Correct logical model size information

Document the logical target parameter count and, where needed, correct the Safetensors index metadata so packed quantized storage is not mistaken for model size. Weight files and shard weight maps are unchanged.

README.md

@@ -18,9 +18,13 @@ tags:
 
 # AX DiffusionGemma 26B-A4B IT MLX 4-bit
 
+> **Parameter count:** approximately 25.82B logical parameters (26B total,
+> approximately 4B active per token). `4-bit` is the quantization precision,
+> not a 4B model-size claim.
+
 > **Transparent upstream mirror — AutomatosX made no model changes.**
 
-This is a revision-pinned, byte-exact mirror of
+This is a revision-pinned, transparent mirror of
 [mlx-community/diffusiongemma-26B-A4B-it-4bit](https://huggingface.co/mlx-community/diffusiongemma-26B-A4B-it-4bit)
 at commit `a7a81407613811e8ba63af92ac0d852b809e191f`.
 
@@ -29,7 +33,9 @@ otherwise alter the weights, configuration, tokenizer, processor, chat
 template, or generation settings. The MLX conversion and quantization were
 performed upstream by MLX Community with `mlx-vlm` 0.6.3. AutomatosX adds only
 this mirror documentation, the preserved upstream model card, a copy of the
-declared license, and machine-readable provenance.
+declared license, machine-readable provenance, and a `total_parameters` field
+in the shard index so the Hub does not mistake packed 4-bit storage for the
+logical model size.
 
 ## Model details
 
@@ -80,8 +86,10 @@ ordinary autoregressive TTFT and decode-throughput comparisons do not apply.
 
 `UPSTREAM_README.md` preserves the original MLX Community model card.
 `ax_provenance.json` pins the source commit and records SHA-256 values and sizes
-for every mirrored artifact. All source artifacts other than the replaced model
-card are byte-identical to the pinned upstream revision.
+for every mirrored artifact. All weight shards and all source artifacts other
+than the replaced model card and shard index are byte-identical to the pinned
+upstream revision. In the index, `weight_map` and `total_size` remain unchanged;
+only the logical `total_parameters` metadata field was added.
 
 This repository exists for revision-pinned cataloging and convenient discovery;
 it does not represent a new model, conversion, quantization, or benchmark claim

ax_provenance.json

@@ -6,7 +6,9 @@
   "source": {
     "repo_id": "mlx-community/diffusiongemma-26B-A4B-it-4bit",
     "revision": "a7a81407613811e8ba63af92ac0d852b809e191f",
-    "model_artifacts_byte_exact": true,
+    "model_artifacts_byte_exact": false,
+    "weight_shards_byte_exact": true,
+    "index_weight_map_byte_exact": true,
     "upstream_readme_preserved_as": "UPSTREAM_README.md",
     "upstream_readme_sha256": "ad3262ee1b00cc856307a469e710991295712fee20480789aa522152785866c8"
   },
@@ -21,7 +23,8 @@
     "merged": false,
     "converted": false,
     "requantized": false,
-    "mtp_added": false
+    "mtp_added": false,
+    "index_metadata_modified": true
   },
   "quantization": {
     "format": "mlx",
@@ -49,8 +52,17 @@
     "README.md",
     "UPSTREAM_README.md",
     "LICENSE",
-    "ax_provenance.json"
+    "ax_provenance.json",
+    "model.safetensors.index.json:metadata.total_parameters"
   ],
+  "destination_index": {
+    "path": "model.safetensors.index.json",
+    "size_bytes": 165404,
+    "sha256": "2b9d39e640173a25acd648ef4d0e98ae4cd06e90b6d6e0b4c0f22111491835d3",
+    "total_parameters": 25823778864,
+    "source_weight_map_unchanged": true,
+    "source_total_size_unchanged": true
+  },
   "excluded_local_only_files": [],
   "mtp": false
 }

model.safetensors.index.json

@@ -1,6 +1,7 @@
 {
     "metadata": {
-        "total_size": 16542844632
+        "total_size": 16542844632,
+        "total_parameters": 25823778864
     },
     "weight_map": {
         "model.decoder.embed_tokens.biases": "model-00001-of-00004.safetensors",
@@ -1651,4 +1652,4 @@
         "model.encoder.vision_tower.std_bias": "model-00004-of-00004.safetensors",
         "model.encoder.vision_tower.std_scale": "model-00004-of-00004.safetensors"
     }
-}
+}