Add umbrella README
Browse files
README.md
ADDED
|
@@ -0,0 +1,66 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
tags:
|
| 4 |
+
- histopathology
|
| 5 |
+
- diffusion
|
| 6 |
+
- spatial-transcriptomics
|
| 7 |
+
- icml-2026-sd4h-workshop
|
| 8 |
+
---
|
| 9 |
+
|
| 10 |
+
# STMDiT — Inference Checkpoints
|
| 11 |
+
|
| 12 |
+
EMA-only inference weights for every model row reported in the ICML 2026 SD4H
|
| 13 |
+
workshop submission *Transcriptomics-Conditioned Virtual Tissue Synthesis via
|
| 14 |
+
Diffusion Transformers*.
|
| 15 |
+
|
| 16 |
+
Each subfolder contains a single model:
|
| 17 |
+
- `model.pt` — EMA-only state dict (PyTorch `.pt`)
|
| 18 |
+
- `training_config.yaml` — the original training YAML
|
| 19 |
+
- `README.md` — per-model card
|
| 20 |
+
|
| 21 |
+
## Models
|
| 22 |
+
|
| 23 |
+
| `row_id` | Paper label |
|
| 24 |
+
|----------|-------------|
|
| 25 |
+
| `pixcell-b` | PixCell-B |
|
| 26 |
+
| `pixcell-flow-b` | PixCell-Flow-B |
|
| 27 |
+
| `adaln-ddpm-p01` | PixCell-GE-B (p=0.1) |
|
| 28 |
+
| `adaln-ddpm-p02` | PixCell-GE-B-p02 |
|
| 29 |
+
| `adaln-ddpm-p03` | PixCell-GE-B-p03 |
|
| 30 |
+
| `adaln-ddpm-p05` | PixCell-GE-B-p05 |
|
| 31 |
+
| `adaln-ddpm-p06` | PixCell-GE-B-p06 |
|
| 32 |
+
| `adaln-flow-p01` | PixCell-Flow-GE-B (p=0.1) |
|
| 33 |
+
| `adaln-flow-p02` | PixCell-Flow-GE-B-p02 |
|
| 34 |
+
| `adaln-flow-p03` | PixCell-Flow-GE-B-p03 |
|
| 35 |
+
| `adaln-flow-p05` | PixCell-Flow-GE-B-p05 |
|
| 36 |
+
| `xattn-direct-p01` | XAttn-Direct (p=0.1) |
|
| 37 |
+
| `xattn-gsa-p01` | XAttn-GSA (p=0.1) |
|
| 38 |
+
| `xattn-perceiver-p01` | XAttn-Perceiver (p=0.1) |
|
| 39 |
+
| `xattn-pma-p01` | XAttn-PMA (p=0.1) |
|
| 40 |
+
| `xattn-perceiver-p05` | XAttn-Perceiver-p05 |
|
| 41 |
+
| `xattn-perceiver-p06` | XAttn-Perceiver-p06 |
|
| 42 |
+
| `xattn-pma-p05` | XAttn-PMA-p05 |
|
| 43 |
+
| `xattn-pma-p06` | XAttn-PMA-p06 |
|
| 44 |
+
| `ptpl-adaln-p05` | PTPL-AdaLN-B (p=0.5) |
|
| 45 |
+
| `ptpl-adaln-p06` | PTPL-AdaLN-B-p06 |
|
| 46 |
+
| `ptpl-adaln-p07` | PTPL-AdaLN-B-p07 |
|
| 47 |
+
| `ptpl-xattn-perceiver-p05` | PTPL-XAttn-Perceiver-B (p=0.5) |
|
| 48 |
+
| `ptpl-xattn-perceiver-p06` | PTPL-XAttn-Perceiver-B-p06 |
|
| 49 |
+
| `ptpl-xattn-perceiver-p07` | PTPL-XAttn-Perceiver-B-p07 |
|
| 50 |
+
| `ptpl-xattn-pma-p05` | PTPL-XAttn-PMA-B (p=0.5) |
|
| 51 |
+
| `ptpl-xattn-pma-p06` | PTPL-XAttn-PMA-B-p06 |
|
| 52 |
+
| `ptpl-xattn-pma-p07` | PTPL-XAttn-PMA-B-p07 |
|
| 53 |
+
|
| 54 |
+
## Usage
|
| 55 |
+
|
| 56 |
+
```python
|
| 57 |
+
from huggingface_hub import snapshot_download
|
| 58 |
+
ckpt_dir = snapshot_download(
|
| 59 |
+
repo_id="stmdit-anon/stmdit-checkpoints",
|
| 60 |
+
allow_patterns="xattn-perceiver-p05/*", # pick one model by row_id
|
| 61 |
+
)
|
| 62 |
+
# ckpt_dir / "xattn-perceiver-p05" / "model.pt"
|
| 63 |
+
```
|
| 64 |
+
|
| 65 |
+
See the public anonymized code repo (linked from the OpenReview submission) for
|
| 66 |
+
the loader, demo notebook, and end-to-end inference example.
|