Update README.md

README.md

@@ -13,8 +13,7 @@ checkpoint takes `(lat, lon)` in degrees and returns a 3,072-dimensional trunk w
 The first 64 dimensions are the default deployment embedding, and other leading prefixes can be selected without
 retraining the encoder.
 
-The model distills four teacher sources: AlphaEarth Foundations, Climplicit, GeoCLIP, and SINR. The released
-lat/lon-only checkpoint is time-invariant. MINDSET contains the cached teacher embeddings used for distillation.
+The model is distilled on the MINDSET dataset which is composed of four teacher sources: AlphaEarth Foundations, Climplicit, GeoCLIP, and SINR
 
 ## Files
 
@@ -22,8 +21,6 @@ lat/lon-only checkpoint is time-invariant. MINDSET contains the cached teacher e
 - `mind.pt` — fp32 PyTorch weights (453,967,275 bytes; pickle-based loading).
 - `mind.onnx` and `mind.onnx.data` — ONNX graph and external weights.
 - `mind.pt2` — PyTorch `ExportedProgram`.
-- `mind_small.safetensors` and `mind_small.pt2` — a 6.4M-parameter distilled student with a
-  128-dimensional output.
 
 ## Usage
 
@@ -33,13 +30,7 @@ The public ONNX release can be loaded directly from the Hub. Install `numpy`, `o
 import os
 import numpy as np
 import onnxruntime as ort
-from huggingface_hub import snapshot_download
 
-folder = snapshot_download(
-    repo_id="taylor-geospatial/MIND",
-    allow_patterns=["mind.onnx", "mind.onnx.data"],
-    token=False,
-)
 session = ort.InferenceSession(os.path.join(folder, "mind.onnx"), providers=["CPUExecutionProvider"])
 coordinates = np.array([[37.77, -122.42], [51.51, -0.13]], dtype=np.float32)  # (lat, lon)
 embedding = session.run(None, {"latlon": coordinates})[0]  # shape [2, 3072]
@@ -49,23 +40,3 @@ deploy_embedding = embedding[:, :64]
 The ONNX graph expects `latlon` with shape `[N, 2]` in `(lat, lon)` order and returns the full 3,072-dimensional trunk.
 Use the leading 64 columns as the deployment embedding. The safetensors and PyTorch files are also available for users
 who have a compatible loader.
-
-## MIND-small
-
-The repository contains `mind_small.safetensors` and `mind_small.pt2` for the 6.4M-parameter
-student. These files require a compatible loader. An ONNX export of MIND-small is not currently available.
-
-## Limitations
-
-MIND is a coordinate-only representation. It does not ingest current imagery, dates, or task labels at inference.
-Its behavior reflects the teacher models and the urban-dense training-coordinate sample in MINDSET. CoordBench scores
-depend on the target, label density, and spatial holdout rule; they do not predict performance for every downstream
-dataset.
-
-## Related releases
-
-- [MINDSET](https://huggingface.co/datasets/taylor-geospatial/MINDSET) — cached teacher embeddings.
-- [CoordBench](https://huggingface.co/datasets/taylor-geospatial/CoordBench) — common evaluation tables.
-- [MIND project page](https://research.taylorgeospatial.org/mind/) — project overview and release links.
-
-MIT licensed.