ios-static: the 6-bit k-means g8 static iOS export before AOT (replaces the 4-bit g32 IR of 2026-09-15)

.gitattributes

@@ -101,3 +101,4 @@ ios-ane-h18p/minicpm5_2b_minicpm5_pal6_g8_static.h18p.aimodelc/main-h18p-delegat
 ios-ane-h18p/minicpm5_2b_minicpm5_pal6_g8_static.h18p.aimodelc/main-h18p-delegates/MPSGraph/mpsExecutable.mpsgraphpackage/original_model_0.mpsgraph filter=lfs diff=lfs merge=lfs -text
 ios-ane-h18p/minicpm5_2b_minicpm5_pal6_g8_static.h18p.aimodelc/main-h18p-delegates/MPSGraph/mpsExecutable.mpsgraphpackage/specialized_model_1.mpsgraph filter=lfs diff=lfs merge=lfs -text
 ios-ane-h18p/minicpm5_2b_minicpm5_pal6_g8_static.h18p.aimodelc/main-h18p.mlirb filter=lfs diff=lfs merge=lfs -text
+ios-static/minicpm5_2b_minicpm5_pal6_g8_static.aimodel/main.mlirb filter=lfs diff=lfs merge=lfs -text

ios-static/metadata.json

@@ -1,9 +1,9 @@
 {
   "metadata_version": "0.2",
   "kind": "llm",
-  "name": "minicpm5_2b_4bit_weight_palettized_group32_static",
+  "name": "minicpm5_2b_minicpm5_pal6_g8_static",
   "assets": {
-    "main": "minicpm5_2b_4bit_weight_palettized_group32_static.aimodel"
+    "main": "minicpm5_2b_minicpm5_pal6_g8_static.aimodel"
   },
   "language": {
     "tokenizer": "openbmb/MiniCPM5-2B",
@@ -20,9 +20,9 @@
     "model_definition": "torch",
     "hf_model_id": "openbmb/MiniCPM5-2B"
   },
-  "compression": "4bit_weight_palettized_group32",
+  "compression": "minicpm5_pal6_g8",
   "compilation": {
-    "date": "2026-09-15T11:08:14.715836+09:00",
+    "date": "2026-09-16T10:55:25.976722+09:00",
     "targets": []
   }
 }

ios-static/minicpm5_2b_4bit_weight_palettized_group32_static.aimodel/main.hash

@@ -1 +0,0 @@
-�C��D\�Em��!�{Ji~���N�iW�

ios-static/minicpm5_2b_4bit_weight_palettized_group32_static.aimodel/metadata.json

@@ -1,8 +0,0 @@
-{
-  "assetVersion": "2.0",
-  "producer": "coreai-core 1.0.0b2",
-  "creationDate": "20260915T020813Z",
-  "author": "OpenBMB",
-  "license": "Apache-2.0",
-  "description": "MiniCPM5-2B is a 2.5B-parameter on-device causal language model from OpenBMB (dense LlamaForCausalLM, 42 layers) with hybrid Think/No-Think reasoning, tool calling and 128K context. Source: https://huggingface.co/openbmb/MiniCPM5-2B"
-}

ios-static/minicpm5_2b_minicpm5_pal6_g8_static.aimodel/main.hash

@@ -0,0 +1 @@
+
͏�RvC���z]�!��K��$'q��J��{UIL

ios-static/{minicpm5_2b_4bit_weight_palettized_group32_static.aimodel → minicpm5_2b_minicpm5_pal6_g8_static.aimodel}/main.mlirb

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8743d67fa5445c9f456deced8521897b4a16691a7ea313b1fd4ef31f6957ef96
-size 1414321247
+oid sha256:01cd8fd8527643a8ffad7a5db92194aa044b8f882427719f964a928b7b55494c
+size 1990517613

ios-static/minicpm5_2b_minicpm5_pal6_g8_static.aimodel/metadata.json

@@ -0,0 +1,5 @@
+{
+  "creationDate" : "20260916T015524Z",
+  "assetVersion" : "2.0",
+  "producer" : "coreai-core 1.0.0b2"
+}