the original LoRA space. Used for projection/unprojection.
|
| 14 |
+
- `files/proj_1000pc.pt`
|
| 15 |
- 64974x1000 dimensional tensor where each row is a 1000-dimensional principal component projection for each identity-encoding model in our dataset of model weights.
|
| 16 |
+
- `files/pinverse.pt`
|
| 17 |
- Precomputed pseudoinverse of 'proj.pt', used for obtaining the classifier weight space directions given labels.
|
| 18 |
+
- `files/identity_df.pt`
|
| 19 |
- A pandas dataframe with each identity model in our dataset of weights and their associated binary labels. Used for getting labels for training linear classifiers.
|
| 20 |
+
- `files/weight_dimensions.pt`
|
| 21 |
- A dictionary of the dimensionality for each LoRA module in the diffusion UNet. Used to save models in Diffusers pipeline format.
|
| 22 |
|
| 23 |
**Drag-Based Manipulation**
|