Model card: draft-zero is public (MIT); state honestly what playing the checkpoints needs

README.md

@@ -53,14 +53,20 @@ self-play card statistics to be trusted: premium removal plays at 45–46% again
 
 ## Using the checkpoints
 
-The checkpoints come from **MageZero 0.1.0**: upstream v0.1.0-alpha plus 7 fork commits,
-pinned at `git+https://github.com/danieljbrooks/MageZero@bcc76de`. **Whether they load
-under MageZero v0.2.0 is untested.** The model is MageZero's 2-layer transformer
-(d_model 512) with a dense feature vocabulary that grew during training. Load them with
-the same MageZero commit to be safe.
-
-The training harness is [draft-zero](https://github.com/danieljbrooks/draft-zero), tag
-`exp1-fdn-generalist`. The repository is currently private; ask the author for access.
+The model is MageZero's 2-layer transformer (d_model 512). Each checkpoint carries its own
+feature vocabulary. Playing games with them needs three things:
+
+| Needed | Where |
+|---|---|
+| MageZero **0.1.0**: upstream v0.1.0-alpha plus 7 fork commits | `pip install "magezero @ git+https://github.com/danieljbrooks/MageZero@bcc76de"` |
+| The **action vocabulary** the policy heads index into | [`assets/vocab/FDN_SPG.tsv`](https://github.com/danieljbrooks/draft-zero/blob/exp1-fdn-generalist/assets/vocab/FDN_SPG.tsv) in draft-zero; point `MZ_ACTION_VOCAB` at it |
+| The **generalist XMage build**: upstream XMage plus one commit that emits actions in that vocabulary | **Not published yet.** Coming as a public fork, and this card will link it. Until then the checkpoints can be inspected but not played. |
+
+Whether the checkpoints load under MageZero v0.2.0 is untested. None of this has been run
+end to end outside the training harness yet.
+
+The training harness is [draft-zero](https://github.com/danieljbrooks/draft-zero) (MIT), tag
+[`exp1-fdn-generalist`](https://github.com/danieljbrooks/draft-zero/tree/exp1-fdn-generalist).
 
 ## Data and license