coreml_ane: W8A16 embed package (int8 weights, fp16 activations) — passed a 384-query paired gate, 96 -> 73 MB

coreml_ane/embed.mlpackage/Data/com.apple.CoreML/model.mlmodel

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6e0d55be95cfdd236581e15174b823b0800c7ff3261ecf6c3afd99281d1eb320
-size 188511
+oid sha256:8010263dfa1be2ada50a341dd4bf1596cddd845088c7de92a799140863c93c45
+size 441263

coreml_ane/embed.mlpackage/Data/com.apple.CoreML/weights/weight.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0be7d62c31e99dfe9a391af25b7a68e01f72e073e74bdd9aade557142ce4c066
-size 93609984
+oid sha256:065695bebebdbea2b803f831cebfc0bd88319d81ff59c303b9edc5cc5b5dedf7
+size 75918464

coreml_ane/embed.mlpackage/Manifest.json

@@ -1,18 +1,18 @@
 {
     "fileFormatVersion": "1.0.0",
     "itemInfoEntries": {
-        "36D22F73-0F7B-46E5-AF85-F37572EBA0A8": {
-            "author": "com.apple.CoreML",
-            "description": "CoreML Model Weights",
-            "name": "weights",
-            "path": "com.apple.CoreML/weights"
-        },
-        "E0E2AA4E-6100-4400-A6BB-B0E7BEA8C872": {
+        "0DDB2BD8-2D40-4072-921F-E9887001F7F3": {
             "author": "com.apple.CoreML",
             "description": "CoreML Model Specification",
             "name": "model.mlmodel",
             "path": "com.apple.CoreML/model.mlmodel"
+        },
+        "3D8BA34E-142B-4952-9EA4-55DF20B3571D": {
+            "author": "com.apple.CoreML",
+            "description": "CoreML Model Weights",
+            "name": "weights",
+            "path": "com.apple.CoreML/weights"
         }
     },
-    "rootModelIdentifier": "E0E2AA4E-6100-4400-A6BB-B0E7BEA8C872"
+    "rootModelIdentifier": "0DDB2BD8-2D40-4072-921F-E9887001F7F3"
 }

coreml_ane/embed.mlpackage/shapes.json

@@ -1,4 +1,7 @@
 {
+  "head": "mean",
+  "output_dim": 768,
+  "per_token": false,
   "shapes": [
     {
       "batch": 96,
@@ -20,6 +23,5 @@
       "seq": 128,
       "function": "b256_s128"
     }
-  ],
-  "output_dim": 768
+  ]
 }