Add pipeline tag and improve metadata
Browse filesHi, I'm Niels from the community science team at Hugging Face. This PR adds the `text-generation` pipeline tag to the model metadata to help users discover it more easily. It also improves the model card layout by providing clear links to the paper, code, and project page, and adds a brief description of the GDDS framework.
README.md
CHANGED
|
@@ -1,26 +1,29 @@
|
|
| 1 |
---
|
| 2 |
-
license: mit
|
| 3 |
library_name: pytorch
|
|
|
|
|
|
|
| 4 |
tags:
|
| 5 |
-
|
| 6 |
-
|
| 7 |
-
|
| 8 |
-
|
| 9 |
-
|
| 10 |
---
|
| 11 |
|
| 12 |
# GDDS Checkpoints
|
| 13 |
|
| 14 |
Official checkpoint bundle for the paper **Generalized Discrete Diffusion from Snapshots**.
|
| 15 |
|
|
|
|
|
|
|
| 16 |
## Model Sources
|
| 17 |
|
| 18 |
-
- Paper: https://huggingface.co/papers/2603.21342
|
| 19 |
-
- arXiv: https://arxiv.org/abs/2603.21342
|
| 20 |
-
- Code: https://github.com/ozekri/gdds
|
| 21 |
-
- Project
|
| 22 |
|
| 23 |
-
## Included
|
| 24 |
|
| 25 |
| File | Method | Notes |
|
| 26 |
| --- | --- | --- |
|
|
@@ -35,8 +38,7 @@ Official checkpoint bundle for the paper **Generalized Discrete Diffusion from S
|
|
| 35 |
|
| 36 |
## Usage
|
| 37 |
|
| 38 |
-
These files are PyTorch Lightning checkpoints intended to be used with the
|
| 39 |
-
`gdds` codebase.
|
| 40 |
|
| 41 |
```bash
|
| 42 |
git clone https://github.com/ozekri/gdds.git
|
|
@@ -44,6 +46,7 @@ cd gdds
|
|
| 44 |
pip install -r requirements.txt
|
| 45 |
pip install -e .
|
| 46 |
|
|
|
|
| 47 |
PYTHONPATH=src python -m discrete_diffusion.evaluations.ppl_eval \
|
| 48 |
data=openwebtext \
|
| 49 |
model=small \
|
|
@@ -51,8 +54,7 @@ PYTHONPATH=src python -m discrete_diffusion.evaluations.ppl_eval \
|
|
| 51 |
eval.checkpoint_path=/path/to/checkpoints/mdlm_500k.ckpt
|
| 52 |
```
|
| 53 |
|
| 54 |
-
For sampling and other evaluations, use the same repository and pass the
|
| 55 |
-
relevant checkpoint path through the evaluation config.
|
| 56 |
|
| 57 |
## Citation
|
| 58 |
|
|
@@ -66,4 +68,4 @@ relevant checkpoint path through the evaluation config.
|
|
| 66 |
primaryClass={stat.ML},
|
| 67 |
url={https://arxiv.org/abs/2603.21342},
|
| 68 |
}
|
| 69 |
-
```
|
|
|
|
| 1 |
---
|
|
|
|
| 2 |
library_name: pytorch
|
| 3 |
+
license: mit
|
| 4 |
+
pipeline_tag: text-generation
|
| 5 |
tags:
|
| 6 |
+
- gdds
|
| 7 |
+
- discrete-diffusion
|
| 8 |
+
- language-modeling
|
| 9 |
+
- research
|
| 10 |
+
- pytorch
|
| 11 |
---
|
| 12 |
|
| 13 |
# GDDS Checkpoints
|
| 14 |
|
| 15 |
Official checkpoint bundle for the paper **Generalized Discrete Diffusion from Snapshots**.
|
| 16 |
|
| 17 |
+
Generalized Discrete Diffusion from Snapshots (GDDS) is a unified framework for discrete diffusion modeling that supports arbitrary noising processes over large discrete state spaces. It introduces a training objective based on snapshot latents rather than the entire noising path, allowing for efficient training and high-quality generation.
|
| 18 |
+
|
| 19 |
## Model Sources
|
| 20 |
|
| 21 |
+
- **Paper:** [Generalized Discrete Diffusion from Snapshots](https://huggingface.co/papers/2603.21342)
|
| 22 |
+
- **arXiv:** [2603.21342](https://arxiv.org/abs/2603.21342)
|
| 23 |
+
- **Code:** [GitHub - ozekri/gdds](https://github.com/ozekri/gdds)
|
| 24 |
+
- **Project Page:** [https://oussamazekri.fr/gdds](https://oussamazekri.fr/gdds)
|
| 25 |
|
| 26 |
+
## Included Checkpoints
|
| 27 |
|
| 28 |
| File | Method | Notes |
|
| 29 |
| --- | --- | --- |
|
|
|
|
| 38 |
|
| 39 |
## Usage
|
| 40 |
|
| 41 |
+
These files are PyTorch Lightning checkpoints intended to be used with the [`gdds`](https://github.com/ozekri/gdds) codebase.
|
|
|
|
| 42 |
|
| 43 |
```bash
|
| 44 |
git clone https://github.com/ozekri/gdds.git
|
|
|
|
| 46 |
pip install -r requirements.txt
|
| 47 |
pip install -e .
|
| 48 |
|
| 49 |
+
# Example evaluation using a checkpoint
|
| 50 |
PYTHONPATH=src python -m discrete_diffusion.evaluations.ppl_eval \
|
| 51 |
data=openwebtext \
|
| 52 |
model=small \
|
|
|
|
| 54 |
eval.checkpoint_path=/path/to/checkpoints/mdlm_500k.ckpt
|
| 55 |
```
|
| 56 |
|
| 57 |
+
For sampling and other evaluations, use the same repository and pass the relevant checkpoint path through the Hydra evaluation config.
|
|
|
|
| 58 |
|
| 59 |
## Citation
|
| 60 |
|
|
|
|
| 68 |
primaryClass={stat.ML},
|
| 69 |
url={https://arxiv.org/abs/2603.21342},
|
| 70 |
}
|
| 71 |
+
```
|