From 79c109613de83dcb8d3eb4fe9d2c88b8c591ee8f Mon Sep 17 00:00:00 2001
From: ryan <rpotter6298@gmail.com>
Date: Sat, 15 Aug 2026 08:45:00 +0200
Subject: [PATCH] Add a bundle-backed config: dataset as smoke.hdb via git-lfs
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Exercises the provider bundle path — verify sha256, unpack to the
content-addressed cache, symlink in at mount="data" so the config metadata_csv
path resolves. The loose-file config stays alongside it as the plain case.
---
 .gitattributes             |   4 ++
 configs/tabular_bundle.hcf |  91 +++++++++++++++++++++++++++++++++++++
 smoke.hdb                  | Bin 0 -> 129 bytes
 3 files changed, 95 insertions(+)
 create mode 100644 .gitattributes
 create mode 100644 configs/tabular_bundle.hcf
 create mode 100644 smoke.hdb

diff --git a/.gitattributes b/.gitattributes
new file mode 100644
index 0000000..bfc1f66
--- /dev/null
+++ b/.gitattributes
@@ -0,0 +1,4 @@
+*.hdb filter=lfs diff=lfs merge=lfs -text
+*.hrb filter=lfs diff=lfs merge=lfs -text
+*.hdb binary
+*.hrb binary
diff --git a/configs/tabular_bundle.hcf b/configs/tabular_bundle.hcf
new file mode 100644
index 0000000..9015721
--- /dev/null
+++ b/configs/tabular_bundle.hcf
@@ -0,0 +1,91 @@
+{
+  "_notes": [
+    "Bundle-backed variant of tabular_smoke: the dataset arrives as smoke.hdb",
+    "rather than as loose files. mount='data' places the bundle's smoke/labels.csv",
+    "at data/smoke/labels.csv, which is what metadata_csv below expects."
+  ],
+  "run_name": "tabular_bundle",
+  "num_classes": 2,
+  "split_identity_level": 1,
+  "eval_stage": "cd_fuse",
+  "save_predictions": true,
+  "seed": 1234,
+  "folds": 3,
+  "fold_seed": 100,
+  "output_root": "results",
+  "data": {
+    "module": "hypertower_core.profiles.generic",
+    "args": {
+      "metadata_csv": "data/smoke/labels.csv",
+      "id_column": "sample_id",
+      "label_col": "diagnosis",
+      "target_type": "classification",
+      "group_column": "patient_id",
+      "cat_cols": [
+        "site"
+      ],
+      "exclude_cols": [
+        "notes"
+      ]
+    }
+  },
+  "towers": [
+    {
+      "name": "cd",
+      "module": "hypertower_core.components.towers.clinical_tower",
+      "class": "ClinicalEncoder",
+      "data_source": "matrix",
+      "args": {
+        "hidden_dim": 64
+      }
+    }
+  ],
+  "stages": [
+    {
+      "name": "cd_warm",
+      "type": "warm",
+      "tower": "cd",
+      "head_name": "cd_aux",
+      "level": "sample",
+      "epochs": 8
+    },
+    {
+      "name": "cd_aux",
+      "type": "head",
+      "input": "cd",
+      "train_with": "cd_fuse"
+    },
+    {
+      "name": "cd_fuse",
+      "type": "fusion",
+      "module": "hypertower_core.components.bridges.mono_bridge",
+      "class": "MonoBridge",
+      "inputs": [
+        "cd"
+      ],
+      "level": "sample",
+      "epochs": 10,
+      "train_towers": true,
+      "args": {
+        "use_ln": false
+      }
+    }
+  ],
+  "training": {
+    "lr": 0.001,
+    "batch_size": 32,
+    "tune_binary_threshold": true
+  },
+  "data_bundles": [
+    {
+      "file": "smoke.hdb",
+      "encrypted": false,
+      "mount": "data",
+      "unpacked_mb": 1,
+      "sha256": "b01dbf19fe3639a1b75fb80d75b31158f78dc42bec3767fe1caf94afd5b8fda5"
+    }
+  ],
+  "dataset": {
+    "n_samples": 400
+  }
+}
\ No newline at end of file
diff --git a/smoke.hdb b/smoke.hdb
new file mode 100644
index 0000000000000000000000000000000000000000..e69442c5a34434f5d088fddfb74a5dfc7eeb24c7
GIT binary patch
literal 129
zcmWN?OA^8$3;@tQr{DsXrVu{84FMv|sB}#2!qe;9ysN!s%$M%xdB|?eeVn%k%ksZ}
zXesk)<Y>iQpysGrdfw3Qv_d6@#Laq}GEhLKxYmGV0WfliwYZI1vB3v!P%<Z%xgtfb
M8I_E0(*dnhKjOD2UH||9

literal 0
HcmV?d00001