Upload folder using huggingface_hub

README.md

@@ -4,18 +4,23 @@ tags:
 - dimensionality-reduction
 - classical
 - TRACERx
+- UMAP
+- PCA
 license: mit
 ---
 
-# CLASSICAL Model - samples mode - 32D
+# Classical Models (PCA + UMAP) - samples mode - 32D
 
-Pre-trained classical model for transcriptomic data compression.
+Pre-trained PCA and UMAP models for transcriptomic data compression.
+
+**UMAP models support transform()** - new data can be projected into the same embedding space.
 
 ## Details
 - **Mode**: samples-centric compression
 - **Dimensions**: 32
 - **Training data**: TRACERx lung cancer transcriptomics
-- **Created**: 2026-01-09T20:53:09.042685
+- **Created**: 2026-01-13T12:07:21.797423
+- **UMAP transform**: Enabled (low_memory=False)
 
 ## Usage
 
@@ -27,9 +32,8 @@ from huggingface_hub import snapshot_download
 local_dir = snapshot_download("jruffle/classical_samples_32d")
 model = joblib.load(f"{local_dir}/model.joblib")
 
-# For classical models (PCA/UMAP):
-# model contains: 'pca', 'umap', 'robust_scaler', 'gene_order'
+# Model contains: 'pca', 'umap', 'robust_scaler', 'gene_order'
 
-# For TabPFN models:
-# model contains: 'tabpfn_embedding', 'pca_final', 'input_scaler', etc.
+# Use UMAP transform on new data:
+new_embeddings = model['umap'].transform(preprocessed_new_data)
 ```

metadata.json

@@ -2,7 +2,8 @@
   "model_type": "classical",
   "mode": "samples",
   "dimensions": 32,
-  "created": "2026-01-09T20:53:09.042969",
+  "created": "2026-01-13T12:07:21.797812",
+  "umap_transform_enabled": true,
   "keys": [
     "robust_scaler",
     "norm_params",

model.joblib

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:62a3ff05206292c7fcc6a0ece6121e1250471c3cf641dcd2566c71580492c159
-size 90839614
+oid sha256:d1740e1307b697df82cb6bf21fb8f9e35635d79d66ccb12659cdf62acca1f063
+size 90758818