Add b2 (beta3-loadable) bundle under gpu-pipelined-b2/gemma4_e2b_qat_decode_int4lin_tbl_aotc_h18p/ (b1 retained)

.gitattributes

@@ -74,3 +74,6 @@ gpu-pipelined/gemma4_e2b_qat_vl_decode_int4linsym_aotc_h18p/tokenizer/tokenizer.
 gpu-pipelined/gemma4_e2b_qat_vl_vision/gemma4_e2b_qat_vl_vision.aimodel/main.mlirb filter=lfs diff=lfs merge=lfs -text
 gpu-pipelined-b2/gemma4_e2b_qat_decode_int4lin_tbl/gemma4_e2b_qat_decode_int4lin_tbl.aimodel/main.mlirb filter=lfs diff=lfs merge=lfs -text
 gpu-pipelined-b2/gemma4_e2b_qat_decode_int4lin_tbl/tokenizer/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+gpu-pipelined-b2/gemma4_e2b_qat_decode_int4lin_tbl_aotc_h18p/gemma4_e2b_qat_decode_int4lin_tbl.h18p.aimodelc/main-h18p-delegates/MPSGraph/mpsExecutable.mpsgraphpackage/original_model_0.mpsgraph filter=lfs diff=lfs merge=lfs -text
+gpu-pipelined-b2/gemma4_e2b_qat_decode_int4lin_tbl_aotc_h18p/gemma4_e2b_qat_decode_int4lin_tbl.h18p.aimodelc/main-h18p-delegates/MPSGraph/mpsExecutable.mpsgraphpackage/specialized_model_1.mpsgraph filter=lfs diff=lfs merge=lfs -text
+gpu-pipelined-b2/gemma4_e2b_qat_decode_int4lin_tbl_aotc_h18p/tokenizer/tokenizer.json filter=lfs diff=lfs merge=lfs -text

gpu-pipelined-b2/gemma4_e2b_qat_decode_int4lin_tbl_aotc_h18p/gemma4_e2b_qat_decode_int4lin_tbl.h18p.aimodelc/main-h18p-delegates/MPSGraph/mpsExecutable.mpsgraphpackage/original_model_0.mpsgraph

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1ca99d0079a429bdcea730f905e184df1c8e7a767c177acf80ed69ee5c4104a6
+size 1112644

gpu-pipelined-b2/gemma4_e2b_qat_decode_int4lin_tbl_aotc_h18p/gemma4_e2b_qat_decode_int4lin_tbl.h18p.aimodelc/main-h18p-delegates/MPSGraph/mpsExecutable.mpsgraphpackage/resources.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fd4360a186c1d3e15f26e7bda51764813d027365f94d58986435e38a4ed41aac
+size 2088055648

gpu-pipelined-b2/gemma4_e2b_qat_decode_int4lin_tbl_aotc_h18p/gemma4_e2b_qat_decode_int4lin_tbl.h18p.aimodelc/main-h18p-delegates/MPSGraph/mpsExecutable.mpsgraphpackage/specialized_model_1.mpsgraph

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b16981e65bc56e963d4e803153c406e43cfd7648ad2c5d00de4365013d737316
+size 2044737

gpu-pipelined-b2/gemma4_e2b_qat_decode_int4lin_tbl_aotc_h18p/gemma4_e2b_qat_decode_int4lin_tbl.h18p.aimodelc/main.hash

@@ -0,0 +1 @@
+&��AY�f��[���\<�n�:-��.'ċ~2�q

gpu-pipelined-b2/gemma4_e2b_qat_decode_int4lin_tbl_aotc_h18p/gemma4_e2b_qat_decode_int4lin_tbl.h18p.aimodelc/metadata.json

@@ -0,0 +1,6 @@
+{
+  "sourceHash" : "FC6830D06F9FD657676FBC3FF33C5E8FFF492E77AC5D5269CEF2153740A225D0",
+  "producer" : "coreai-build-3600.67.5.8.1",
+  "assetVersion" : "2.0",
+  "creationDate" : "20260713T005618Z"
+}

gpu-pipelined-b2/gemma4_e2b_qat_decode_int4lin_tbl_aotc_h18p/gemma4_e2b_qat_decode_int4lin_tbl.h18p.aimodelc/stats.json

@@ -0,0 +1 @@
+{"storageTypes":[{"count":2292056064,"name":"Int4"},{"count":474594853,"name":"Float16"},{"count":474,"name":"Int32"},{"name":"Float32","count":388},{"count":89,"name":"UInt32"},{"name":"UInt64","count":1}],"computeTypes":["Bool","Float16","Float32","Int32","Int4","Int8","UInt32","UInt64"],"operationDistribution":[{"name":"constant","count":6161},{"name":"mul","count":1288},{"count":1161,"name":"reshape"},{"name":"broadcast_to","count":1054},{"count":1046,"name":"broadcasting_mul"},{"name":"cast","count":1013},{"count":810,"name":"slice"},{"name":"add","count":710},{"name":"broadcast_in_dims","count":580},{"name":"get_shape","count":530},{"count":519,"name":"reduce"},{"count":471,"name":"concat"},{"name":"broadcasting_add","count":433},{"name":"broadcast_shapes","count":388},{"count":347,"name":"batch_matmul"},{"name":"broadcasting_batch_matmul","count":347},{"name":"transpose","count":293},{"name":"broadcasting_divide","count":278},{"name":"divide","count":278},{"name":"blockwise_shift_scale","count":277},{"name":"reduce_sum","count":277},{"count":243,"name":"gather_nd"},{"name":"reduce_mean","count":242},{"count":242,"name":"reduce_product"},{"name":"rsqrt","count":227},{"name":"gelu","count":70},{"count":54,"name":"broadcasting_sub"},{"name":"sub","count":54},{"count":35,"name":"exp"},{"name":"softmax","count":35},{"count":34,"name":"gather_along_axis"},{"name":"read_handle","count":32},{"name":"shrink_dims","count":32},{"count":30,"name":"broadcasting_pow"},{"name":"pow","count":30},{"name":"slice_update","count":30},{"count":30,"name":"write_handle"},{"name":"pad","count":24},{"count":5,"name":"not"},{"name":"broadcasting_greater","count":3},{"name":"greater","count":3},{"count":2,"name":"cos"},{"count":2,"name":"expand_dims"},{"count":2,"name":"range"},{"name":"sin","count":2},{"name":"broadcasting_logical_xor","count":1},{"name":"create_token","count":1},{"name":"logical_xor","count":1},{"name":"tanh","count":1}]}

gpu-pipelined-b2/gemma4_e2b_qat_decode_int4lin_tbl_aotc_h18p/metadata.json

@@ -0,0 +1,28 @@
+{
+  "metadata_version": "0.2",
+  "kind": "llm",
+  "name": "gemma4_e2b_qat_decode_int4lin_tbl_aotc_h18p",
+  "assets": {
+    "main": "gemma4_e2b_qat_decode_int4lin_tbl.h18p.aimodelc"
+  },
+  "language": {
+    "tokenizer": "google/gemma-4-E2B-it-qat-q4_0-unquantized",
+    "vocab_size": 262144,
+    "max_context_length": 4096,
+    "embedded_tokenizer": true,
+    "function_map": {
+      "main": [
+        "main"
+      ]
+    }
+  },
+  "source": {
+    "model_definition": "torch",
+    "hf_model_id": "google/gemma-4-E2B-it-qat-q4_0-unquantized"
+  },
+  "compression": null,
+  "compilation": {
+    "date": "2026-07-12T21:29:03.089320+00:00",
+    "targets": []
+  }
+}

gpu-pipelined-b2/gemma4_e2b_qat_decode_int4lin_tbl_aotc_h18p/tokenizer/tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cc8d3a0ce36466ccc1278bf987df5f71db1719b9ca6b4118264f45cb627bfe0f
+size 32169626

gpu-pipelined-b2/gemma4_e2b_qat_decode_int4lin_tbl_aotc_h18p/tokenizer/tokenizer_config.json

@@ -0,0 +1,96 @@
+{
+  "audio_token": "<|audio|>",
+  "backend": "tokenizers",
+  "boa_token": "<|audio>",
+  "boi_token": "<|image>",
+  "bos_token": "<bos>",
+  "eoa_token": "<audio|>",
+  "eoc_token": "<channel|>",
+  "eoi_token": "<image|>",
+  "eos_token": "<eos>",
+  "eot_token": "<turn|>",
+  "escape_token": "<|\"|>",
+  "etc_token": "<tool_call|>",
+  "etd_token": "<tool|>",
+  "etr_token": "<tool_response|>",
+  "extra_special_tokens": [
+    "<|video|>"
+  ],
+  "image_token": "<|image|>",
+  "is_local": false,
+  "local_files_only": false,
+  "mask_token": "<mask>",
+  "model_max_length": 1000000000000000019884624838656,
+  "model_specific_special_tokens": {
+    "audio_token": "<|audio|>",
+    "boa_token": "<|audio>",
+    "boi_token": "<|image>",
+    "eoa_token": "<audio|>",
+    "eoc_token": "<channel|>",
+    "eoi_token": "<image|>",
+    "eot_token": "<turn|>",
+    "escape_token": "<|\"|>",
+    "etc_token": "<tool_call|>",
+    "etd_token": "<tool|>",
+    "etr_token": "<tool_response|>",
+    "image_token": "<|image|>",
+    "soc_token": "<|channel>",
+    "sot_token": "<|turn>",
+    "stc_token": "<|tool_call>",
+    "std_token": "<|tool>",
+    "str_token": "<|tool_response>",
+    "think_token": "<|think|>"
+  },
+  "pad_token": "<pad>",
+  "padding_side": "left",
+  "processor_class": "Gemma4Processor",
+  "response_schema": {
+    "properties": {
+      "content": {
+        "type": "string"
+      },
+      "role": {
+        "const": "assistant"
+      },
+      "thinking": {
+        "type": "string"
+      },
+      "tool_calls": {
+        "items": {
+          "properties": {
+            "function": {
+              "properties": {
+                "arguments": {
+                  "additionalProperties": {},
+                  "type": "object",
+                  "x-parser": "gemma4-tool-call"
+                },
+                "name": {
+                  "type": "string"
+                }
+              },
+              "type": "object",
+              "x-regex": "call\\:(?P<name>\\w+)(?P<arguments>\\{.*\\})"
+            },
+            "type": {
+              "const": "function"
+            }
+          },
+          "type": "object"
+        },
+        "type": "array",
+        "x-regex-iterator": "<\\|tool_call>(.*?)<tool_call\\|>"
+      }
+    },
+    "type": "object",
+    "x-regex": "(\\<\\|channel\\>thought\\n(?P<thinking>.*?)\\<channel\\|\\>)?(?P<tool_calls>\\<\\|tool_call\\>.*\\<tool_call\\|\\>)?(?P<content>(?:(?!\\<turn\\|\\>)(?!\\<\\|tool_response\\>).)+)?(?:\\<turn\\|\\>|\\<\\|tool_response\\>)?"
+  },
+  "soc_token": "<|channel>",
+  "sot_token": "<|turn>",
+  "stc_token": "<|tool_call>",
+  "std_token": "<|tool>",
+  "str_token": "<|tool_response>",
+  "think_token": "<|think|>",
+  "tokenizer_class": "GemmaTokenizer",
+  "unk_token": "<unk>"
+}