Clean legacy artifact as safetensors

README.md

@@ -0,0 +1,32 @@
+---
+library_name: safetensors
+tags:
+  - mamba
+  - sequence-model
+  - strategy-game
+  - 4x
+license: other
+---
+
+# 4x Mamba
+
+This is the legacy public 4X strategy-game Mamba checkpoint.
+
+The repository has been cleaned so the public artifact is weights-only:
+
+- `model.safetensors` - model weights only
+- `config.json` - minimal model/checkpoint metadata
+
+No optimizer state, scheduler state, training corpus, simulator source, AWS scripts, benchmark harness, or internal project files are included in the cleaned artifact.
+
+## Model
+
+- Architecture: legacy Mamba-style sequence model
+- Parameters: `27,648,512`
+- Original training metadata:
+  - Global step: `47,500`
+  - Epoch: `3`
+  - Validation loss: `0.7532`
+  - Validation accuracy: `0.8130`
+
+This is a research checkpoint trained around synthetic 4X strategy-game traces. It is not an official game integration.

config.json

@@ -0,0 +1,12 @@
+{
+  "architecture": "LegacyMambaModel",
+  "format": "state_dict",
+  "parameter_count": 27648512,
+  "source_checkpoint": "legacy_public_hf_checkpoint",
+  "training_metadata": {
+    "epoch": 3,
+    "global_step": 47500,
+    "val_acc": 0.8129544417492206,
+    "val_loss": 0.753193209497191
+  }
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b88dc52e39e230488264183c73a4d2ea22bed08dc7135ac734df952b5eb65d97
+size 110611296