add PDB → features quickstart, link to release
Browse files
README.md
CHANGED
|
@@ -15,10 +15,10 @@ Part of the [k-fold-structure release](https://github.com/<TODO-github-org>/k-fo
|
|
| 15 |
|
| 16 |
## Files
|
| 17 |
|
| 18 |
-
- `35M.ckpt`
|
| 19 |
-
- `config.yaml`
|
| 20 |
-
- `backbone_tokenizer.pt`
|
| 21 |
-
- `fullatom_tokenizer.pt`
|
| 22 |
|
| 23 |
## Quickstart
|
| 24 |
|
|
@@ -29,7 +29,7 @@ git clone https://github.com/<TODO-github-org>/k-fold-structure-release.git
|
|
| 29 |
cd k-fold-structure-release
|
| 30 |
```
|
| 31 |
|
| 32 |
-
### Option A
|
| 33 |
|
| 34 |
```python
|
| 35 |
import sys
|
|
@@ -47,7 +47,7 @@ print(features.shape)
|
|
| 47 |
runs PDB → (seq, bb, fa) tokens through the tokenizer VQ-VAEs, then forwards
|
| 48 |
through the encoder.
|
| 49 |
|
| 50 |
-
### Option B
|
| 51 |
|
| 52 |
If you already have token IDs (e.g., from `k-fold-structure/repsp-triprorep-tokens`):
|
| 53 |
|
|
|
|
| 15 |
|
| 16 |
## Files
|
| 17 |
|
| 18 |
+
- `35M.ckpt`: full Lightning checkpoint (for re-training / fine-tuning)
|
| 19 |
+
- `config.yaml`: model + data config (sanitized: replace `/path/to/...` with your local paths)
|
| 20 |
+
- `backbone_tokenizer.pt`: backbone (bb) token VQ-VAE (from [StructTokenBench](https://github.com/KatarinaYuan/StructTokenBench))
|
| 21 |
+
- `fullatom_tokenizer.pt`: full-atom (fa) token VQ-VAE
|
| 22 |
|
| 23 |
## Quickstart
|
| 24 |
|
|
|
|
| 29 |
cd k-fold-structure-release
|
| 30 |
```
|
| 31 |
|
| 32 |
+
### Option A: PDB → per-residue features (one-shot)
|
| 33 |
|
| 34 |
```python
|
| 35 |
import sys
|
|
|
|
| 47 |
runs PDB → (seq, bb, fa) tokens through the tokenizer VQ-VAEs, then forwards
|
| 48 |
through the encoder.
|
| 49 |
|
| 50 |
+
### Option B: token IDs → features (pre-tokenized)
|
| 51 |
|
| 52 |
If you already have token IDs (e.g., from `k-fold-structure/repsp-triprorep-tokens`):
|
| 53 |
|