Use the published GitHub and Dataset paths

README.md

@@ -69,7 +69,8 @@ a forecasting result.
 
 ## Use
 
-Install the code and download the checkpoint:
+Install the published code, then download the current Encoder and Dataset
+trees into the runtime locations used by that code:
 
 ```bash
 git clone https://github.com/risenyard/egms-qa
@@ -78,21 +79,30 @@ pip install -e .
 
 hf download risenyard/egms-qa-encoder \
     --local-dir data/encoder/checkpoint
+hf download risenyard/egms-qa-dataset \
+    --repo-type dataset --local-dir release/egms-qa-dataset
+python -m egms_qa.release audit \
+    --release-dir release/egms-qa-dataset
+python -m egms_qa.release install \
+    --release-dir release/egms-qa-dataset --target-root .
 ```
 
-Extract tokens from an already prepared compatible NPZ tile collection:
+Extract tokens from the released tiles:
 
 ```bash
 python -m egms_encoder.extract_tokens \
     --checkpoint data/encoder/checkpoint/encoder.safetensors \
     --model-config data/encoder/checkpoint/config.json \
-    --manifest path/to/split.parquet \
-    --data-config path/to/data_config.json \
+    --normalization data/encoder/checkpoint/normalization.json \
+    --manifest data/encoder/manifest/split.parquet \
+    --data-config data/encoder/manifest/data_config.json \
     --output-dir outputs/tokens
 ```
 
-The manifest and data config describe the user's tiles; they are not properties
-of the checkpoint. In the released dataset, stored `[0,294)` is bitwise equal
+For another compatible NPZ collection, replace only `--manifest` and
+`--data-config` with that collection's contract paths. These files describe the
+tiles and are not properties of the checkpoint. In the released dataset,
+stored `[0,294)` is bitwise equal
 to `[8,302)` on the original 304-step prepared axis; the data config retains
 the original offset for physical-time calculations. The released training data
 and reference manifest are available separately in