Add pipeline tag and link to original Relational Transformer paper
#1
by nielsr HF Staff - opened
README.md
CHANGED
|
@@ -1,6 +1,12 @@
|
|
| 1 |
---
|
| 2 |
-
|
|
|
|
| 3 |
library_name: pytorch
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 4 |
tags:
|
| 5 |
- relational-data
|
| 6 |
- tabular
|
|
@@ -9,30 +15,34 @@ tags:
|
|
| 9 |
- relational-transformer
|
| 10 |
- relbench
|
| 11 |
- synthetic-data
|
| 12 |
-
datasets:
|
| 13 |
-
- kvignesh1420/plurel
|
| 14 |
-
metrics:
|
| 15 |
-
- roc_auc
|
| 16 |
-
- r_squared
|
| 17 |
---
|
| 18 |
|
| 19 |
# Relational Transformer — PluRel Checkpoints
|
| 20 |
|
| 21 |
-
Relational Transformer (RT) model checkpoints pretrained on synthetic relational databases generated by [PluRel](https://huggingface.co/datasets/kvignesh1420/plurel)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 22 |
|
| 23 |
-
> **PluRel: Synthetic Data unlocks Scaling Laws for Relational Foundation Models**
|
| 24 |
-
> Kothapalli, Ranjan, Hudovernik, Dwivedi, Hoffart, Guestrin, Leskovec — arXiv:2602.04029 (2026)
|
| 25 |
|
| 26 |
-
[](https://huggingface.co/datasets/kvignesh1420/plurel)
|
| 30 |
|
| 31 |
---
|
| 32 |
|
| 33 |
## Model Architecture
|
| 34 |
|
| 35 |
-
The Relational Transformer operates on multi-tabular relational databases, treating rows across linked tables as a sequence via BFS-ordered context sampling.
|
| 36 |
|
| 37 |
| Hyperparameter | Value |
|
| 38 |
|----------------|-------|
|
|
@@ -44,7 +54,7 @@ The Relational Transformer operates on multi-tabular relational databases, treat
|
|
| 44 |
| Text encoder | `all-MiniLM-L12-v2` (d_text = 384) |
|
| 45 |
| Max BFS width | 128 |
|
| 46 |
|
| 47 |
-
The architecture and training loop build on the [Relational Transformer
|
| 48 |
|
| 49 |
---
|
| 50 |
|
|
@@ -53,4 +63,5 @@ The architecture and training loop build on the [Relational Transformer (ICLR 20
|
|
| 53 |
```bash
|
| 54 |
huggingface-cli download kvignesh1420/relational-transformer-plurel \
|
| 55 |
--repo-type model \
|
| 56 |
-
--local-dir ~/scratch/rt_hf_ckpts
|
|
|
|
|
|
| 1 |
---
|
| 2 |
+
datasets:
|
| 3 |
+
- kvignesh1420/plurel
|
| 4 |
library_name: pytorch
|
| 5 |
+
license: mit
|
| 6 |
+
pipeline_tag: other
|
| 7 |
+
metrics:
|
| 8 |
+
- roc_auc
|
| 9 |
+
- r_squared
|
| 10 |
tags:
|
| 11 |
- relational-data
|
| 12 |
- tabular
|
|
|
|
| 15 |
- relational-transformer
|
| 16 |
- relbench
|
| 17 |
- synthetic-data
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 18 |
---
|
| 19 |
|
| 20 |
# Relational Transformer — PluRel Checkpoints
|
| 21 |
|
| 22 |
+
Relational Transformer (RT) model checkpoints pretrained on synthetic relational databases generated by [PluRel](https://huggingface.co/datasets/kvignesh1420/plurel).
|
| 23 |
+
|
| 24 |
+
Relational Transformer is a foundation model architecture for relational data that enables zero-shot transfer across heterogeneous schemas and tasks. It was introduced in:
|
| 25 |
+
|
| 26 |
+
> **Relational Transformer: Toward Zero-Shot Foundation Models for Relational Data**
|
| 27 |
+
> Rishabh Ranjan, Valter Hudovernik, Mark Znidar, Charilaos Kanatsoulis, Roshan Upendra, Mahmoud Mohammadi, Joe Meyer, Tom Palczewski, Carlos Guestrin, Jure Leskovec — [arXiv:2510.06377](https://arxiv.org/abs/2510.06377) (ICLR 2026)
|
| 28 |
+
|
| 29 |
+
The checkpoints provided in this repository were trained using the methodology described in:
|
| 30 |
|
| 31 |
+
> **PluRel: Synthetic Data unlocks Scaling Laws for Relational Foundation Models**
|
| 32 |
+
> Kothapalli, Ranjan, Hudovernik, Dwivedi, Hoffart, Guestrin, Leskovec — [arXiv:2602.04029](https://arxiv.org/abs/2602.04029) (2026)
|
| 33 |
|
| 34 |
+
[](https://arxiv.org/abs/2510.06377)
|
| 35 |
+
[](https://github.com/snap-stanford/relational-transformer)
|
| 36 |
+
[](https://arxiv.org/abs/2602.04029)
|
| 37 |
+
[](https://snap-stanford.github.io/plurel/)
|
| 38 |
+
[](https://github.com/snap-stanford/plurel)
|
| 39 |
[](https://huggingface.co/datasets/kvignesh1420/plurel)
|
| 40 |
|
| 41 |
---
|
| 42 |
|
| 43 |
## Model Architecture
|
| 44 |
|
| 45 |
+
The Relational Transformer operates on multi-tabular relational databases, treating rows across linked tables as a sequence via BFS-ordered context sampling. It utilizes a Relational Attention mechanism over columns, rows, and primary-foreign key links.
|
| 46 |
|
| 47 |
| Hyperparameter | Value |
|
| 48 |
|----------------|-------|
|
|
|
|
| 54 |
| Text encoder | `all-MiniLM-L12-v2` (d_text = 384) |
|
| 55 |
| Max BFS width | 128 |
|
| 56 |
|
| 57 |
+
The architecture and training loop build on the [Relational Transformer](https://github.com/snap-stanford/relational-transformer) codebase.
|
| 58 |
|
| 59 |
---
|
| 60 |
|
|
|
|
| 63 |
```bash
|
| 64 |
huggingface-cli download kvignesh1420/relational-transformer-plurel \
|
| 65 |
--repo-type model \
|
| 66 |
+
--local-dir ~/scratch/rt_hf_ckpts
|
| 67 |
+
```
|