Add causal and bidirectional context configs

README.md

@@ -11,14 +11,22 @@ tags:
 - hg38
 - parquet
 configs:
-- config_name: canonical
+- config_name: canonical-causal
   data_files:
   - split: all
-    path: functional-upstream-gamba.parquet
-- config_name: noncoding-added
+    path: functional-upstream-gamba-causal.parquet
+- config_name: canonical-bidi
   data_files:
   - split: all
-    path: functional-upstream-gamba-noncoding-added.parquet
+    path: functional-upstream-gamba-bidi.parquet
+- config_name: noncoding-added-causal
+  data_files:
+  - split: all
+    path: functional-upstream-gamba-noncoding-added-causal.parquet
+- config_name: noncoding-added-bidi
+  data_files:
+  - split: all
+    path: functional-upstream-gamba-noncoding-added-bidi.parquet
 ---
 
 # GAMBA Functional Regions: Feature vs Upstream
@@ -32,13 +40,23 @@ not overlap retained anchors from the same category.
 
 | Config | Train rows | Test rows | Total |
 |---|---:|---:|---:|
-| `canonical` | 135,692 | 33,982 | 169,674 |
-| `noncoding-added` | 150,726 | 37,834 | 188,560 |
+| `canonical-causal`, `canonical-bidi` | 135,692 | 33,982 | 169,674 |
+| `noncoding-added-causal`, `noncoding-added-bidi` | 150,726 | 37,834 | 188,560 |
 
-The `noncoding-added` config is an explicit extension from
+The `noncoding-added-*` configs are an explicit extension from
 [Gamba-Processing](https://github.com/TaykhoomDalal/Gamba-Processing), not an
 additional category emitted by GAMBA's current `create_eval_data.py`.
 
+## Choosing a context
+
+The GAMBA paper uses context geometry matched to the model. Choose
+`*-causal` for Evo2 or another left-to-right model: the ROI is at the end of
+the strand-oriented window, so all preceding bases are usable context. Choose
+`*-bidi` for GAMBA encoders, the distilled student, GPN-Star, PhyloGPN, or
+another masked/bidirectional model: the ROI is centered so the model sees both
+flanks. Using a causal file for a bidirectional model removes its right-side
+context and is not the paper protocol.
+
 ## Loading
 
 ```python
@@ -46,7 +64,7 @@ from datasets import load_dataset
 
 canonical = load_dataset(
     "Taykhoom/functional-upstream-gamba",
-    "canonical",
+    "canonical-bidi",
     split="all",
 )
 train = canonical.filter(lambda row: row["split"] == "train")
@@ -63,7 +81,8 @@ the column as shown above.
 | Column | Description |
 |---|---|
 | `split`, `label` | Probe/fine-tuning `train` or held-out `test`; labels are `feature` or `upstream`. |
-| `sequence` | Strand-oriented GAMBA context of at most 2,048 bp. |
+| `sequence` | Strand-oriented causal or bidirectional context of at most 2,048 bp. |
+| `context_policy` | `causal` for `*-causal`; `symmetric` for `*-bidi`. |
 | `pair_id` | Shared identifier for a feature and its matched upstream control. |
 | `category`, `scope` | Functional class and pooling scope. |
 | `chrom`, `start`, `end`, `strand` | Zero-based, half-open hg38 feature/control coordinates. |

functional-upstream-gamba-bidi.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e2a85348a801d0553062910aec012f29838c780ef605c9a49eb36fa74b6ac05a
+size 119330690

functional-upstream-gamba-causal.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1cca1a4e047e3420be28a85c31d74137470ce75abd5c4e5e59ef49545a607e2d
+size 114548467

functional-upstream-gamba-noncoding-added-bidi.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8ce0957ef79ffefecd5409527c5905846a986d76f8ad94092a9652d52637ca85
+size 132422261

functional-upstream-gamba-noncoding-added-causal.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:242066fed5369bb6f5471465db3ce2cc37d4d01139195f45e3f4a0c5a1606010
+size 124940309