Upload folder using huggingface_hub

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+tokenizer.json filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,87 @@
+---
+license: mit
+tags:
+- coreml
+- ANE
+- Apple Neural Engine
+- Qwen2.5
+- VibeThinker
+- math
+- reasoning
+---
+# anemll-WeiboAI-VibeThinker-1.5B-ctx2048
+
+CoreML conversion of [WeiboAI/VibeThinker-1.5B](https://huggingface.co/WeiboAI/VibeThinker-1.5B) for Apple Neural Engine inference, converted using [ANEMLL](https://github.com/anemll/anemll) v0.3.5.
+
+## Original Model
+
+**VibeThinker-1.5B** is a 1.5B parameter dense language model by WeiboAI, fine-tuned from Qwen2.5-Math-1.5B for competitive math and algorithm coding. Despite its small size, it achieves remarkable results:
+
+| Benchmark | VibeThinker-1.5B | DeepSeek R1 (671B) |
+|-----------|----------------:|-------------------:|
+| AIME24    | 80.3            | 79.8               |
+| AIME25    | 74.4            | 70.0               |
+| HMMT25    | 50.4            | 41.7               |
+
+Best results with English prompts on competitive-style math and coding problems.
+
+## Conversion Details
+
+| Parameter | Value |
+|-----------|-------|
+| Architecture | Qwen 2.5 |
+| Context Length | 2048 |
+| Batch Size | 64 |
+| Chunks | 4 (hybrid) |
+| FFN Quantization | LUT6 (Apple Neural Engine), per-channel group size 4 |
+| LM Head Quantization | LUT6 (Apple Neural Engine), per-channel group size 4 |
+| Embeddings | Unquantized |
+| Argmax in Model | No |
+
+### Hybrid FP32 First Chunk
+
+This model uses a **hybrid precision** layout. Chunk 1 of 4 (`qwen25_FFN_PF_lut6_chunk_01of04.mlmodelc`) runs the first transformer layer's attention in **FP32** (unquantized), while chunks 2-4 use standard LUT6 quantization.
+
+**Why:** VibeThinker has unusually high-magnitude Q/K projection biases in layer 0, causing FP16 attention logit overflow (`+inf`) before masking. This leads to softmax saturation and catastrophic output divergence. Running first-layer attention at full precision eliminates the overflow and recovers near-exact parity with the original Hugging Face model.
+
+Chunk 1 is significantly smaller (~15 MB vs ~310-340 MB for the other chunks) since it only contains the first-layer attention residual path.
+
+### Recommended Sampling
+
+```yaml
+temperature: 0.6
+top_p: 0.95
+top_k: 0
+do_sample: true
+```
+
+## Quick Start
+
+```bash
+pip install coremltools transformers
+
+# Basic chat
+python chat.py --meta ./meta.yaml
+
+# Full conversation mode with history
+python chat_full.py --meta ./meta.yaml
+```
+
+> Note: First load takes time as macOS places the model on the Neural Engine. Subsequent loads are fast.
+
+### iOS/macOS App
+
+Try the ANEMLL Chat app on TestFlight:
+1. Install [TestFlight](https://apps.apple.com/app/testflight/id899247664)
+2. Join beta: [TestFlight Link](https://testflight.apple.com/join/jrQq1D1C)
+3. Add this model via its HuggingFace URL
+
+## Links
+
+- [ANEMLL GitHub](https://github.com/anemll/anemll)
+- [ANEMLL Models on HuggingFace](https://huggingface.co/anemll)
+- [Original Model: WeiboAI/VibeThinker-1.5B](https://huggingface.co/WeiboAI/VibeThinker-1.5B)
+
+## License
+
+MIT (ANEMLL conversion). The original model is also MIT licensed.

chat_template.jinja

@@ -0,0 +1,54 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0]['role'] == 'system' %}
+        {{- messages[0]['content'] }}
+    {%- else %}
+        {{- '' }}
+    {%- endif %}
+    {{- "\n\n# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0]['role'] == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0]['content'] + '<|im_end|>\n' }}
+    {%- else %}
+        {{- '<|im_start|>system\nPlease reason step by step, and put your final answer within \\boxed{}.<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) or (message.role == "assistant" and not message.tool_calls) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {{- '<|im_start|>' + message.role }}
+        {%- if message.content %}
+            {{- '\n' + message.content }}
+        {%- endif %}
+        {%- for tool_call in message.tool_calls %}
+            {%- if tool_call.function is defined %}
+                {%- set tool_call = tool_call.function %}
+            {%- endif %}
+            {{- '\n<tool_call>\n{"name": "' }}
+            {{- tool_call.name }}
+            {{- '", "arguments": ' }}
+            {{- tool_call.arguments | tojson }}
+            {{- '}\n</tool_call>' }}
+        {%- endfor %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if (loop.index0 == 0) or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+{%- endif %}

config.json

@@ -0,0 +1,4 @@
+{
+  "tokenizer_class": "LlamaTokenizer",
+  "model_type": "llama"
+}

meta.yaml

@@ -0,0 +1,56 @@
+model_info:
+  name: anemll-WeiboAI-VibeThinker-1.5B-ctx2048
+  version: 0.3.5
+  description: 'Demonstarates running WeiboAI-VibeThinker-1.5B on Apple Neural Engine
+
+    Context length: 2048
+
+    Batch size: 64
+
+    Chunks: 4
+
+    '
+  license: MIT
+  author: Anemll
+  framework: Core ML
+  language: Python
+  architecture: qwen2
+  parameters:
+    context_length: 2048
+    batch_size: 64
+    lut_embeddings: none
+    lut_ffn: 6
+    lut_ffn_per_channel: 4
+    lut_lmhead: 6
+    lut_lmhead_per_channel: 4
+    num_chunks: 4
+    model_prefix: qwen25
+    embeddings: qwen25_embeddings.mlmodelc
+    lm_head: qwen25_lm_head_lut6.mlmodelc
+    ffn: qwen25_FFN_PF_lut6_chunk_01of04.mlmodelc
+    split_lm_head: 16
+    vocab_size: 151936
+    lm_head_chunk_sizes:
+    - 9496
+    - 9496
+    - 9496
+    - 9496
+    - 9496
+    - 9496
+    - 9496
+    - 9496
+    - 9496
+    - 9496
+    - 9496
+    - 9496
+    - 9496
+    - 9496
+    - 9496
+    - 9496
+    prefill_dynamic_slice: true
+    argmax_in_model: false
+    recommended_sampling:
+      do_sample: true
+      temperature: 0.6
+      top_p: 0.95
+      top_k: 0

qwen25_FFN_PF_lut6_chunk_01of04.mlmodelc/analytics/coremldata.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d3c52c3d9dc9db6e235a0044fc97a3f960045bc6c01227f971e0a2941026a942
+size 243

qwen25_FFN_PF_lut6_chunk_01of04.mlmodelc/coremldata.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1aa2254d0b7cb3cdc9bcf1ee539fed260d554dc0563f3f650138d581dd7c2f59
+size 542

qwen25_FFN_PF_lut6_chunk_01of04.mlmodelc/metadata.json

@@ -0,0 +1,304 @@
+[
+  {
+    "metadataOutputVersion" : "3.0",
+    "storagePrecision" : "Mixed (Float16, Float32)",
+    "outputSchema" : [
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 1536)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 1536]",
+        "name" : "output_hidden_states",
+        "type" : "MultiArray"
+      }
+    ],
+    "modelParameters" : [
+
+    ],
+    "specificationVersion" : 9,
+    "functions" : [
+      {
+        "inputSchema" : [
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 1536)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 1536]",
+            "name" : "hidden_states",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Int32",
+            "formattedType" : "MultiArray (Int32 1)",
+            "shortDescription" : "",
+            "shape" : "[1]",
+            "name" : "position_ids",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 2048)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 1, 2048]",
+            "name" : "causal_mask",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Int32",
+            "formattedType" : "MultiArray (Int32 1)",
+            "shortDescription" : "",
+            "shape" : "[1]",
+            "name" : "current_pos",
+            "type" : "MultiArray"
+          }
+        ],
+        "computePrecision" : "Mixed (Float16, Float32, Int32)",
+        "storagePrecision" : "Mixed (Float16, Float32)",
+        "stateSchema" : [
+          {
+            "dataType" : "Float16",
+            "isOptional" : "0",
+            "formattedType" : "State (Float16 56 × 2 × 2048 × 128)",
+            "shortDescription" : "",
+            "shape" : "[56, 2, 2048, 128]",
+            "name" : "model_model_kv_cache_0",
+            "type" : "State"
+          }
+        ],
+        "outputSchema" : [
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 1536)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 1536]",
+            "name" : "output_hidden_states",
+            "type" : "MultiArray"
+          }
+        ],
+        "name" : "infer",
+        "mlProgramOperationTypeHistogram" : {
+          "Ios18.expandDims" : 3,
+          "Ios18.softmax" : 1,
+          "Ios18.mul" : 9,
+          "Ios18.matmul" : 2,
+          "Identity" : 1,
+          "Ios18.greaterEqual" : 1,
+          "Select" : 1,
+          "Ios18.readState" : 3,
+          "Tile" : 2,
+          "Ios18.gather" : 2,
+          "Ios18.add" : 6,
+          "Ios18.layerNorm" : 1,
+          "Ios18.sliceUpdate" : 2,
+          "Ios18.writeState" : 2,
+          "Ios18.reshape" : 8,
+          "Ios18.conv" : 4,
+          "Ios18.concat" : 7,
+          "Ios18.transpose" : 4,
+          "Ios18.cast" : 18,
+          "Ios18.sliceByIndex" : 7,
+          "Ios18.squeeze" : 2
+        }
+      },
+      {
+        "inputSchema" : [
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 64 × 1536)",
+            "shortDescription" : "",
+            "shape" : "[1, 64, 1536]",
+            "name" : "hidden_states",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Int32",
+            "formattedType" : "MultiArray (Int32 64)",
+            "shortDescription" : "",
+            "shape" : "[64]",
+            "name" : "position_ids",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 64 × 2048)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 64, 2048]",
+            "name" : "causal_mask",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Int32",
+            "formattedType" : "MultiArray (Int32 1)",
+            "shortDescription" : "",
+            "shape" : "[1]",
+            "name" : "current_pos",
+            "type" : "MultiArray"
+          }
+        ],
+        "computePrecision" : "Mixed (Float16, Float32, Int32)",
+        "storagePrecision" : "Mixed (Float16, Float32)",
+        "stateSchema" : [
+          {
+            "dataType" : "Float16",
+            "isOptional" : "0",
+            "formattedType" : "State (Float16 56 × 2 × 2048 × 128)",
+            "shortDescription" : "",
+            "shape" : "[56, 2, 2048, 128]",
+            "name" : "model_model_kv_cache_0",
+            "type" : "State"
+          }
+        ],
+        "outputSchema" : [
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 64 × 1536)",
+            "shortDescription" : "",
+            "shape" : "[1, 64, 1536]",
+            "name" : "output_hidden_states",
+            "type" : "MultiArray"
+          }
+        ],
+        "name" : "prefill",
+        "mlProgramOperationTypeHistogram" : {
+          "Ios18.expandDims" : 3,
+          "Ios18.softmax" : 1,
+          "Ios18.mul" : 9,
+          "Ios18.matmul" : 2,
+          "Ios18.greaterEqual" : 1,
+          "Select" : 1,
+          "Ios18.readState" : 3,
+          "Tile" : 2,
+          "Ios18.gather" : 2,
+          "Ios18.add" : 6,
+          "Ios18.layerNorm" : 1,
+          "Ios18.sliceUpdate" : 2,
+          "Ios18.writeState" : 2,
+          "Ios18.reshape" : 10,
+          "Ios18.conv" : 4,
+          "Ios18.concat" : 7,
+          "Ios18.transpose" : 9,
+          "Ios18.cast" : 18,
+          "Ios18.sliceByIndex" : 7,
+          "Ios18.squeeze" : 2
+        }
+      }
+    ],
+    "mlProgramOperationTypeHistogram" : {
+      "Ios18.expandDims" : 3,
+      "Ios18.softmax" : 1,
+      "Ios18.mul" : 9,
+      "Ios18.matmul" : 2,
+      "Identity" : 1,
+      "Ios18.greaterEqual" : 1,
+      "Select" : 1,
+      "Ios18.readState" : 3,
+      "Tile" : 2,
+      "Ios18.gather" : 2,
+      "Ios18.add" : 6,
+      "Ios18.layerNorm" : 1,
+      "Ios18.sliceUpdate" : 2,
+      "Ios18.writeState" : 2,
+      "Ios18.reshape" : 8,
+      "Ios18.conv" : 4,
+      "Ios18.concat" : 7,
+      "Ios18.transpose" : 4,
+      "Ios18.cast" : 18,
+      "Ios18.sliceByIndex" : 7,
+      "Ios18.squeeze" : 2
+    },
+    "isUpdatable" : "0",
+    "stateSchema" : [
+      {
+        "dataType" : "Float16",
+        "isOptional" : "0",
+        "formattedType" : "State (Float16 56 × 2 × 2048 × 128)",
+        "shortDescription" : "",
+        "shape" : "[56, 2, 2048, 128]",
+        "name" : "model_model_kv_cache_0",
+        "type" : "State"
+      }
+    ],
+    "availability" : {
+      "macOS" : "15.0",
+      "tvOS" : "18.0",
+      "visionOS" : "2.0",
+      "watchOS" : "11.0",
+      "iOS" : "18.0",
+      "macCatalyst" : "18.0"
+    },
+    "computePrecision" : "Mixed (Float16, Float32, Int32)",
+    "modelType" : {
+      "name" : "MLModelType_mlProgram"
+    },
+    "inputSchema" : [
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 1536)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 1536]",
+        "name" : "hidden_states",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Int32",
+        "formattedType" : "MultiArray (Int32 1)",
+        "shortDescription" : "",
+        "shape" : "[1]",
+        "name" : "position_ids",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 2048)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 1, 2048]",
+        "name" : "causal_mask",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Int32",
+        "formattedType" : "MultiArray (Int32 1)",
+        "shortDescription" : "",
+        "shape" : "[1]",
+        "name" : "current_pos",
+        "type" : "MultiArray"
+      }
+    ],
+    "defaultFunctionName" : "infer",
+    "generatedClassName" : "qwen25_FFN_PF_lut6_chunk_01of04",
+    "userDefinedMetadata" : {
+
+    },
+    "method" : "predict"
+  }
+]

qwen25_FFN_PF_lut6_chunk_01of04.mlmodelc/model.mil

@@ -0,0 +1,477 @@
+program(1.3)
+[buildInfo = dict<string, string>({{"coremlc-component-MIL", "3510.2.1"}, {"coremlc-version", "3500.32.1"}})]
+{
+    func infer<ios18>(tensor<fp16, [1, 1, 1, 2048]> causal_mask, tensor<int32, [1]> current_pos, tensor<fp16, [1, 1, 1536]> hidden_states, state<tensor<fp16, [56, 2, 2048, 128]>> model_model_kv_cache_0, tensor<int32, [1]> position_ids) {
+            string cast_0_dtype_0 = const()[name = string("cast_0_dtype_0"), val = string("fp32")];
+            string cast_1_dtype_0 = const()[name = string("cast_1_dtype_0"), val = string("fp32")];
+            tensor<fp16, [1536]> model_model_layers_0_self_attn_q_proj_bias = const()[name = string("model_model_layers_0_self_attn_q_proj_bias"), val = tensor<fp16, [1536]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(64)))];
+            tensor<fp16, [1536, 1536, 1, 1]> model_model_layers_0_self_attn_q_proj_weight = const()[name = string("model_model_layers_0_self_attn_q_proj_weight"), val = tensor<fp16, [1536, 1536, 1, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(3200)))];
+            tensor<fp16, [256]> model_model_layers_0_self_attn_k_proj_bias = const()[name = string("model_model_layers_0_self_attn_k_proj_bias"), val = tensor<fp16, [256]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(4721856)))];
+            tensor<fp16, [256, 1536, 1, 1]> model_model_layers_0_self_attn_k_proj_weight = const()[name = string("model_model_layers_0_self_attn_k_proj_weight"), val = tensor<fp16, [256, 1536, 1, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(4722432)))];
+            tensor<fp16, [256]> model_model_layers_0_self_attn_v_proj_bias = const()[name = string("model_model_layers_0_self_attn_v_proj_bias"), val = tensor<fp16, [256]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(5508928)))];
+            tensor<fp16, [256, 1536, 1, 1]> model_model_layers_0_self_attn_v_proj_weight = const()[name = string("model_model_layers_0_self_attn_v_proj_weight"), val = tensor<fp16, [256, 1536, 1, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(5509504)))];
+            fp32 var_45 = const()[name = string("op_45"), val = fp32(0x1.0c6f7ap-20)];
+            int32 var_48 = const()[name = string("op_48"), val = int32(-1)];
+            fp32 const_0_promoted = const()[name = string("const_0_promoted"), val = fp32(-0x1p+0)];
+            tensor<fp32, [1, 1, 1536]> cast_0 = cast(dtype = cast_0_dtype_0, x = hidden_states)[name = string("cast_31")];
+            tensor<fp32, [1, 1, 1536]> var_50 = mul(x = cast_0, y = const_0_promoted)[name = string("op_50")];
+            bool input_1_interleave_0 = const()[name = string("input_1_interleave_0"), val = bool(false)];
+            tensor<fp32, [1, 1, 3072]> input_1 = concat(axis = var_48, interleave = input_1_interleave_0, values = (cast_0, var_50))[name = string("input_1")];
+            tensor<int32, [1]> normed_1_axes_0 = const()[name = string("normed_1_axes_0"), val = tensor<int32, [1]>([-1])];
+            tensor<fp32, [1, 1, 3072]> normed_1 = layer_norm(axes = normed_1_axes_0, epsilon = var_45, x = input_1)[name = string("normed_1")];
+            tensor<int32, [3]> normed_begin_0 = const()[name = string("normed_begin_0"), val = tensor<int32, [3]>([0, 0, 0])];
+            tensor<int32, [3]> normed_end_0 = const()[name = string("normed_end_0"), val = tensor<int32, [3]>([1, 1, 1536])];
+            tensor<bool, [3]> normed_end_mask_0 = const()[name = string("normed_end_mask_0"), val = tensor<bool, [3]>([true, true, false])];
+            tensor<fp32, [1, 1, 1536]> normed = slice_by_index(begin = normed_begin_0, end = normed_end_0, end_mask = normed_end_mask_0, x = normed_1)[name = string("normed")];
+            tensor<fp32, [1536]> const_3_promoted = const()[name = string("const_3_promoted"), val = tensor<fp32, [1536]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(6296000)))];
+            tensor<fp32, [1, 1, 1536]> hidden_states_1 = mul(x = normed, y = const_3_promoted)[name = string("hidden_states")];
+            tensor<fp32, [1, 4096, 128]> var_64 = const()[name = string("op_64"), val = tensor<fp32, [1, 4096, 128]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(6302208)))];
+            int32 var_72_batch_dims_0 = const()[name = string("op_72_batch_dims_0"), val = int32(0)];
+            bool var_72_validate_indices_0 = const()[name = string("op_72_validate_indices_0"), val = bool(false)];
+            int32 greater_equal_0_y_0 = const()[name = string("greater_equal_0_y_0"), val = int32(0)];
+            tensor<bool, [1]> greater_equal_0 = greater_equal(x = current_pos, y = greater_equal_0_y_0)[name = string("greater_equal_0")];
+            int32 slice_by_index_0 = const()[name = string("slice_by_index_0"), val = int32(4096)];
+            tensor<int32, [1]> add_0 = add(x = current_pos, y = slice_by_index_0)[name = string("add_0")];
+            tensor<int32, [1]> select_0 = select(a = current_pos, b = add_0, cond = greater_equal_0)[name = string("select_0")];
+            int32 var_72_axis_1 = const()[name = string("op_72_axis_1"), val = int32(1)];
+            tensor<fp32, [1, 1, 128]> var_72 = gather(axis = var_72_axis_1, batch_dims = var_72_batch_dims_0, indices = select_0, validate_indices = var_72_validate_indices_0, x = var_64)[name = string("op_72")];
+            tensor<int32, [4]> var_77 = const()[name = string("op_77"), val = tensor<int32, [4]>([1, 1, 1, -1])];
+            tensor<fp32, [1, 1, 1, 128]> sin_1 = reshape(shape = var_77, x = var_72)[name = string("sin_1")];
+            tensor<fp32, [1, 4096, 128]> var_79 = const()[name = string("op_79"), val = tensor<fp32, [1, 4096, 128]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(8399424)))];
+            int32 var_87_batch_dims_0 = const()[name = string("op_87_batch_dims_0"), val = int32(0)];
+            bool var_87_validate_indices_0 = const()[name = string("op_87_validate_indices_0"), val = bool(false)];
+            int32 var_87_axis_1 = const()[name = string("op_87_axis_1"), val = int32(1)];
+            tensor<fp32, [1, 1, 128]> var_87 = gather(axis = var_87_axis_1, batch_dims = var_87_batch_dims_0, indices = select_0, validate_indices = var_87_validate_indices_0, x = var_79)[name = string("op_87")];
+            tensor<int32, [4]> var_92 = const()[name = string("op_92"), val = tensor<int32, [4]>([1, 1, 1, -1])];
+            tensor<fp32, [1, 1, 1, 128]> cos_1 = reshape(shape = var_92, x = var_87)[name = string("cos_1")];
+            string cos_dtype_0 = const()[name = string("cos_dtype_0"), val = string("fp16")];
+            string sin_dtype_0 = const()[name = string("sin_dtype_0"), val = string("fp16")];
+            tensor<int32, [3]> var_107 = const()[name = string("op_107"), val = tensor<int32, [3]>([0, 2, 1])];
+            tensor<int32, [1]> var_110_axes_0 = const()[name = string("op_110_axes_0"), val = tensor<int32, [1]>([2])];
+            tensor<fp32, [1, 1536, 1]> var_108 = transpose(perm = var_107, x = hidden_states_1)[name = string("transpose_3")];
+            tensor<fp32, [1, 1536, 1, 1]> var_110 = expand_dims(axes = var_110_axes_0, x = var_108)[name = string("op_110")];
+            string input_3_dtype_0 = const()[name = string("input_3_dtype_0"), val = string("fp16")];
+            string var_126_pad_type_0 = const()[name = string("op_126_pad_type_0"), val = string("valid")];
+            tensor<int32, [2]> var_126_strides_0 = const()[name = string("op_126_strides_0"), val = tensor<int32, [2]>([1, 1])];
+            tensor<int32, [4]> var_126_pad_0 = const()[name = string("op_126_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
+            tensor<int32, [2]> var_126_dilations_0 = const()[name = string("op_126_dilations_0"), val = tensor<int32, [2]>([1, 1])];
+            int32 var_126_groups_0 = const()[name = string("op_126_groups_0"), val = int32(1)];
+            tensor<fp16, [1, 1536, 1, 1]> input_3 = cast(dtype = input_3_dtype_0, x = var_110)[name = string("cast_27")];
+            tensor<fp16, [1, 1536, 1, 1]> var_126 = conv(bias = model_model_layers_0_self_attn_q_proj_bias, dilations = var_126_dilations_0, groups = var_126_groups_0, pad = var_126_pad_0, pad_type = var_126_pad_type_0, strides = var_126_strides_0, weight = model_model_layers_0_self_attn_q_proj_weight, x = input_3)[name = string("op_126")];
+            tensor<int32, [4]> var_131 = const()[name = string("op_131"), val = tensor<int32, [4]>([1, 12, 1, 128])];
+            tensor<fp16, [1, 12, 1, 128]> var_132 = reshape(shape = var_131, x = var_126)[name = string("op_132")];
+            string var_148_pad_type_0 = const()[name = string("op_148_pad_type_0"), val = string("valid")];
+            tensor<int32, [2]> var_148_strides_0 = const()[name = string("op_148_strides_0"), val = tensor<int32, [2]>([1, 1])];
+            tensor<int32, [4]> var_148_pad_0 = const()[name = string("op_148_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
+            tensor<int32, [2]> var_148_dilations_0 = const()[name = string("op_148_dilations_0"), val = tensor<int32, [2]>([1, 1])];
+            int32 var_148_groups_0 = const()[name = string("op_148_groups_0"), val = int32(1)];
+            tensor<fp16, [1, 256, 1, 1]> var_148 = conv(bias = model_model_layers_0_self_attn_k_proj_bias, dilations = var_148_dilations_0, groups = var_148_groups_0, pad = var_148_pad_0, pad_type = var_148_pad_type_0, strides = var_148_strides_0, weight = model_model_layers_0_self_attn_k_proj_weight, x = input_3)[name = string("op_148")];
+            tensor<int32, [4]> var_153 = const()[name = string("op_153"), val = tensor<int32, [4]>([1, 2, 1, 128])];
+            tensor<fp16, [1, 2, 1, 128]> var_154 = reshape(shape = var_153, x = var_148)[name = string("op_154")];
+            string var_170_pad_type_0 = const()[name = string("op_170_pad_type_0"), val = string("valid")];
+            tensor<int32, [2]> var_170_strides_0 = const()[name = string("op_170_strides_0"), val = tensor<int32, [2]>([1, 1])];
+            tensor<int32, [4]> var_170_pad_0 = const()[name = string("op_170_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
+            tensor<int32, [2]> var_170_dilations_0 = const()[name = string("op_170_dilations_0"), val = tensor<int32, [2]>([1, 1])];
+            int32 var_170_groups_0 = const()[name = string("op_170_groups_0"), val = int32(1)];
+            tensor<fp16, [1, 256, 1, 1]> var_170 = conv(bias = model_model_layers_0_self_attn_v_proj_bias, dilations = var_170_dilations_0, groups = var_170_groups_0, pad = var_170_pad_0, pad_type = var_170_pad_type_0, strides = var_170_strides_0, weight = model_model_layers_0_self_attn_v_proj_weight, x = input_3)[name = string("op_170")];
+            tensor<int32, [4]> var_175 = const()[name = string("op_175"), val = tensor<int32, [4]>([1, 2, 1, 128])];
+            tensor<fp16, [1, 2, 1, 128]> var_176 = reshape(shape = var_175, x = var_170)[name = string("op_176")];
+            tensor<fp16, [1, 1, 1, 128]> cos = cast(dtype = cos_dtype_0, x = cos_1)[name = string("cast_29")];
+            tensor<fp16, [1, 12, 1, 128]> var_182 = mul(x = var_132, y = cos)[name = string("op_182")];
+            tensor<int32, [4]> x1_1_begin_0 = const()[name = string("x1_1_begin_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
+            tensor<int32, [4]> x1_1_end_0 = const()[name = string("x1_1_end_0"), val = tensor<int32, [4]>([1, 12, 1, 64])];
+            tensor<bool, [4]> x1_1_end_mask_0 = const()[name = string("x1_1_end_mask_0"), val = tensor<bool, [4]>([true, true, true, false])];
+            tensor<fp16, [1, 12, 1, 64]> x1_1 = slice_by_index(begin = x1_1_begin_0, end = x1_1_end_0, end_mask = x1_1_end_mask_0, x = var_132)[name = string("x1_1")];
+            tensor<int32, [4]> x2_1_begin_0 = const()[name = string("x2_1_begin_0"), val = tensor<int32, [4]>([0, 0, 0, 64])];
+            tensor<int32, [4]> x2_1_end_0 = const()[name = string("x2_1_end_0"), val = tensor<int32, [4]>([1, 12, 1, 128])];
+            tensor<bool, [4]> x2_1_end_mask_0 = const()[name = string("x2_1_end_mask_0"), val = tensor<bool, [4]>([true, true, true, true])];
+            tensor<fp16, [1, 12, 1, 64]> x2_1 = slice_by_index(begin = x2_1_begin_0, end = x2_1_end_0, end_mask = x2_1_end_mask_0, x = var_132)[name = string("x2_1")];
+            fp16 const_6_promoted = const()[name = string("const_6_promoted"), val = fp16(-0x1p+0)];
+            tensor<fp16, [1, 12, 1, 64]> var_203 = mul(x = x2_1, y = const_6_promoted)[name = string("op_203")];
+            int32 var_205 = const()[name = string("op_205"), val = int32(-1)];
+            bool var_206_interleave_0 = const()[name = string("op_206_interleave_0"), val = bool(false)];
+            tensor<fp16, [1, 12, 1, 128]> var_206 = concat(axis = var_205, interleave = var_206_interleave_0, values = (var_203, x1_1))[name = string("op_206")];
+            tensor<fp16, [1, 1, 1, 128]> sin = cast(dtype = sin_dtype_0, x = sin_1)[name = string("cast_28")];
+            tensor<fp16, [1, 12, 1, 128]> var_207 = mul(x = var_206, y = sin)[name = string("op_207")];
+            tensor<fp16, [1, 12, 1, 128]> query_states = add(x = var_182, y = var_207)[name = string("query_states")];
+            tensor<fp16, [1, 2, 1, 128]> var_210 = mul(x = var_154, y = cos)[name = string("op_210")];
+            tensor<int32, [4]> x1_begin_0 = const()[name = string("x1_begin_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
+            tensor<int32, [4]> x1_end_0 = const()[name = string("x1_end_0"), val = tensor<int32, [4]>([1, 2, 1, 64])];
+            tensor<bool, [4]> x1_end_mask_0 = const()[name = string("x1_end_mask_0"), val = tensor<bool, [4]>([true, true, true, false])];
+            tensor<fp16, [1, 2, 1, 64]> x1 = slice_by_index(begin = x1_begin_0, end = x1_end_0, end_mask = x1_end_mask_0, x = var_154)[name = string("x1")];
+            tensor<int32, [4]> x2_begin_0 = const()[name = string("x2_begin_0"), val = tensor<int32, [4]>([0, 0, 0, 64])];
+            tensor<int32, [4]> x2_end_0 = const()[name = string("x2_end_0"), val = tensor<int32, [4]>([1, 2, 1, 128])];
+            tensor<bool, [4]> x2_end_mask_0 = const()[name = string("x2_end_mask_0"), val = tensor<bool, [4]>([true, true, true, true])];
+            tensor<fp16, [1, 2, 1, 64]> x2 = slice_by_index(begin = x2_begin_0, end = x2_end_0, end_mask = x2_end_mask_0, x = var_154)[name = string("x2")];
+            fp16 const_9_promoted = const()[name = string("const_9_promoted"), val = fp16(-0x1p+0)];
+            tensor<fp16, [1, 2, 1, 64]> var_231 = mul(x = x2, y = const_9_promoted)[name = string("op_231")];
+            int32 var_233 = const()[name = string("op_233"), val = int32(-1)];
+            bool var_234_interleave_0 = const()[name = string("op_234_interleave_0"), val = bool(false)];
+            tensor<fp16, [1, 2, 1, 128]> var_234 = concat(axis = var_233, interleave = var_234_interleave_0, values = (var_231, x1))[name = string("op_234")];
+            tensor<fp16, [1, 2, 1, 128]> var_235 = mul(x = var_234, y = sin)[name = string("op_235")];
+            tensor<fp16, [1, 2, 1, 128]> key_states_1 = add(x = var_210, y = var_235)[name = string("key_states_1")];
+            int32 var_239 = const()[name = string("op_239"), val = int32(1)];
+            tensor<int32, [1]> var_240 = add(x = current_pos, y = var_239)[name = string("op_240")];
+            tensor<fp16, [56, 2, 2048, 128]> read_state_0 = read_state(input = model_model_kv_cache_0)[name = string("read_state_0")];
+            string cast_7_dtype_0 = const()[name = string("cast_7_dtype_0"), val = string("fp32")];
+            tensor<int32, [1]> expand_dims_0 = const()[name = string("expand_dims_0"), val = tensor<int32, [1]>([0])];
+            tensor<int32, [1]> expand_dims_1 = const()[name = string("expand_dims_1"), val = tensor<int32, [1]>([0])];
+            tensor<int32, [1]> expand_dims_3 = const()[name = string("expand_dims_3"), val = tensor<int32, [1]>([0])];
+            tensor<int32, [1]> expand_dims_4 = const()[name = string("expand_dims_4"), val = tensor<int32, [1]>([1])];
+            int32 concat_2_axis_0 = const()[name = string("concat_2_axis_0"), val = int32(0)];
+            bool concat_2_interleave_0 = const()[name = string("concat_2_interleave_0"), val = bool(false)];
+            tensor<int32, [4]> concat_2 = concat(axis = concat_2_axis_0, interleave = concat_2_interleave_0, values = (expand_dims_0, expand_dims_1, current_pos, expand_dims_3))[name = string("concat_2")];
+            tensor<int32, [1]> concat_3_values1_0 = const()[name = string("concat_3_values1_0"), val = tensor<int32, [1]>([0])];
+            tensor<int32, [1]> concat_3_values3_0 = const()[name = string("concat_3_values3_0"), val = tensor<int32, [1]>([0])];
+            int32 concat_3_axis_0 = const()[name = string("concat_3_axis_0"), val = int32(0)];
+            bool concat_3_interleave_0 = const()[name = string("concat_3_interleave_0"), val = bool(false)];
+            tensor<int32, [4]> concat_3 = concat(axis = concat_3_axis_0, interleave = concat_3_interleave_0, values = (expand_dims_4, concat_3_values1_0, var_240, concat_3_values3_0))[name = string("concat_3")];
+            string key_states_1_promoted_dtype_0 = const()[name = string("key_states_1_promoted_dtype_0"), val = string("fp32")];
+            tensor<int32, [4]> model_model_kv_cache_0_internal_tensor_assign_1_stride_0 = const()[name = string("model_model_kv_cache_0_internal_tensor_assign_1_stride_0"), val = tensor<int32, [4]>([1, 1, 1, 1])];
+            tensor<bool, [4]> model_model_kv_cache_0_internal_tensor_assign_1_begin_mask_0 = const()[name = string("model_model_kv_cache_0_internal_tensor_assign_1_begin_mask_0"), val = tensor<bool, [4]>([false, false, false, false])];
+            tensor<bool, [4]> model_model_kv_cache_0_internal_tensor_assign_1_end_mask_0 = const()[name = string("model_model_kv_cache_0_internal_tensor_assign_1_end_mask_0"), val = tensor<bool, [4]>([false, true, false, true])];
+            tensor<bool, [4]> model_model_kv_cache_0_internal_tensor_assign_1_squeeze_mask_0 = const()[name = string("model_model_kv_cache_0_internal_tensor_assign_1_squeeze_mask_0"), val = tensor<bool, [4]>([false, false, false, false])];
+            tensor<fp32, [1, 2, 1, 128]> key_states_1_promoted = cast(dtype = key_states_1_promoted_dtype_0, x = key_states_1)[name = string("cast_25")];
+            tensor<fp32, [56, 2, 2048, 128]> cast_7 = cast(dtype = cast_7_dtype_0, x = read_state_0)[name = string("cast_26")];
+            tensor<fp32, [56, 2, 2048, 128]> model_model_kv_cache_0_internal_tensor_assign_1 = slice_update(begin = concat_2, begin_mask = model_model_kv_cache_0_internal_tensor_assign_1_begin_mask_0, end = concat_3, end_mask = model_model_kv_cache_0_internal_tensor_assign_1_end_mask_0, squeeze_mask = model_model_kv_cache_0_internal_tensor_assign_1_squeeze_mask_0, stride = model_model_kv_cache_0_internal_tensor_assign_1_stride_0, update = key_states_1_promoted, x = cast_7)[name = string("model_model_kv_cache_0_internal_tensor_assign_1")];
+            string cast_8_dtype_0 = const()[name = string("cast_8_dtype_0"), val = string("fp16")];
+            tensor<fp16, [56, 2, 2048, 128]> cast_8 = cast(dtype = cast_8_dtype_0, x = model_model_kv_cache_0_internal_tensor_assign_1)[name = string("cast_24")];
+            write_state(data = cast_8, input = model_model_kv_cache_0)[name = string("coreml_update_state_4_write_state")];
+            tensor<fp16, [56, 2, 2048, 128]> coreml_update_state_2 = read_state(input = model_model_kv_cache_0)[name = string("coreml_update_state_4")];
+            string model_model_kv_cache_0_internal_tensor_assign_1_dtype_0 = const()[name = string("model_model_kv_cache_0_internal_tensor_assign_1_dtype_0"), val = string("fp32")];
+            tensor<int32, [1]> expand_dims_6 = const()[name = string("expand_dims_6"), val = tensor<int32, [1]>([28])];
+            tensor<int32, [1]> expand_dims_7 = const()[name = string("expand_dims_7"), val = tensor<int32, [1]>([0])];
+            tensor<int32, [1]> expand_dims_9 = const()[name = string("expand_dims_9"), val = tensor<int32, [1]>([0])];
+            tensor<int32, [1]> expand_dims_10 = const()[name = string("expand_dims_10"), val = tensor<int32, [1]>([29])];
+            int32 concat_6_axis_0 = const()[name = string("concat_6_axis_0"), val = int32(0)];
+            bool concat_6_interleave_0 = const()[name = string("concat_6_interleave_0"), val = bool(false)];
+            tensor<int32, [4]> concat_6 = concat(axis = concat_6_axis_0, interleave = concat_6_interleave_0, values = (expand_dims_6, expand_dims_7, current_pos, expand_dims_9))[name = string("concat_6")];
+            tensor<int32, [1]> concat_7_values1_0 = const()[name = string("concat_7_values1_0"), val = tensor<int32, [1]>([0])];
+            tensor<int32, [1]> concat_7_values3_0 = const()[name = string("concat_7_values3_0"), val = tensor<int32, [1]>([0])];
+            int32 concat_7_axis_0 = const()[name = string("concat_7_axis_0"), val = int32(0)];
+            bool concat_7_interleave_0 = const()[name = string("concat_7_interleave_0"), val = bool(false)];
+            tensor<int32, [4]> concat_7 = concat(axis = concat_7_axis_0, interleave = concat_7_interleave_0, values = (expand_dims_10, concat_7_values1_0, var_240, concat_7_values3_0))[name = string("concat_7")];
+            string value_states_1_promoted_dtype_0 = const()[name = string("value_states_1_promoted_dtype_0"), val = string("fp32")];
+            tensor<int32, [4]> model_model_kv_cache_0_internal_tensor_assign_2_stride_0 = const()[name = string("model_model_kv_cache_0_internal_tensor_assign_2_stride_0"), val = tensor<int32, [4]>([1, 1, 1, 1])];
+            tensor<bool, [4]> model_model_kv_cache_0_internal_tensor_assign_2_begin_mask_0 = const()[name = string("model_model_kv_cache_0_internal_tensor_assign_2_begin_mask_0"), val = tensor<bool, [4]>([false, false, false, false])];
+            tensor<bool, [4]> model_model_kv_cache_0_internal_tensor_assign_2_end_mask_0 = const()[name = string("model_model_kv_cache_0_internal_tensor_assign_2_end_mask_0"), val = tensor<bool, [4]>([false, true, false, true])];
+            tensor<bool, [4]> model_model_kv_cache_0_internal_tensor_assign_2_squeeze_mask_0 = const()[name = string("model_model_kv_cache_0_internal_tensor_assign_2_squeeze_mask_0"), val = tensor<bool, [4]>([false, false, false, false])];
+            tensor<fp32, [1, 2, 1, 128]> value_states_1_promoted = cast(dtype = value_states_1_promoted_dtype_0, x = var_176)[name = string("cast_22")];
+            tensor<fp32, [56, 2, 2048, 128]> model_model_kv_cache_0_internal_tensor_assign_1_1 = cast(dtype = model_model_kv_cache_0_internal_tensor_assign_1_dtype_0, x = coreml_update_state_2)[name = string("cast_23")];
+            tensor<fp32, [56, 2, 2048, 128]> model_model_kv_cache_0_internal_tensor_assign_2 = slice_update(begin = concat_6, begin_mask = model_model_kv_cache_0_internal_tensor_assign_2_begin_mask_0, end = concat_7, end_mask = model_model_kv_cache_0_internal_tensor_assign_2_end_mask_0, squeeze_mask = model_model_kv_cache_0_internal_tensor_assign_2_squeeze_mask_0, stride = model_model_kv_cache_0_internal_tensor_assign_2_stride_0, update = value_states_1_promoted, x = model_model_kv_cache_0_internal_tensor_assign_1_1)[name = string("model_model_kv_cache_0_internal_tensor_assign_2")];
+            string cast_9_dtype_0 = const()[name = string("cast_9_dtype_0"), val = string("fp16")];
+            tensor<fp16, [56, 2, 2048, 128]> cast_9 = cast(dtype = cast_9_dtype_0, x = model_model_kv_cache_0_internal_tensor_assign_2)[name = string("cast_21")];
+            write_state(data = cast_9, input = model_model_kv_cache_0)[name = string("coreml_update_state_5_write_state")];
+            tensor<fp16, [56, 2, 2048, 128]> coreml_update_state_3 = read_state(input = model_model_kv_cache_0)[name = string("coreml_update_state_5")];
+            string model_model_kv_cache_0_internal_tensor_assign_2_dtype_0 = const()[name = string("model_model_kv_cache_0_internal_tensor_assign_2_dtype_0"), val = string("fp32")];
+            tensor<int32, [4]> var_290_begin_0 = const()[name = string("op_290_begin_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
+            tensor<int32, [4]> var_290_end_0 = const()[name = string("op_290_end_0"), val = tensor<int32, [4]>([1, 2, 2048, 128])];
+            tensor<bool, [4]> var_290_end_mask_0 = const()[name = string("op_290_end_mask_0"), val = tensor<bool, [4]>([false, true, true, true])];
+            tensor<fp32, [56, 2, 2048, 128]> model_model_kv_cache_0_internal_tensor_assign_2_1 = cast(dtype = model_model_kv_cache_0_internal_tensor_assign_2_dtype_0, x = coreml_update_state_3)[name = string("cast_20")];
+            tensor<fp32, [1, 2, 2048, 128]> var_290 = slice_by_index(begin = var_290_begin_0, end = var_290_end_0, end_mask = var_290_end_mask_0, x = model_model_kv_cache_0_internal_tensor_assign_2_1)[name = string("op_290")];
+            tensor<int32, [1]> K_layer_cache_axes_0 = const()[name = string("K_layer_cache_axes_0"), val = tensor<int32, [1]>([0])];
+            tensor<fp32, [2, 2048, 128]> K_layer_cache = squeeze(axes = K_layer_cache_axes_0, x = var_290)[name = string("K_layer_cache")];
+            tensor<int32, [4]> var_297_begin_0 = const()[name = string("op_297_begin_0"), val = tensor<int32, [4]>([28, 0, 0, 0])];
+            tensor<int32, [4]> var_297_end_0 = const()[name = string("op_297_end_0"), val = tensor<int32, [4]>([29, 2, 2048, 128])];
+            tensor<bool, [4]> var_297_end_mask_0 = const()[name = string("op_297_end_mask_0"), val = tensor<bool, [4]>([false, true, true, true])];
+            tensor<fp32, [1, 2, 2048, 128]> var_297 = slice_by_index(begin = var_297_begin_0, end = var_297_end_0, end_mask = var_297_end_mask_0, x = model_model_kv_cache_0_internal_tensor_assign_2_1)[name = string("op_297")];
+            tensor<int32, [1]> V_layer_cache_axes_0 = const()[name = string("V_layer_cache_axes_0"), val = tensor<int32, [1]>([0])];
+            tensor<fp32, [2, 2048, 128]> V_layer_cache = squeeze(axes = V_layer_cache_axes_0, x = var_297)[name = string("V_layer_cache")];
+            tensor<int32, [1]> x_3_axes_0 = const()[name = string("x_3_axes_0"), val = tensor<int32, [1]>([1])];
+            tensor<fp32, [2, 1, 2048, 128]> x_3 = expand_dims(axes = x_3_axes_0, x = K_layer_cache)[name = string("x_3")];
+            tensor<int32, [4]> var_334 = const()[name = string("op_334"), val = tensor<int32, [4]>([1, 6, 1, 1])];
+            tensor<fp32, [2, 6, 2048, 128]> x_5 = tile(reps = var_334, x = x_3)[name = string("x_5")];
+            tensor<int32, [4]> var_346 = const()[name = string("op_346"), val = tensor<int32, [4]>([1, -1, 2048, 128])];
+            tensor<fp32, [1, 12, 2048, 128]> key_states = reshape(shape = var_346, x = x_5)[name = string("key_states")];
+            tensor<int32, [1]> x_9_axes_0 = const()[name = string("x_9_axes_0"), val = tensor<int32, [1]>([1])];
+            tensor<fp32, [2, 1, 2048, 128]> x_9 = expand_dims(axes = x_9_axes_0, x = V_layer_cache)[name = string("x_9")];
+            tensor<int32, [4]> var_354 = const()[name = string("op_354"), val = tensor<int32, [4]>([1, 6, 1, 1])];
+            tensor<fp32, [2, 6, 2048, 128]> x = tile(reps = var_354, x = x_9)[name = string("x")];
+            tensor<int32, [4]> var_366 = const()[name = string("op_366"), val = tensor<int32, [4]>([1, -1, 2048, 128])];
+            tensor<fp32, [1, 12, 2048, 128]> value_states = reshape(shape = var_366, x = x)[name = string("value_states")];
+            string q_dtype_0 = const()[name = string("q_dtype_0"), val = string("fp32")];
+            bool var_389_transpose_x_1 = const()[name = string("op_389_transpose_x_1"), val = bool(false)];
+            bool var_389_transpose_y_1 = const()[name = string("op_389_transpose_y_1"), val = bool(true)];
+            tensor<fp32, [1, 12, 1, 128]> q = cast(dtype = q_dtype_0, x = query_states)[name = string("cast_19")];
+            tensor<fp32, [1, 12, 1, 2048]> var_389 = matmul(transpose_x = var_389_transpose_x_1, transpose_y = var_389_transpose_y_1, x = q, y = key_states)[name = string("op_389")];
+            fp32 var_390 = const()[name = string("op_390"), val = fp32(0x1.6a09e6p-4)];
+            tensor<fp32, [1, 12, 1, 2048]> attn_logits_1 = mul(x = var_389, y = var_390)[name = string("attn_logits_1")];
+            tensor<fp32, [1, 1, 1, 2048]> cast_1 = cast(dtype = cast_1_dtype_0, x = causal_mask)[name = string("cast_30")];
+            tensor<fp32, [1, 12, 1, 2048]> attn_logits = add(x = attn_logits_1, y = cast_1)[name = string("attn_logits")];
+            int32 var_417 = const()[name = string("op_417"), val = int32(-1)];
+            tensor<fp32, [1, 12, 1, 2048]> var_419 = softmax(axis = var_417, x = attn_logits)[name = string("op_419")];
+            string attn_weights_dtype_0 = const()[name = string("attn_weights_dtype_0"), val = string("fp16")];
+            string var_429_dtype_0 = const()[name = string("op_429_dtype_0"), val = string("fp32")];
+            bool attn_output_1_transpose_x_0 = const()[name = string("attn_output_1_transpose_x_0"), val = bool(false)];
+            bool attn_output_1_transpose_y_0 = const()[name = string("attn_output_1_transpose_y_0"), val = bool(false)];
+            tensor<fp16, [1, 12, 1, 2048]> attn_weights = cast(dtype = attn_weights_dtype_0, x = var_419)[name = string("cast_18")];
+            tensor<fp32, [1, 12, 1, 2048]> var_429 = cast(dtype = var_429_dtype_0, x = attn_weights)[name = string("cast_17")];
+            tensor<fp32, [1, 12, 1, 128]> attn_output_1 = matmul(transpose_x = attn_output_1_transpose_x_0, transpose_y = attn_output_1_transpose_y_0, x = var_429, y = value_states)[name = string("attn_output_1")];
+            string attn_output_3_dtype_0 = const()[name = string("attn_output_3_dtype_0"), val = string("fp16")];
+            tensor<int32, [4]> var_443_perm_0 = const()[name = string("op_443_perm_0"), val = tensor<int32, [4]>([0, 2, 1, 3])];
+            tensor<int32, [3]> var_447 = const()[name = string("op_447"), val = tensor<int32, [3]>([1, 1, 1536])];
+            tensor<fp16, [1, 12, 1, 128]> attn_output_3 = cast(dtype = attn_output_3_dtype_0, x = attn_output_1)[name = string("cast_16")];
+            tensor<fp16, [1, 1, 12, 128]> var_443 = transpose(perm = var_443_perm_0, x = attn_output_3)[name = string("transpose_2")];
+            tensor<fp16, [1, 1, 1536]> attn_output_7 = reshape(shape = var_447, x = var_443)[name = string("attn_output_7")];
+            tensor<int32, [3]> var_452 = const()[name = string("op_452"), val = tensor<int32, [3]>([0, 2, 1])];
+            tensor<fp16, [1536, 1536, 1]> squeeze_0 = const()[name = string("squeeze_0"), val = tensor<fp16, [1536, 1536, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(10496640)))];
+            string var_468_pad_type_0 = const()[name = string("op_468_pad_type_0"), val = string("valid")];
+            int32 var_468_groups_0 = const()[name = string("op_468_groups_0"), val = int32(1)];
+            tensor<int32, [1]> var_468_strides_0 = const()[name = string("op_468_strides_0"), val = tensor<int32, [1]>([1])];
+            tensor<int32, [2]> var_468_pad_0 = const()[name = string("op_468_pad_0"), val = tensor<int32, [2]>([0, 0])];
+            tensor<int32, [1]> var_468_dilations_0 = const()[name = string("op_468_dilations_0"), val = tensor<int32, [1]>([1])];
+            tensor<fp16, [1, 1536, 1]> var_453 = transpose(perm = var_452, x = attn_output_7)[name = string("transpose_1")];
+            tensor<fp16, [1, 1536, 1]> var_468 = conv(dilations = var_468_dilations_0, groups = var_468_groups_0, pad = var_468_pad_0, pad_type = var_468_pad_type_0, strides = var_468_strides_0, weight = squeeze_0, x = var_453)[name = string("op_468")];
+            tensor<int32, [3]> var_472 = const()[name = string("op_472"), val = tensor<int32, [3]>([0, 2, 1])];
+            string attn_out_promoted_dtype_0 = const()[name = string("attn_out_promoted_dtype_0"), val = string("fp32")];
+            tensor<fp16, [1, 1, 1536]> attn_out = transpose(perm = var_472, x = var_468)[name = string("transpose_0")];
+            tensor<fp32, [1, 1, 1536]> attn_out_promoted = cast(dtype = attn_out_promoted_dtype_0, x = attn_out)[name = string("cast_15")];
+            tensor<fp32, [1, 1, 1536]> output_hidden_states_type_fp32 = add(x = cast_0, y = attn_out_promoted)[name = string("op_475")];
+            string cast_13_dtype_0 = const()[name = string("cast_13_dtype_0"), val = string("fp16")];
+            tensor<fp16, [1, 1, 1536]> output_hidden_states = cast(dtype = cast_13_dtype_0, x = output_hidden_states_type_fp32)[name = string("cast_14")];
+            tensor<int32, [1]> position_ids_tmp = identity(x = position_ids)[name = string("position_ids_tmp")];
+        } -> (output_hidden_states);
+    func prefill<ios18>(tensor<fp16, [1, 1, 64, 2048]> causal_mask, tensor<int32, [1]> current_pos, tensor<fp16, [1, 64, 1536]> hidden_states, state<tensor<fp16, [56, 2, 2048, 128]>> model_model_kv_cache_0, tensor<int32, [64]> position_ids) {
+            string cast_0_dtype_0 = const()[name = string("cast_0_dtype_0"), val = string("fp32")];
+            string cast_1_dtype_0 = const()[name = string("cast_1_dtype_0"), val = string("fp32")];
+            tensor<fp16, [1536]> model_model_layers_0_self_attn_q_proj_bias = const()[name = string("model_model_layers_0_self_attn_q_proj_bias"), val = tensor<fp16, [1536]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(64)))];
+            tensor<fp16, [1536, 1536, 1, 1]> model_model_layers_0_self_attn_q_proj_weight = const()[name = string("model_model_layers_0_self_attn_q_proj_weight"), val = tensor<fp16, [1536, 1536, 1, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(3200)))];
+            tensor<fp16, [256]> model_model_layers_0_self_attn_k_proj_bias = const()[name = string("model_model_layers_0_self_attn_k_proj_bias"), val = tensor<fp16, [256]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(4721856)))];
+            tensor<fp16, [256, 1536, 1, 1]> model_model_layers_0_self_attn_k_proj_weight = const()[name = string("model_model_layers_0_self_attn_k_proj_weight"), val = tensor<fp16, [256, 1536, 1, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(4722432)))];
+            tensor<fp16, [256]> model_model_layers_0_self_attn_v_proj_bias = const()[name = string("model_model_layers_0_self_attn_v_proj_bias"), val = tensor<fp16, [256]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(5508928)))];
+            tensor<fp16, [256, 1536, 1, 1]> model_model_layers_0_self_attn_v_proj_weight = const()[name = string("model_model_layers_0_self_attn_v_proj_weight"), val = tensor<fp16, [256, 1536, 1, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(5509504)))];
+            fp32 var_45 = const()[name = string("op_45"), val = fp32(0x1.0c6f7ap-20)];
+            int32 var_48 = const()[name = string("op_48"), val = int32(-1)];
+            fp32 const_0_promoted = const()[name = string("const_0_promoted"), val = fp32(-0x1p+0)];
+            tensor<fp32, [1, 64, 1536]> cast_0 = cast(dtype = cast_0_dtype_0, x = hidden_states)[name = string("cast_31")];
+            tensor<fp32, [1, 64, 1536]> var_50 = mul(x = cast_0, y = const_0_promoted)[name = string("op_50")];
+            bool input_1_interleave_0 = const()[name = string("input_1_interleave_0"), val = bool(false)];
+            tensor<fp32, [1, 64, 3072]> input_1 = concat(axis = var_48, interleave = input_1_interleave_0, values = (cast_0, var_50))[name = string("input_1")];
+            tensor<int32, [1]> normed_1_axes_0 = const()[name = string("normed_1_axes_0"), val = tensor<int32, [1]>([-1])];
+            tensor<fp32, [1, 64, 3072]> normed_1 = layer_norm(axes = normed_1_axes_0, epsilon = var_45, x = input_1)[name = string("normed_1")];
+            tensor<int32, [3]> normed_begin_0 = const()[name = string("normed_begin_0"), val = tensor<int32, [3]>([0, 0, 0])];
+            tensor<int32, [3]> normed_end_0 = const()[name = string("normed_end_0"), val = tensor<int32, [3]>([1, 64, 1536])];
+            tensor<bool, [3]> normed_end_mask_0 = const()[name = string("normed_end_mask_0"), val = tensor<bool, [3]>([true, true, false])];
+            tensor<fp32, [1, 64, 1536]> normed = slice_by_index(begin = normed_begin_0, end = normed_end_0, end_mask = normed_end_mask_0, x = normed_1)[name = string("normed")];
+            tensor<fp32, [1536]> const_3_promoted = const()[name = string("const_3_promoted"), val = tensor<fp32, [1536]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(6296000)))];
+            tensor<fp32, [1, 64, 1536]> hidden_states_1 = mul(x = normed, y = const_3_promoted)[name = string("hidden_states")];
+            tensor<fp32, [1, 4096, 128]> var_69 = const()[name = string("op_69"), val = tensor<fp32, [1, 4096, 128]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(8399424)))];
+            int32 var_77_batch_dims_0 = const()[name = string("op_77_batch_dims_0"), val = int32(0)];
+            bool var_77_validate_indices_0 = const()[name = string("op_77_validate_indices_0"), val = bool(false)];
+            int32 greater_equal_0_y_0 = const()[name = string("greater_equal_0_y_0"), val = int32(0)];
+            tensor<bool, [64]> greater_equal_0 = greater_equal(x = position_ids, y = greater_equal_0_y_0)[name = string("greater_equal_0")];
+            int32 slice_by_index_8 = const()[name = string("slice_by_index_8"), val = int32(4096)];
+            tensor<int32, [64]> add_0 = add(x = position_ids, y = slice_by_index_8)[name = string("add_0")];
+            tensor<int32, [64]> select_0 = select(a = position_ids, b = add_0, cond = greater_equal_0)[name = string("select_0")];
+            int32 var_77_axis_1 = const()[name = string("op_77_axis_1"), val = int32(1)];
+            tensor<fp32, [1, 64, 128]> var_77 = gather(axis = var_77_axis_1, batch_dims = var_77_batch_dims_0, indices = select_0, validate_indices = var_77_validate_indices_0, x = var_69)[name = string("op_77")];
+            tensor<int32, [4]> var_81 = const()[name = string("op_81"), val = tensor<int32, [4]>([1, 64, 1, 128])];
+            tensor<fp32, [1, 64, 1, 128]> cos_1 = reshape(shape = var_81, x = var_77)[name = string("cos_1")];
+            tensor<fp32, [1, 4096, 128]> var_83 = const()[name = string("op_83"), val = tensor<fp32, [1, 4096, 128]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(6302208)))];
+            int32 var_91_batch_dims_0 = const()[name = string("op_91_batch_dims_0"), val = int32(0)];
+            bool var_91_validate_indices_0 = const()[name = string("op_91_validate_indices_0"), val = bool(false)];
+            int32 var_91_axis_1 = const()[name = string("op_91_axis_1"), val = int32(1)];
+            tensor<fp32, [1, 64, 128]> var_91 = gather(axis = var_91_axis_1, batch_dims = var_91_batch_dims_0, indices = select_0, validate_indices = var_91_validate_indices_0, x = var_83)[name = string("op_91")];
+            tensor<int32, [4]> var_95 = const()[name = string("op_95"), val = tensor<int32, [4]>([1, 64, 1, 128])];
+            tensor<fp32, [1, 64, 1, 128]> sin_1 = reshape(shape = var_95, x = var_91)[name = string("sin_1")];
+            string cos_3_dtype_0 = const()[name = string("cos_3_dtype_0"), val = string("fp16")];
+            string sin_3_dtype_0 = const()[name = string("sin_3_dtype_0"), val = string("fp16")];
+            tensor<int32, [3]> var_116 = const()[name = string("op_116"), val = tensor<int32, [3]>([0, 2, 1])];
+            tensor<int32, [1]> var_119_axes_0 = const()[name = string("op_119_axes_0"), val = tensor<int32, [1]>([2])];
+            tensor<fp32, [1, 1536, 64]> var_117 = transpose(perm = var_116, x = hidden_states_1)[name = string("transpose_10")];
+            tensor<fp32, [1, 1536, 1, 64]> var_119 = expand_dims(axes = var_119_axes_0, x = var_117)[name = string("op_119")];
+            string input_3_dtype_0 = const()[name = string("input_3_dtype_0"), val = string("fp16")];
+            string query_states_1_pad_type_0 = const()[name = string("query_states_1_pad_type_0"), val = string("valid")];
+            tensor<int32, [2]> query_states_1_strides_0 = const()[name = string("query_states_1_strides_0"), val = tensor<int32, [2]>([1, 1])];
+            tensor<int32, [4]> query_states_1_pad_0 = const()[name = string("query_states_1_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
+            tensor<int32, [2]> query_states_1_dilations_0 = const()[name = string("query_states_1_dilations_0"), val = tensor<int32, [2]>([1, 1])];
+            int32 query_states_1_groups_0 = const()[name = string("query_states_1_groups_0"), val = int32(1)];
+            tensor<fp16, [1, 1536, 1, 64]> input_3 = cast(dtype = input_3_dtype_0, x = var_119)[name = string("cast_27")];
+            tensor<fp16, [1, 1536, 1, 64]> query_states_1 = conv(bias = model_model_layers_0_self_attn_q_proj_bias, dilations = query_states_1_dilations_0, groups = query_states_1_groups_0, pad = query_states_1_pad_0, pad_type = query_states_1_pad_type_0, strides = query_states_1_strides_0, weight = model_model_layers_0_self_attn_q_proj_weight, x = input_3)[name = string("query_states_1")];
+            string key_states_1_pad_type_0 = const()[name = string("key_states_1_pad_type_0"), val = string("valid")];
+            tensor<int32, [2]> key_states_1_strides_0 = const()[name = string("key_states_1_strides_0"), val = tensor<int32, [2]>([1, 1])];
+            tensor<int32, [4]> key_states_1_pad_0 = const()[name = string("key_states_1_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
+            tensor<int32, [2]> key_states_1_dilations_0 = const()[name = string("key_states_1_dilations_0"), val = tensor<int32, [2]>([1, 1])];
+            int32 key_states_1_groups_0 = const()[name = string("key_states_1_groups_0"), val = int32(1)];
+            tensor<fp16, [1, 256, 1, 64]> key_states_1 = conv(bias = model_model_layers_0_self_attn_k_proj_bias, dilations = key_states_1_dilations_0, groups = key_states_1_groups_0, pad = key_states_1_pad_0, pad_type = key_states_1_pad_type_0, strides = key_states_1_strides_0, weight = model_model_layers_0_self_attn_k_proj_weight, x = input_3)[name = string("key_states_1")];
+            string value_states_1_pad_type_0 = const()[name = string("value_states_1_pad_type_0"), val = string("valid")];
+            tensor<int32, [2]> value_states_1_strides_0 = const()[name = string("value_states_1_strides_0"), val = tensor<int32, [2]>([1, 1])];
+            tensor<int32, [4]> value_states_1_pad_0 = const()[name = string("value_states_1_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
+            tensor<int32, [2]> value_states_1_dilations_0 = const()[name = string("value_states_1_dilations_0"), val = tensor<int32, [2]>([1, 1])];
+            int32 value_states_1_groups_0 = const()[name = string("value_states_1_groups_0"), val = int32(1)];
+            tensor<fp16, [1, 256, 1, 64]> value_states_1 = conv(bias = model_model_layers_0_self_attn_v_proj_bias, dilations = value_states_1_dilations_0, groups = value_states_1_groups_0, pad = value_states_1_pad_0, pad_type = value_states_1_pad_type_0, strides = value_states_1_strides_0, weight = model_model_layers_0_self_attn_v_proj_weight, x = input_3)[name = string("value_states_1")];
+            tensor<int32, [4]> var_161 = const()[name = string("op_161"), val = tensor<int32, [4]>([1, 12, 128, 64])];
+            tensor<fp16, [1, 12, 128, 64]> var_162 = reshape(shape = var_161, x = query_states_1)[name = string("op_162")];
+            tensor<int32, [4]> var_167 = const()[name = string("op_167"), val = tensor<int32, [4]>([0, 1, 3, 2])];
+            tensor<int32, [4]> var_172 = const()[name = string("op_172"), val = tensor<int32, [4]>([1, 2, 128, 64])];
+            tensor<fp16, [1, 2, 128, 64]> var_173 = reshape(shape = var_172, x = key_states_1)[name = string("op_173")];
+            tensor<int32, [4]> var_178 = const()[name = string("op_178"), val = tensor<int32, [4]>([0, 1, 3, 2])];
+            tensor<int32, [4]> var_183 = const()[name = string("op_183"), val = tensor<int32, [4]>([1, 2, 128, 64])];
+            tensor<fp16, [1, 2, 128, 64]> var_184 = reshape(shape = var_183, x = value_states_1)[name = string("op_184")];
+            tensor<int32, [4]> var_189 = const()[name = string("op_189"), val = tensor<int32, [4]>([0, 1, 3, 2])];
+            tensor<int32, [4]> var_195 = const()[name = string("op_195"), val = tensor<int32, [4]>([0, 2, 1, 3])];
+            tensor<int32, [4]> var_201 = const()[name = string("op_201"), val = tensor<int32, [4]>([0, 2, 1, 3])];
+            tensor<fp16, [1, 64, 1, 128]> cos_3 = cast(dtype = cos_3_dtype_0, x = cos_1)[name = string("cast_29")];
+            tensor<fp16, [1, 1, 64, 128]> cos = transpose(perm = var_195, x = cos_3)[name = string("transpose_6")];
+            tensor<fp16, [1, 12, 64, 128]> q_1 = transpose(perm = var_167, x = var_162)[name = string("transpose_9")];
+            tensor<fp16, [1, 12, 64, 128]> var_203 = mul(x = q_1, y = cos)[name = string("op_203")];
+            tensor<int32, [4]> x1_1_begin_0 = const()[name = string("x1_1_begin_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
+            tensor<int32, [4]> x1_1_end_0 = const()[name = string("x1_1_end_0"), val = tensor<int32, [4]>([1, 12, 64, 64])];
+            tensor<bool, [4]> x1_1_end_mask_0 = const()[name = string("x1_1_end_mask_0"), val = tensor<bool, [4]>([true, true, true, false])];
+            tensor<fp16, [1, 12, 64, 64]> x1_1 = slice_by_index(begin = x1_1_begin_0, end = x1_1_end_0, end_mask = x1_1_end_mask_0, x = q_1)[name = string("x1_1")];
+            tensor<int32, [4]> x2_1_begin_0 = const()[name = string("x2_1_begin_0"), val = tensor<int32, [4]>([0, 0, 0, 64])];
+            tensor<int32, [4]> x2_1_end_0 = const()[name = string("x2_1_end_0"), val = tensor<int32, [4]>([1, 12, 64, 128])];
+            tensor<bool, [4]> x2_1_end_mask_0 = const()[name = string("x2_1_end_mask_0"), val = tensor<bool, [4]>([true, true, true, true])];
+            tensor<fp16, [1, 12, 64, 64]> x2_1 = slice_by_index(begin = x2_1_begin_0, end = x2_1_end_0, end_mask = x2_1_end_mask_0, x = q_1)[name = string("x2_1")];
+            fp16 const_8_promoted = const()[name = string("const_8_promoted"), val = fp16(-0x1p+0)];
+            tensor<fp16, [1, 12, 64, 64]> var_224 = mul(x = x2_1, y = const_8_promoted)[name = string("op_224")];
+            int32 var_226 = const()[name = string("op_226"), val = int32(-1)];
+            bool var_227_interleave_0 = const()[name = string("op_227_interleave_0"), val = bool(false)];
+            tensor<fp16, [1, 12, 64, 128]> var_227 = concat(axis = var_226, interleave = var_227_interleave_0, values = (var_224, x1_1))[name = string("op_227")];
+            tensor<fp16, [1, 64, 1, 128]> sin_3 = cast(dtype = sin_3_dtype_0, x = sin_1)[name = string("cast_28")];
+            tensor<fp16, [1, 1, 64, 128]> sin = transpose(perm = var_201, x = sin_3)[name = string("transpose_5")];
+            tensor<fp16, [1, 12, 64, 128]> var_228 = mul(x = var_227, y = sin)[name = string("op_228")];
+            tensor<fp16, [1, 12, 64, 128]> query_states_3 = add(x = var_203, y = var_228)[name = string("query_states_3")];
+            tensor<fp16, [1, 2, 64, 128]> k_1 = transpose(perm = var_178, x = var_173)[name = string("transpose_8")];
+            tensor<fp16, [1, 2, 64, 128]> var_231 = mul(x = k_1, y = cos)[name = string("op_231")];
+            tensor<int32, [4]> x1_begin_0 = const()[name = string("x1_begin_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
+            tensor<int32, [4]> x1_end_0 = const()[name = string("x1_end_0"), val = tensor<int32, [4]>([1, 2, 64, 64])];
+            tensor<bool, [4]> x1_end_mask_0 = const()[name = string("x1_end_mask_0"), val = tensor<bool, [4]>([true, true, true, false])];
+            tensor<fp16, [1, 2, 64, 64]> x1 = slice_by_index(begin = x1_begin_0, end = x1_end_0, end_mask = x1_end_mask_0, x = k_1)[name = string("x1")];
+            tensor<int32, [4]> x2_begin_0 = const()[name = string("x2_begin_0"), val = tensor<int32, [4]>([0, 0, 0, 64])];
+            tensor<int32, [4]> x2_end_0 = const()[name = string("x2_end_0"), val = tensor<int32, [4]>([1, 2, 64, 128])];
+            tensor<bool, [4]> x2_end_mask_0 = const()[name = string("x2_end_mask_0"), val = tensor<bool, [4]>([true, true, true, true])];
+            tensor<fp16, [1, 2, 64, 64]> x2 = slice_by_index(begin = x2_begin_0, end = x2_end_0, end_mask = x2_end_mask_0, x = k_1)[name = string("x2")];
+            fp16 const_11_promoted = const()[name = string("const_11_promoted"), val = fp16(-0x1p+0)];
+            tensor<fp16, [1, 2, 64, 64]> var_252 = mul(x = x2, y = const_11_promoted)[name = string("op_252")];
+            int32 var_254 = const()[name = string("op_254"), val = int32(-1)];
+            bool var_255_interleave_0 = const()[name = string("op_255_interleave_0"), val = bool(false)];
+            tensor<fp16, [1, 2, 64, 128]> var_255 = concat(axis = var_254, interleave = var_255_interleave_0, values = (var_252, x1))[name = string("op_255")];
+            tensor<fp16, [1, 2, 64, 128]> var_256 = mul(x = var_255, y = sin)[name = string("op_256")];
+            tensor<fp16, [1, 2, 64, 128]> key_states_3 = add(x = var_231, y = var_256)[name = string("key_states_3")];
+            tensor<int32, [1]> seq_length = const()[name = string("seq_length"), val = tensor<int32, [1]>([64])];
+            tensor<int32, [1]> var_278 = add(x = current_pos, y = seq_length)[name = string("op_278")];
+            tensor<fp16, [56, 2, 2048, 128]> read_state_0 = read_state(input = model_model_kv_cache_0)[name = string("read_state_0")];
+            string cast_7_dtype_0 = const()[name = string("cast_7_dtype_0"), val = string("fp32")];
+            tensor<int32, [1]> expand_dims_0 = const()[name = string("expand_dims_0"), val = tensor<int32, [1]>([0])];
+            tensor<int32, [1]> expand_dims_1 = const()[name = string("expand_dims_1"), val = tensor<int32, [1]>([0])];
+            tensor<int32, [1]> expand_dims_3 = const()[name = string("expand_dims_3"), val = tensor<int32, [1]>([0])];
+            tensor<int32, [1]> expand_dims_4 = const()[name = string("expand_dims_4"), val = tensor<int32, [1]>([1])];
+            int32 concat_2_axis_0 = const()[name = string("concat_2_axis_0"), val = int32(0)];
+            bool concat_2_interleave_0 = const()[name = string("concat_2_interleave_0"), val = bool(false)];
+            tensor<int32, [4]> concat_2 = concat(axis = concat_2_axis_0, interleave = concat_2_interleave_0, values = (expand_dims_0, expand_dims_1, current_pos, expand_dims_3))[name = string("concat_2")];
+            tensor<int32, [1]> concat_3_values1_0 = const()[name = string("concat_3_values1_0"), val = tensor<int32, [1]>([0])];
+            tensor<int32, [1]> concat_3_values3_0 = const()[name = string("concat_3_values3_0"), val = tensor<int32, [1]>([0])];
+            int32 concat_3_axis_0 = const()[name = string("concat_3_axis_0"), val = int32(0)];
+            bool concat_3_interleave_0 = const()[name = string("concat_3_interleave_0"), val = bool(false)];
+            tensor<int32, [4]> concat_3 = concat(axis = concat_3_axis_0, interleave = concat_3_interleave_0, values = (expand_dims_4, concat_3_values1_0, var_278, concat_3_values3_0))[name = string("concat_3")];
+            string key_states_5_promoted_dtype_0 = const()[name = string("key_states_5_promoted_dtype_0"), val = string("fp32")];
+            tensor<int32, [4]> model_model_kv_cache_0_internal_tensor_assign_1_stride_0 = const()[name = string("model_model_kv_cache_0_internal_tensor_assign_1_stride_0"), val = tensor<int32, [4]>([1, 1, 1, 1])];
+            tensor<bool, [4]> model_model_kv_cache_0_internal_tensor_assign_1_begin_mask_0 = const()[name = string("model_model_kv_cache_0_internal_tensor_assign_1_begin_mask_0"), val = tensor<bool, [4]>([false, false, false, false])];
+            tensor<bool, [4]> model_model_kv_cache_0_internal_tensor_assign_1_end_mask_0 = const()[name = string("model_model_kv_cache_0_internal_tensor_assign_1_end_mask_0"), val = tensor<bool, [4]>([false, true, false, true])];
+            tensor<bool, [4]> model_model_kv_cache_0_internal_tensor_assign_1_squeeze_mask_0 = const()[name = string("model_model_kv_cache_0_internal_tensor_assign_1_squeeze_mask_0"), val = tensor<bool, [4]>([false, false, false, false])];
+            tensor<fp32, [1, 2, 64, 128]> key_states_5_promoted = cast(dtype = key_states_5_promoted_dtype_0, x = key_states_3)[name = string("cast_25")];
+            tensor<fp32, [56, 2, 2048, 128]> cast_7 = cast(dtype = cast_7_dtype_0, x = read_state_0)[name = string("cast_26")];
+            tensor<fp32, [56, 2, 2048, 128]> model_model_kv_cache_0_internal_tensor_assign_1 = slice_update(begin = concat_2, begin_mask = model_model_kv_cache_0_internal_tensor_assign_1_begin_mask_0, end = concat_3, end_mask = model_model_kv_cache_0_internal_tensor_assign_1_end_mask_0, squeeze_mask = model_model_kv_cache_0_internal_tensor_assign_1_squeeze_mask_0, stride = model_model_kv_cache_0_internal_tensor_assign_1_stride_0, update = key_states_5_promoted, x = cast_7)[name = string("model_model_kv_cache_0_internal_tensor_assign_1")];
+            string cast_8_dtype_0 = const()[name = string("cast_8_dtype_0"), val = string("fp16")];
+            tensor<fp16, [56, 2, 2048, 128]> cast_8 = cast(dtype = cast_8_dtype_0, x = model_model_kv_cache_0_internal_tensor_assign_1)[name = string("cast_24")];
+            write_state(data = cast_8, input = model_model_kv_cache_0)[name = string("coreml_update_state_6_write_state")];
+            tensor<fp16, [56, 2, 2048, 128]> coreml_update_state_2 = read_state(input = model_model_kv_cache_0)[name = string("coreml_update_state_6")];
+            string model_model_kv_cache_0_internal_tensor_assign_1_dtype_0 = const()[name = string("model_model_kv_cache_0_internal_tensor_assign_1_dtype_0"), val = string("fp32")];
+            tensor<int32, [1]> expand_dims_6 = const()[name = string("expand_dims_6"), val = tensor<int32, [1]>([28])];
+            tensor<int32, [1]> expand_dims_7 = const()[name = string("expand_dims_7"), val = tensor<int32, [1]>([0])];
+            tensor<int32, [1]> expand_dims_9 = const()[name = string("expand_dims_9"), val = tensor<int32, [1]>([0])];
+            tensor<int32, [1]> expand_dims_10 = const()[name = string("expand_dims_10"), val = tensor<int32, [1]>([29])];
+            int32 concat_6_axis_0 = const()[name = string("concat_6_axis_0"), val = int32(0)];
+            bool concat_6_interleave_0 = const()[name = string("concat_6_interleave_0"), val = bool(false)];
+            tensor<int32, [4]> concat_6 = concat(axis = concat_6_axis_0, interleave = concat_6_interleave_0, values = (expand_dims_6, expand_dims_7, current_pos, expand_dims_9))[name = string("concat_6")];
+            tensor<int32, [1]> concat_7_values1_0 = const()[name = string("concat_7_values1_0"), val = tensor<int32, [1]>([0])];
+            tensor<int32, [1]> concat_7_values3_0 = const()[name = string("concat_7_values3_0"), val = tensor<int32, [1]>([0])];
+            int32 concat_7_axis_0 = const()[name = string("concat_7_axis_0"), val = int32(0)];
+            bool concat_7_interleave_0 = const()[name = string("concat_7_interleave_0"), val = bool(false)];
+            tensor<int32, [4]> concat_7 = concat(axis = concat_7_axis_0, interleave = concat_7_interleave_0, values = (expand_dims_10, concat_7_values1_0, var_278, concat_7_values3_0))[name = string("concat_7")];
+            string value_states_5_promoted_dtype_0 = const()[name = string("value_states_5_promoted_dtype_0"), val = string("fp32")];
+            tensor<int32, [4]> model_model_kv_cache_0_internal_tensor_assign_2_stride_0 = const()[name = string("model_model_kv_cache_0_internal_tensor_assign_2_stride_0"), val = tensor<int32, [4]>([1, 1, 1, 1])];
+            tensor<bool, [4]> model_model_kv_cache_0_internal_tensor_assign_2_begin_mask_0 = const()[name = string("model_model_kv_cache_0_internal_tensor_assign_2_begin_mask_0"), val = tensor<bool, [4]>([false, false, false, false])];
+            tensor<bool, [4]> model_model_kv_cache_0_internal_tensor_assign_2_end_mask_0 = const()[name = string("model_model_kv_cache_0_internal_tensor_assign_2_end_mask_0"), val = tensor<bool, [4]>([false, true, false, true])];
+            tensor<bool, [4]> model_model_kv_cache_0_internal_tensor_assign_2_squeeze_mask_0 = const()[name = string("model_model_kv_cache_0_internal_tensor_assign_2_squeeze_mask_0"), val = tensor<bool, [4]>([false, false, false, false])];
+            tensor<fp16, [1, 2, 64, 128]> value_states_3 = transpose(perm = var_189, x = var_184)[name = string("transpose_7")];
+            tensor<fp32, [1, 2, 64, 128]> value_states_5_promoted = cast(dtype = value_states_5_promoted_dtype_0, x = value_states_3)[name = string("cast_22")];
+            tensor<fp32, [56, 2, 2048, 128]> model_model_kv_cache_0_internal_tensor_assign_1_1 = cast(dtype = model_model_kv_cache_0_internal_tensor_assign_1_dtype_0, x = coreml_update_state_2)[name = string("cast_23")];
+            tensor<fp32, [56, 2, 2048, 128]> model_model_kv_cache_0_internal_tensor_assign_2 = slice_update(begin = concat_6, begin_mask = model_model_kv_cache_0_internal_tensor_assign_2_begin_mask_0, end = concat_7, end_mask = model_model_kv_cache_0_internal_tensor_assign_2_end_mask_0, squeeze_mask = model_model_kv_cache_0_internal_tensor_assign_2_squeeze_mask_0, stride = model_model_kv_cache_0_internal_tensor_assign_2_stride_0, update = value_states_5_promoted, x = model_model_kv_cache_0_internal_tensor_assign_1_1)[name = string("model_model_kv_cache_0_internal_tensor_assign_2")];
+            string cast_9_dtype_0 = const()[name = string("cast_9_dtype_0"), val = string("fp16")];
+            tensor<fp16, [56, 2, 2048, 128]> cast_9 = cast(dtype = cast_9_dtype_0, x = model_model_kv_cache_0_internal_tensor_assign_2)[name = string("cast_21")];
+            write_state(data = cast_9, input = model_model_kv_cache_0)[name = string("coreml_update_state_7_write_state")];
+            tensor<fp16, [56, 2, 2048, 128]> coreml_update_state_3 = read_state(input = model_model_kv_cache_0)[name = string("coreml_update_state_7")];
+            string model_model_kv_cache_0_internal_tensor_assign_2_dtype_0 = const()[name = string("model_model_kv_cache_0_internal_tensor_assign_2_dtype_0"), val = string("fp32")];
+            tensor<int32, [4]> var_327_begin_0 = const()[name = string("op_327_begin_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
+            tensor<int32, [4]> var_327_end_0 = const()[name = string("op_327_end_0"), val = tensor<int32, [4]>([1, 2, 2048, 128])];
+            tensor<bool, [4]> var_327_end_mask_0 = const()[name = string("op_327_end_mask_0"), val = tensor<bool, [4]>([false, true, true, true])];
+            tensor<fp32, [56, 2, 2048, 128]> model_model_kv_cache_0_internal_tensor_assign_2_1 = cast(dtype = model_model_kv_cache_0_internal_tensor_assign_2_dtype_0, x = coreml_update_state_3)[name = string("cast_20")];
+            tensor<fp32, [1, 2, 2048, 128]> var_327 = slice_by_index(begin = var_327_begin_0, end = var_327_end_0, end_mask = var_327_end_mask_0, x = model_model_kv_cache_0_internal_tensor_assign_2_1)[name = string("op_327")];
+            tensor<int32, [1]> K_layer_cache_axes_0 = const()[name = string("K_layer_cache_axes_0"), val = tensor<int32, [1]>([0])];
+            tensor<fp32, [2, 2048, 128]> K_layer_cache = squeeze(axes = K_layer_cache_axes_0, x = var_327)[name = string("K_layer_cache")];
+            tensor<int32, [4]> var_334_begin_0 = const()[name = string("op_334_begin_0"), val = tensor<int32, [4]>([28, 0, 0, 0])];
+            tensor<int32, [4]> var_334_end_0 = const()[name = string("op_334_end_0"), val = tensor<int32, [4]>([29, 2, 2048, 128])];
+            tensor<bool, [4]> var_334_end_mask_0 = const()[name = string("op_334_end_mask_0"), val = tensor<bool, [4]>([false, true, true, true])];
+            tensor<fp32, [1, 2, 2048, 128]> var_334 = slice_by_index(begin = var_334_begin_0, end = var_334_end_0, end_mask = var_334_end_mask_0, x = model_model_kv_cache_0_internal_tensor_assign_2_1)[name = string("op_334")];
+            tensor<int32, [1]> V_layer_cache_axes_0 = const()[name = string("V_layer_cache_axes_0"), val = tensor<int32, [1]>([0])];
+            tensor<fp32, [2, 2048, 128]> V_layer_cache = squeeze(axes = V_layer_cache_axes_0, x = var_334)[name = string("V_layer_cache")];
+            tensor<int32, [1]> x_3_axes_0 = const()[name = string("x_3_axes_0"), val = tensor<int32, [1]>([1])];
+            tensor<fp32, [2, 1, 2048, 128]> x_3 = expand_dims(axes = x_3_axes_0, x = K_layer_cache)[name = string("x_3")];
+            tensor<int32, [4]> var_363 = const()[name = string("op_363"), val = tensor<int32, [4]>([1, 6, 1, 1])];
+            tensor<fp32, [2, 6, 2048, 128]> x_5 = tile(reps = var_363, x = x_3)[name = string("x_5")];
+            tensor<int32, [4]> var_375 = const()[name = string("op_375"), val = tensor<int32, [4]>([1, -1, 2048, 128])];
+            tensor<fp32, [1, 12, 2048, 128]> key_states = reshape(shape = var_375, x = x_5)[name = string("key_states")];
+            tensor<int32, [1]> x_9_axes_0 = const()[name = string("x_9_axes_0"), val = tensor<int32, [1]>([1])];
+            tensor<fp32, [2, 1, 2048, 128]> x_9 = expand_dims(axes = x_9_axes_0, x = V_layer_cache)[name = string("x_9")];
+            tensor<int32, [4]> var_383 = const()[name = string("op_383"), val = tensor<int32, [4]>([1, 6, 1, 1])];
+            tensor<fp32, [2, 6, 2048, 128]> x = tile(reps = var_383, x = x_9)[name = string("x")];
+            string q_dtype_0 = const()[name = string("q_dtype_0"), val = string("fp32")];
+            bool var_418_transpose_x_1 = const()[name = string("op_418_transpose_x_1"), val = bool(false)];
+            bool var_418_transpose_y_1 = const()[name = string("op_418_transpose_y_1"), val = bool(true)];
+            tensor<fp32, [1, 12, 64, 128]> q = cast(dtype = q_dtype_0, x = query_states_3)[name = string("cast_19")];
+            tensor<fp32, [1, 12, 64, 2048]> var_418 = matmul(transpose_x = var_418_transpose_x_1, transpose_y = var_418_transpose_y_1, x = q, y = key_states)[name = string("op_418")];
+            fp32 var_419 = const()[name = string("op_419"), val = fp32(0x1.6a09e6p-4)];
+            tensor<fp32, [1, 12, 64, 2048]> attn_logits_1 = mul(x = var_418, y = var_419)[name = string("attn_logits_1")];
+            tensor<fp32, [1, 1, 64, 2048]> cast_1 = cast(dtype = cast_1_dtype_0, x = causal_mask)[name = string("cast_30")];
+            tensor<fp32, [1, 12, 64, 2048]> attn_logits = add(x = attn_logits_1, y = cast_1)[name = string("attn_logits")];
+            int32 var_446 = const()[name = string("op_446"), val = int32(-1)];
+            tensor<fp32, [1, 12, 64, 2048]> var_448 = softmax(axis = var_446, x = attn_logits)[name = string("op_448")];
+            string attn_weights_dtype_0 = const()[name = string("attn_weights_dtype_0"), val = string("fp16")];
+            string var_458_dtype_0 = const()[name = string("op_458_dtype_0"), val = string("fp32")];
+            tensor<int32, [3]> concat_12 = const()[name = string("concat_12"), val = tensor<int32, [3]>([12, 64, 2048])];
+            tensor<fp16, [1, 12, 64, 2048]> attn_weights = cast(dtype = attn_weights_dtype_0, x = var_448)[name = string("cast_18")];
+            tensor<fp32, [1, 12, 64, 2048]> var_458 = cast(dtype = var_458_dtype_0, x = attn_weights)[name = string("cast_17")];
+            tensor<fp32, [12, 64, 2048]> reshape_0 = reshape(shape = concat_12, x = var_458)[name = string("reshape_0")];
+            tensor<int32, [3]> concat_13 = const()[name = string("concat_13"), val = tensor<int32, [3]>([12, 2048, 128])];
+            tensor<fp32, [12, 2048, 128]> reshape_1 = reshape(shape = concat_13, x = x)[name = string("reshape_1")];
+            bool matmul_0_transpose_x_0 = const()[name = string("matmul_0_transpose_x_0"), val = bool(false)];
+            bool matmul_0_transpose_y_0 = const()[name = string("matmul_0_transpose_y_0"), val = bool(false)];
+            tensor<fp32, [12, 64, 128]> matmul_0 = matmul(transpose_x = matmul_0_transpose_x_0, transpose_y = matmul_0_transpose_y_0, x = reshape_0, y = reshape_1)[name = string("matmul_0")];
+            tensor<int32, [4]> concat_17 = const()[name = string("concat_17"), val = tensor<int32, [4]>([1, 12, 64, 128])];
+            tensor<fp32, [1, 12, 64, 128]> reshape_2 = reshape(shape = concat_17, x = matmul_0)[name = string("reshape_2")];
+            string attn_output_1_dtype_0 = const()[name = string("attn_output_1_dtype_0"), val = string("fp16")];
+            tensor<int32, [4]> var_475_perm_0 = const()[name = string("op_475_perm_0"), val = tensor<int32, [4]>([0, 2, 1, 3])];
+            tensor<int32, [3]> var_494 = const()[name = string("op_494"), val = tensor<int32, [3]>([1, 64, 1536])];
+            tensor<fp16, [1, 12, 64, 128]> attn_output_1 = cast(dtype = attn_output_1_dtype_0, x = reshape_2)[name = string("cast_16")];
+            tensor<fp16, [1, 64, 12, 128]> var_475 = transpose(perm = var_475_perm_0, x = attn_output_1)[name = string("transpose_4")];
+            tensor<fp16, [1, 64, 1536]> attn_output_5 = reshape(shape = var_494, x = var_475)[name = string("attn_output_5")];
+            tensor<int32, [3]> var_499 = const()[name = string("op_499"), val = tensor<int32, [3]>([0, 2, 1])];
+            tensor<fp16, [1536, 1536, 1]> squeeze_0 = const()[name = string("squeeze_0"), val = tensor<fp16, [1536, 1536, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(10496640)))];
+            string var_515_pad_type_0 = const()[name = string("op_515_pad_type_0"), val = string("valid")];
+            int32 var_515_groups_0 = const()[name = string("op_515_groups_0"), val = int32(1)];
+            tensor<int32, [1]> var_515_strides_0 = const()[name = string("op_515_strides_0"), val = tensor<int32, [1]>([1])];
+            tensor<int32, [2]> var_515_pad_0 = const()[name = string("op_515_pad_0"), val = tensor<int32, [2]>([0, 0])];
+            tensor<int32, [1]> var_515_dilations_0 = const()[name = string("op_515_dilations_0"), val = tensor<int32, [1]>([1])];
+            tensor<fp16, [1, 1536, 64]> var_500 = transpose(perm = var_499, x = attn_output_5)[name = string("transpose_3")];
+            tensor<fp16, [1, 1536, 64]> var_515 = conv(dilations = var_515_dilations_0, groups = var_515_groups_0, pad = var_515_pad_0, pad_type = var_515_pad_type_0, strides = var_515_strides_0, weight = squeeze_0, x = var_500)[name = string("op_515")];
+            tensor<int32, [3]> var_519 = const()[name = string("op_519"), val = tensor<int32, [3]>([0, 2, 1])];
+            string attn_out_promoted_dtype_0 = const()[name = string("attn_out_promoted_dtype_0"), val = string("fp32")];
+            tensor<fp16, [1, 64, 1536]> attn_out = transpose(perm = var_519, x = var_515)[name = string("transpose_2")];
+            tensor<fp32, [1, 64, 1536]> attn_out_promoted = cast(dtype = attn_out_promoted_dtype_0, x = attn_out)[name = string("cast_15")];
+            tensor<fp32, [1, 64, 1536]> output_hidden_states_type_fp32 = add(x = cast_0, y = attn_out_promoted)[name = string("op_522")];
+            string cast_13_dtype_0 = const()[name = string("cast_13_dtype_0"), val = string("fp16")];
+            tensor<fp16, [1, 64, 1536]> output_hidden_states = cast(dtype = cast_13_dtype_0, x = output_hidden_states_type_fp32)[name = string("cast_14")];
+        } -> (output_hidden_states);
+}

qwen25_FFN_PF_lut6_chunk_01of04.mlmodelc/weights/weight.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d314de0a6d18482f94083f7823f2f18e20c4235d3ab352572e023e1bd3bf066b
+size 15215296

qwen25_FFN_PF_lut6_chunk_02of04.mlmodelc/analytics/coremldata.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a14f122173665390a88e43c7657f13ae0fdb992c8458f5bc2829e1d4b8f6e09d
+size 243

qwen25_FFN_PF_lut6_chunk_02of04.mlmodelc/coremldata.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1aa2254d0b7cb3cdc9bcf1ee539fed260d554dc0563f3f650138d581dd7c2f59
+size 542

qwen25_FFN_PF_lut6_chunk_02of04.mlmodelc/metadata.json

@@ -0,0 +1,304 @@
+[
+  {
+    "metadataOutputVersion" : "3.0",
+    "storagePrecision" : "Mixed (Float16, Palettized (12 bits), Palettized (15 bits), Palettized (18 bits), UInt6)",
+    "outputSchema" : [
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 1536)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 1536]",
+        "name" : "output_hidden_states",
+        "type" : "MultiArray"
+      }
+    ],
+    "modelParameters" : [
+
+    ],
+    "specificationVersion" : 9,
+    "functions" : [
+      {
+        "inputSchema" : [
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 1536)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 1536]",
+            "name" : "hidden_states",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Int32",
+            "formattedType" : "MultiArray (Int32 1)",
+            "shortDescription" : "",
+            "shape" : "[1]",
+            "name" : "position_ids",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 2048)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 1, 2048]",
+            "name" : "causal_mask",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Int32",
+            "formattedType" : "MultiArray (Int32 1)",
+            "shortDescription" : "",
+            "shape" : "[1]",
+            "name" : "current_pos",
+            "type" : "MultiArray"
+          }
+        ],
+        "computePrecision" : "Mixed (Float16, Int32, UInt16)",
+        "storagePrecision" : "Mixed (Float16, Palettized (12 bits), Palettized (15 bits), Palettized (18 bits), UInt6)",
+        "stateSchema" : [
+          {
+            "dataType" : "Float16",
+            "isOptional" : "0",
+            "formattedType" : "State (Float16 56 × 2 × 2048 × 128)",
+            "shortDescription" : "",
+            "shape" : "[56, 2, 2048, 128]",
+            "name" : "model_model_kv_cache_0",
+            "type" : "State"
+          }
+        ],
+        "outputSchema" : [
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 1536)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 1536]",
+            "name" : "output_hidden_states",
+            "type" : "MultiArray"
+          }
+        ],
+        "name" : "infer",
+        "mlProgramOperationTypeHistogram" : {
+          "Ios18.expandDims" : 37,
+          "Ios18.softmax" : 9,
+          "Ios18.mul" : 111,
+          "Ios18.matmul" : 18,
+          "Identity" : 1,
+          "Ios18.readState" : 19,
+          "Tile" : 18,
+          "Ios18.gather" : 2,
+          "Ios18.add" : 47,
+          "Ios18.layerNorm" : 19,
+          "Ios18.sliceUpdate" : 18,
+          "Ios18.reshape" : 56,
+          "Ios18.writeState" : 18,
+          "Ios18.constexprLutToDense" : 66,
+          "Ios18.conv" : 66,
+          "Ios18.concat" : 73,
+          "Ios18.silu" : 10,
+          "Ios18.cast" : 1,
+          "Ios18.transpose" : 56,
+          "Ios18.sliceByIndex" : 73,
+          "Ios18.squeeze" : 28
+        }
+      },
+      {
+        "inputSchema" : [
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 64 × 1536)",
+            "shortDescription" : "",
+            "shape" : "[1, 64, 1536]",
+            "name" : "hidden_states",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Int32",
+            "formattedType" : "MultiArray (Int32 64)",
+            "shortDescription" : "",
+            "shape" : "[64]",
+            "name" : "position_ids",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 64 × 2048)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 64, 2048]",
+            "name" : "causal_mask",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Int32",
+            "formattedType" : "MultiArray (Int32 1)",
+            "shortDescription" : "",
+            "shape" : "[1]",
+            "name" : "current_pos",
+            "type" : "MultiArray"
+          }
+        ],
+        "computePrecision" : "Mixed (Float16, Int32, UInt16)",
+        "storagePrecision" : "Mixed (Float16, Palettized (12 bits), Palettized (15 bits), Palettized (18 bits), UInt6)",
+        "stateSchema" : [
+          {
+            "dataType" : "Float16",
+            "isOptional" : "0",
+            "formattedType" : "State (Float16 56 × 2 × 2048 × 128)",
+            "shortDescription" : "",
+            "shape" : "[56, 2, 2048, 128]",
+            "name" : "model_model_kv_cache_0",
+            "type" : "State"
+          }
+        ],
+        "outputSchema" : [
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 64 × 1536)",
+            "shortDescription" : "",
+            "shape" : "[1, 64, 1536]",
+            "name" : "output_hidden_states",
+            "type" : "MultiArray"
+          }
+        ],
+        "name" : "prefill",
+        "mlProgramOperationTypeHistogram" : {
+          "Ios18.expandDims" : 37,
+          "Ios18.softmax" : 9,
+          "Ios18.mul" : 111,
+          "Ios18.matmul" : 18,
+          "Ios18.readState" : 19,
+          "Tile" : 18,
+          "Ios18.gather" : 2,
+          "Ios18.add" : 47,
+          "Ios18.layerNorm" : 19,
+          "Ios18.sliceUpdate" : 18,
+          "Ios18.reshape" : 74,
+          "Ios18.writeState" : 18,
+          "Ios18.constexprLutToDense" : 66,
+          "Ios18.conv" : 66,
+          "Ios18.concat" : 73,
+          "Ios18.silu" : 10,
+          "Ios18.cast" : 1,
+          "Ios18.transpose" : 85,
+          "Ios18.sliceByIndex" : 73,
+          "Ios18.squeeze" : 28
+        }
+      }
+    ],
+    "mlProgramOperationTypeHistogram" : {
+      "Ios18.expandDims" : 37,
+      "Ios18.softmax" : 9,
+      "Ios18.mul" : 111,
+      "Ios18.matmul" : 18,
+      "Identity" : 1,
+      "Ios18.readState" : 19,
+      "Tile" : 18,
+      "Ios18.gather" : 2,
+      "Ios18.add" : 47,
+      "Ios18.layerNorm" : 19,
+      "Ios18.sliceUpdate" : 18,
+      "Ios18.reshape" : 56,
+      "Ios18.writeState" : 18,
+      "Ios18.constexprLutToDense" : 66,
+      "Ios18.conv" : 66,
+      "Ios18.concat" : 73,
+      "Ios18.silu" : 10,
+      "Ios18.cast" : 1,
+      "Ios18.transpose" : 56,
+      "Ios18.sliceByIndex" : 73,
+      "Ios18.squeeze" : 28
+    },
+    "isUpdatable" : "0",
+    "stateSchema" : [
+      {
+        "dataType" : "Float16",
+        "isOptional" : "0",
+        "formattedType" : "State (Float16 56 × 2 × 2048 × 128)",
+        "shortDescription" : "",
+        "shape" : "[56, 2, 2048, 128]",
+        "name" : "model_model_kv_cache_0",
+        "type" : "State"
+      }
+    ],
+    "availability" : {
+      "macOS" : "15.0",
+      "tvOS" : "18.0",
+      "visionOS" : "2.0",
+      "watchOS" : "11.0",
+      "iOS" : "18.0",
+      "macCatalyst" : "18.0"
+    },
+    "computePrecision" : "Mixed (Float16, Int32, UInt16)",
+    "modelType" : {
+      "name" : "MLModelType_mlProgram"
+    },
+    "inputSchema" : [
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 1536)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 1536]",
+        "name" : "hidden_states",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Int32",
+        "formattedType" : "MultiArray (Int32 1)",
+        "shortDescription" : "",
+        "shape" : "[1]",
+        "name" : "position_ids",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 2048)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 1, 2048]",
+        "name" : "causal_mask",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Int32",
+        "formattedType" : "MultiArray (Int32 1)",
+        "shortDescription" : "",
+        "shape" : "[1]",
+        "name" : "current_pos",
+        "type" : "MultiArray"
+      }
+    ],
+    "defaultFunctionName" : "infer",
+    "generatedClassName" : "qwen25_FFN_PF_lut6_chunk_02of04",
+    "userDefinedMetadata" : {
+
+    },
+    "method" : "predict"
+  }
+]

qwen25_FFN_PF_lut6_chunk_02of04.mlmodelc/weights/weight.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a06512dcf66a966f683d674d36fea9915f3e09e74b336a6c057e58fb416f2790
+size 356278592

qwen25_FFN_PF_lut6_chunk_03of04.mlmodelc/analytics/coremldata.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c4d4436f03fb7a5f92fdb0bfaf8b347ac9fa2cf858630616487c00dae6397d0b
+size 243

qwen25_FFN_PF_lut6_chunk_03of04.mlmodelc/coremldata.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1aa2254d0b7cb3cdc9bcf1ee539fed260d554dc0563f3f650138d581dd7c2f59
+size 542

qwen25_FFN_PF_lut6_chunk_03of04.mlmodelc/metadata.json

@@ -0,0 +1,310 @@
+[
+  {
+    "metadataOutputVersion" : "3.0",
+    "storagePrecision" : "Mixed (Float16, Palettized (12 bits), Palettized (15 bits), Palettized (18 bits), UInt6)",
+    "outputSchema" : [
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 1536)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 1536]",
+        "name" : "output_hidden_states",
+        "type" : "MultiArray"
+      }
+    ],
+    "modelParameters" : [
+
+    ],
+    "specificationVersion" : 9,
+    "functions" : [
+      {
+        "inputSchema" : [
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 1536)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 1536]",
+            "name" : "hidden_states",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Int32",
+            "formattedType" : "MultiArray (Int32 1)",
+            "shortDescription" : "",
+            "shape" : "[1]",
+            "name" : "position_ids",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 2048)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 1, 2048]",
+            "name" : "causal_mask",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Int32",
+            "formattedType" : "MultiArray (Int32 1)",
+            "shortDescription" : "",
+            "shape" : "[1]",
+            "name" : "current_pos",
+            "type" : "MultiArray"
+          }
+        ],
+        "computePrecision" : "Mixed (Float16, Int16, Int32, UInt16)",
+        "storagePrecision" : "Mixed (Float16, Palettized (12 bits), Palettized (15 bits), Palettized (18 bits), UInt6)",
+        "stateSchema" : [
+          {
+            "dataType" : "Float16",
+            "isOptional" : "0",
+            "formattedType" : "State (Float16 56 × 2 × 2048 × 128)",
+            "shortDescription" : "",
+            "shape" : "[56, 2, 2048, 128]",
+            "name" : "model_model_kv_cache_0",
+            "type" : "State"
+          }
+        ],
+        "outputSchema" : [
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 1536)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 1536]",
+            "name" : "output_hidden_states",
+            "type" : "MultiArray"
+          }
+        ],
+        "name" : "infer",
+        "mlProgramOperationTypeHistogram" : {
+          "Ios18.expandDims" : 36,
+          "Ios18.mul" : 108,
+          "Ios18.softmax" : 9,
+          "Ios18.matmul" : 18,
+          "Identity" : 1,
+          "Ios18.greaterEqual" : 2,
+          "Select" : 2,
+          "Ios18.readState" : 19,
+          "Tile" : 18,
+          "Ios18.gather" : 2,
+          "Ios18.add" : 48,
+          "Ios18.layerNorm" : 18,
+          "Ios18.sliceUpdate" : 18,
+          "Ios18.writeState" : 18,
+          "Ios18.reshape" : 56,
+          "Ios18.constexprLutToDense" : 63,
+          "Ios18.conv" : 63,
+          "Ios18.concat" : 72,
+          "Ios18.transpose" : 54,
+          "Ios18.cast" : 5,
+          "Ios18.silu" : 9,
+          "Ios18.sliceByIndex" : 72,
+          "Ios18.squeeze" : 27
+        }
+      },
+      {
+        "inputSchema" : [
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 64 × 1536)",
+            "shortDescription" : "",
+            "shape" : "[1, 64, 1536]",
+            "name" : "hidden_states",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Int32",
+            "formattedType" : "MultiArray (Int32 64)",
+            "shortDescription" : "",
+            "shape" : "[64]",
+            "name" : "position_ids",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 64 × 2048)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 64, 2048]",
+            "name" : "causal_mask",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Int32",
+            "formattedType" : "MultiArray (Int32 1)",
+            "shortDescription" : "",
+            "shape" : "[1]",
+            "name" : "current_pos",
+            "type" : "MultiArray"
+          }
+        ],
+        "computePrecision" : "Mixed (Float16, Int16, Int32, UInt16)",
+        "storagePrecision" : "Mixed (Float16, Palettized (12 bits), Palettized (15 bits), Palettized (18 bits), UInt6)",
+        "stateSchema" : [
+          {
+            "dataType" : "Float16",
+            "isOptional" : "0",
+            "formattedType" : "State (Float16 56 × 2 × 2048 × 128)",
+            "shortDescription" : "",
+            "shape" : "[56, 2, 2048, 128]",
+            "name" : "model_model_kv_cache_0",
+            "type" : "State"
+          }
+        ],
+        "outputSchema" : [
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 64 × 1536)",
+            "shortDescription" : "",
+            "shape" : "[1, 64, 1536]",
+            "name" : "output_hidden_states",
+            "type" : "MultiArray"
+          }
+        ],
+        "name" : "prefill",
+        "mlProgramOperationTypeHistogram" : {
+          "Ios18.expandDims" : 36,
+          "Ios18.mul" : 108,
+          "Ios18.softmax" : 9,
+          "Ios18.matmul" : 18,
+          "Ios18.greaterEqual" : 2,
+          "Select" : 2,
+          "Ios18.readState" : 19,
+          "Tile" : 18,
+          "Ios18.gather" : 2,
+          "Ios18.add" : 48,
+          "Ios18.layerNorm" : 18,
+          "Ios18.sliceUpdate" : 18,
+          "Ios18.writeState" : 18,
+          "Ios18.reshape" : 74,
+          "Ios18.constexprLutToDense" : 63,
+          "Ios18.conv" : 63,
+          "Ios18.concat" : 72,
+          "Ios18.transpose" : 83,
+          "Ios18.cast" : 5,
+          "Ios18.silu" : 9,
+          "Ios18.sliceByIndex" : 72,
+          "Ios18.squeeze" : 27
+        }
+      }
+    ],
+    "mlProgramOperationTypeHistogram" : {
+      "Ios18.expandDims" : 36,
+      "Ios18.mul" : 108,
+      "Ios18.softmax" : 9,
+      "Ios18.matmul" : 18,
+      "Identity" : 1,
+      "Ios18.greaterEqual" : 2,
+      "Select" : 2,
+      "Ios18.readState" : 19,
+      "Tile" : 18,
+      "Ios18.gather" : 2,
+      "Ios18.add" : 48,
+      "Ios18.layerNorm" : 18,
+      "Ios18.sliceUpdate" : 18,
+      "Ios18.writeState" : 18,
+      "Ios18.reshape" : 56,
+      "Ios18.constexprLutToDense" : 63,
+      "Ios18.conv" : 63,
+      "Ios18.concat" : 72,
+      "Ios18.transpose" : 54,
+      "Ios18.cast" : 5,
+      "Ios18.silu" : 9,
+      "Ios18.sliceByIndex" : 72,
+      "Ios18.squeeze" : 27
+    },
+    "isUpdatable" : "0",
+    "stateSchema" : [
+      {
+        "dataType" : "Float16",
+        "isOptional" : "0",
+        "formattedType" : "State (Float16 56 × 2 × 2048 × 128)",
+        "shortDescription" : "",
+        "shape" : "[56, 2, 2048, 128]",
+        "name" : "model_model_kv_cache_0",
+        "type" : "State"
+      }
+    ],
+    "availability" : {
+      "macOS" : "15.0",
+      "tvOS" : "18.0",
+      "visionOS" : "2.0",
+      "watchOS" : "11.0",
+      "iOS" : "18.0",
+      "macCatalyst" : "18.0"
+    },
+    "computePrecision" : "Mixed (Float16, Int16, Int32, UInt16)",
+    "modelType" : {
+      "name" : "MLModelType_mlProgram"
+    },
+    "inputSchema" : [
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 1536)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 1536]",
+        "name" : "hidden_states",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Int32",
+        "formattedType" : "MultiArray (Int32 1)",
+        "shortDescription" : "",
+        "shape" : "[1]",
+        "name" : "position_ids",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 2048)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 1, 2048]",
+        "name" : "causal_mask",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Int32",
+        "formattedType" : "MultiArray (Int32 1)",
+        "shortDescription" : "",
+        "shape" : "[1]",
+        "name" : "current_pos",
+        "type" : "MultiArray"
+      }
+    ],
+    "defaultFunctionName" : "infer",
+    "generatedClassName" : "qwen25_FFN_PF_lut6_chunk_03of04",
+    "userDefinedMetadata" : {
+
+    },
+    "method" : "predict"
+  }
+]

qwen25_FFN_PF_lut6_chunk_03of04.mlmodelc/weights/weight.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a8f76651ba7e9a1fc7a573f9b1563d1487ceaefdb9910abb19338d50cdaa27d7
+size 324686720

qwen25_FFN_PF_lut6_chunk_04of04.mlmodelc/analytics/coremldata.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:104fa0152c2e4c23fafa8288d7fe075d8dc4913b12ce06c2a5a15e9d0bb6d006
+size 243

qwen25_FFN_PF_lut6_chunk_04of04.mlmodelc/coremldata.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:825ce4430834eab994271dcd253984969881cc4f6123795ede120cd626019557
+size 542

qwen25_FFN_PF_lut6_chunk_04of04.mlmodelc/metadata.json

@@ -0,0 +1,310 @@
+[
+  {
+    "metadataOutputVersion" : "3.0",
+    "storagePrecision" : "Mixed (Float16, Palettized (12 bits), Palettized (15 bits), Palettized (18 bits), UInt6)",
+    "outputSchema" : [
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 1536)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 1536]",
+        "name" : "output_hidden_states",
+        "type" : "MultiArray"
+      }
+    ],
+    "modelParameters" : [
+
+    ],
+    "specificationVersion" : 9,
+    "functions" : [
+      {
+        "inputSchema" : [
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 1536)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 1536]",
+            "name" : "hidden_states",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Int32",
+            "formattedType" : "MultiArray (Int32 1)",
+            "shortDescription" : "",
+            "shape" : "[1]",
+            "name" : "position_ids",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 2048)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 1, 2048]",
+            "name" : "causal_mask",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Int32",
+            "formattedType" : "MultiArray (Int32 1)",
+            "shortDescription" : "",
+            "shape" : "[1]",
+            "name" : "current_pos",
+            "type" : "MultiArray"
+          }
+        ],
+        "computePrecision" : "Mixed (Float16, Int16, Int32, UInt16)",
+        "storagePrecision" : "Mixed (Float16, Palettized (12 bits), Palettized (15 bits), Palettized (18 bits), UInt6)",
+        "stateSchema" : [
+          {
+            "dataType" : "Float16",
+            "isOptional" : "0",
+            "formattedType" : "State (Float16 56 × 2 × 2048 × 128)",
+            "shortDescription" : "",
+            "shape" : "[56, 2, 2048, 128]",
+            "name" : "model_model_kv_cache_0",
+            "type" : "State"
+          }
+        ],
+        "outputSchema" : [
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 1536)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 1536]",
+            "name" : "output_hidden_states",
+            "type" : "MultiArray"
+          }
+        ],
+        "name" : "infer",
+        "mlProgramOperationTypeHistogram" : {
+          "Ios18.expandDims" : 36,
+          "Ios18.mul" : 110,
+          "Ios18.softmax" : 9,
+          "Ios18.matmul" : 18,
+          "Identity" : 1,
+          "Ios18.greaterEqual" : 2,
+          "Select" : 2,
+          "Ios18.readState" : 19,
+          "Tile" : 18,
+          "Ios18.gather" : 2,
+          "Ios18.add" : 48,
+          "Ios18.layerNorm" : 19,
+          "Ios18.sliceUpdate" : 18,
+          "Ios18.writeState" : 18,
+          "Ios18.reshape" : 56,
+          "Ios18.constexprLutToDense" : 63,
+          "Ios18.conv" : 63,
+          "Ios18.concat" : 73,
+          "Ios18.transpose" : 54,
+          "Ios18.cast" : 5,
+          "Ios18.silu" : 9,
+          "Ios18.sliceByIndex" : 73,
+          "Ios18.squeeze" : 27
+        }
+      },
+      {
+        "inputSchema" : [
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 64 × 1536)",
+            "shortDescription" : "",
+            "shape" : "[1, 64, 1536]",
+            "name" : "hidden_states",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Int32",
+            "formattedType" : "MultiArray (Int32 64)",
+            "shortDescription" : "",
+            "shape" : "[64]",
+            "name" : "position_ids",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 64 × 2048)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 64, 2048]",
+            "name" : "causal_mask",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Int32",
+            "formattedType" : "MultiArray (Int32 1)",
+            "shortDescription" : "",
+            "shape" : "[1]",
+            "name" : "current_pos",
+            "type" : "MultiArray"
+          }
+        ],
+        "computePrecision" : "Mixed (Float16, Int16, Int32, UInt16)",
+        "storagePrecision" : "Mixed (Float16, Palettized (12 bits), Palettized (15 bits), Palettized (18 bits), UInt6)",
+        "stateSchema" : [
+          {
+            "dataType" : "Float16",
+            "isOptional" : "0",
+            "formattedType" : "State (Float16 56 × 2 × 2048 × 128)",
+            "shortDescription" : "",
+            "shape" : "[56, 2, 2048, 128]",
+            "name" : "model_model_kv_cache_0",
+            "type" : "State"
+          }
+        ],
+        "outputSchema" : [
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 1536)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 1536]",
+            "name" : "output_hidden_states",
+            "type" : "MultiArray"
+          }
+        ],
+        "name" : "prefill",
+        "mlProgramOperationTypeHistogram" : {
+          "Ios18.expandDims" : 36,
+          "Ios18.mul" : 108,
+          "Ios18.softmax" : 9,
+          "Ios18.matmul" : 18,
+          "Ios18.greaterEqual" : 2,
+          "Select" : 2,
+          "Ios18.readState" : 19,
+          "Tile" : 18,
+          "Ios18.gather" : 2,
+          "Ios18.add" : 48,
+          "Ios18.layerNorm" : 18,
+          "Ios18.sliceUpdate" : 18,
+          "Ios18.writeState" : 18,
+          "Ios18.reshape" : 74,
+          "Ios18.constexprLutToDense" : 63,
+          "Ios18.conv" : 63,
+          "Ios18.concat" : 72,
+          "Ios18.transpose" : 83,
+          "Ios18.cast" : 5,
+          "Ios18.silu" : 9,
+          "Ios18.sliceByIndex" : 73,
+          "Ios18.squeeze" : 27
+        }
+      }
+    ],
+    "mlProgramOperationTypeHistogram" : {
+      "Ios18.expandDims" : 36,
+      "Ios18.mul" : 110,
+      "Ios18.softmax" : 9,
+      "Ios18.matmul" : 18,
+      "Identity" : 1,
+      "Ios18.greaterEqual" : 2,
+      "Select" : 2,
+      "Ios18.readState" : 19,
+      "Tile" : 18,
+      "Ios18.gather" : 2,
+      "Ios18.add" : 48,
+      "Ios18.layerNorm" : 19,
+      "Ios18.sliceUpdate" : 18,
+      "Ios18.writeState" : 18,
+      "Ios18.reshape" : 56,
+      "Ios18.constexprLutToDense" : 63,
+      "Ios18.conv" : 63,
+      "Ios18.concat" : 73,
+      "Ios18.transpose" : 54,
+      "Ios18.cast" : 5,
+      "Ios18.silu" : 9,
+      "Ios18.sliceByIndex" : 73,
+      "Ios18.squeeze" : 27
+    },
+    "isUpdatable" : "0",
+    "stateSchema" : [
+      {
+        "dataType" : "Float16",
+        "isOptional" : "0",
+        "formattedType" : "State (Float16 56 × 2 × 2048 × 128)",
+        "shortDescription" : "",
+        "shape" : "[56, 2, 2048, 128]",
+        "name" : "model_model_kv_cache_0",
+        "type" : "State"
+      }
+    ],
+    "availability" : {
+      "macOS" : "15.0",
+      "tvOS" : "18.0",
+      "visionOS" : "2.0",
+      "watchOS" : "11.0",
+      "iOS" : "18.0",
+      "macCatalyst" : "18.0"
+    },
+    "computePrecision" : "Mixed (Float16, Int16, Int32, UInt16)",
+    "modelType" : {
+      "name" : "MLModelType_mlProgram"
+    },
+    "inputSchema" : [
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 1536)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 1536]",
+        "name" : "hidden_states",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Int32",
+        "formattedType" : "MultiArray (Int32 1)",
+        "shortDescription" : "",
+        "shape" : "[1]",
+        "name" : "position_ids",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 2048)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 1, 2048]",
+        "name" : "causal_mask",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Int32",
+        "formattedType" : "MultiArray (Int32 1)",
+        "shortDescription" : "",
+        "shape" : "[1]",
+        "name" : "current_pos",
+        "type" : "MultiArray"
+      }
+    ],
+    "defaultFunctionName" : "infer",
+    "generatedClassName" : "qwen25_FFN_PF_lut6_chunk_04of04",
+    "userDefinedMetadata" : {
+
+    },
+    "method" : "predict"
+  }
+]

qwen25_FFN_PF_lut6_chunk_04of04.mlmodelc/weights/weight.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1e40dc68aefefa40d76f49d6768845a1e6150e4e6842c503242f89f29402ea66
+size 324689856

qwen25_embeddings.mlmodelc/analytics/coremldata.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2adb2e815f1acd83c04da15fd47dccb1abd586430f59930f9f86f6ba494e5c41
+size 243

qwen25_embeddings.mlmodelc/coremldata.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5d123b29b77702fbcf07fc14b74612a9e77ed3c953ce2549f6b898907c84d248
+size 560

qwen25_embeddings.mlmodelc/metadata.json

@@ -0,0 +1,71 @@
+[
+  {
+    "shortDescription" : "Anemll Model (Embeddings) converted to CoreML",
+    "metadataOutputVersion" : "3.0",
+    "outputSchema" : [
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16)",
+        "shortDescription" : "",
+        "shape" : "[]",
+        "name" : "hidden_states",
+        "type" : "MultiArray"
+      }
+    ],
+    "version" : "0.1.1",
+    "modelParameters" : [
+
+    ],
+    "author" : "Converted with Anemll v0.1.1",
+    "specificationVersion" : 9,
+    "storagePrecision" : "Float16",
+    "mlProgramOperationTypeHistogram" : {
+      "Ios18.greaterEqual" : 1,
+      "Ios18.add" : 1,
+      "Select" : 1,
+      "Ios18.gather" : 1
+    },
+    "computePrecision" : "Mixed (Float16, Int32)",
+    "stateSchema" : [
+
+    ],
+    "isUpdatable" : "0",
+    "availability" : {
+      "macOS" : "15.0",
+      "tvOS" : "18.0",
+      "visionOS" : "2.0",
+      "watchOS" : "11.0",
+      "iOS" : "18.0",
+      "macCatalyst" : "18.0"
+    },
+    "modelType" : {
+      "name" : "MLModelType_mlProgram"
+    },
+    "inputSchema" : [
+      {
+        "shortDescription" : "",
+        "dataType" : "Int32",
+        "hasShapeFlexibility" : "1",
+        "isOptional" : "0",
+        "shapeFlexibility" : "1 × 1 | 1 × 64",
+        "formattedType" : "MultiArray (Int32 1 × 1)",
+        "type" : "MultiArray",
+        "shape" : "[1, 1]",
+        "name" : "input_ids",
+        "enumeratedShapes" : "[[1, 1], [1, 64]]"
+      }
+    ],
+    "userDefinedMetadata" : {
+      "com.anemll.context_length" : "2048",
+      "com.github.apple.coremltools.conversion_date" : "2026-02-09",
+      "com.github.apple.coremltools.source_dialect" : "TorchScript",
+      "com.github.apple.coremltools.source" : "torch==2.5.0",
+      "com.github.apple.coremltools.version" : "9.0",
+      "com.anemll.info" : "Converted with Anemll v0.1.1"
+    },
+    "generatedClassName" : "qwen25_embeddings",
+    "method" : "predict"
+  }
+]

qwen25_embeddings.mlmodelc/model.mil

@@ -0,0 +1,16 @@
+program(1.3)
+[buildInfo = dict<string, string>({{"coremlc-component-MIL", "3510.2.1"}, {"coremlc-version", "3500.32.1"}, {"coremltools-component-torch", "2.5.0"}, {"coremltools-source-dialect", "TorchScript"}, {"coremltools-version", "9.0"}})]
+{
+    func main<ios18>(tensor<int32, [1, ?]> input_ids) [FlexibleShapeInformation = tuple<tuple<string, dict<string, tensor<int32, [?]>>>, tuple<string, dict<string, dict<string, tensor<int32, [?]>>>>>((("DefaultShapes", {{"input_ids", [1, 1]}}), ("EnumeratedShapes", {{"79ae981e", {{"input_ids", [1, 1]}}}, {"ed9b58c8", {{"input_ids", [1, 64]}}}})))] {
+            int32 hidden_states_batch_dims_0 = const()[name = string("hidden_states_batch_dims_0"), val = int32(0)];
+            bool hidden_states_validate_indices_0 = const()[name = string("hidden_states_validate_indices_0"), val = bool(false)];
+            tensor<fp16, [151936, 1536]> embed_tokens_weight_to_fp16 = const()[name = string("embed_tokens_weight_to_fp16"), val = tensor<fp16, [151936, 1536]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(64)))];
+            int32 greater_equal_0_y_0 = const()[name = string("greater_equal_0_y_0"), val = int32(0)];
+            tensor<bool, [1, ?]> greater_equal_0 = greater_equal(x = input_ids, y = greater_equal_0_y_0)[name = string("greater_equal_0")];
+            int32 slice_by_index_0 = const()[name = string("slice_by_index_0"), val = int32(151936)];
+            tensor<int32, [1, ?]> add_0 = add(x = input_ids, y = slice_by_index_0)[name = string("add_0")];
+            tensor<int32, [1, ?]> select_0 = select(a = input_ids, b = add_0, cond = greater_equal_0)[name = string("select_0")];
+            int32 hidden_states_cast_fp16_axis_0 = const()[name = string("hidden_states_cast_fp16_axis_0"), val = int32(0)];
+            tensor<fp16, [1, ?, 1536]> hidden_states = gather(axis = hidden_states_cast_fp16_axis_0, batch_dims = hidden_states_batch_dims_0, indices = select_0, validate_indices = hidden_states_validate_indices_0, x = embed_tokens_weight_to_fp16)[name = string("hidden_states_cast_fp16")];
+        } -> (hidden_states);
+}

qwen25_embeddings.mlmodelc/weights/weight.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:00bdf93723628bffe53590ce127d0d86d860126ac21b6f0fdd6125972e08c1b6
+size 466747520

qwen25_lm_head_lut6.mlmodelc/analytics/coremldata.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6ef4207ad9c8ea6bcaccc71df07ccde81c1655d65be962d8b4eba7aba13b82af
+size 243

qwen25_lm_head_lut6.mlmodelc/coremldata.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a93c407091446dd6ddeefe23d2be735003856f183f30a58a0bbd4b35872b7ff1
+size 763

qwen25_lm_head_lut6.mlmodelc/metadata.json

@@ -0,0 +1,215 @@
+[
+  {
+    "metadataOutputVersion" : "3.0",
+    "storagePrecision" : "Mixed (Float16, Palettized (18 bits), UInt6)",
+    "outputSchema" : [
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 9496)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 9496]",
+        "name" : "logits1",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 9496)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 9496]",
+        "name" : "logits2",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 9496)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 9496]",
+        "name" : "logits3",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 9496)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 9496]",
+        "name" : "logits4",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 9496)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 9496]",
+        "name" : "logits5",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 9496)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 9496]",
+        "name" : "logits6",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 9496)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 9496]",
+        "name" : "logits7",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 9496)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 9496]",
+        "name" : "logits8",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 9496)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 9496]",
+        "name" : "logits9",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 9496)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 9496]",
+        "name" : "logits10",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 9496)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 9496]",
+        "name" : "logits11",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 9496)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 9496]",
+        "name" : "logits12",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 9496)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 9496]",
+        "name" : "logits13",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 9496)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 9496]",
+        "name" : "logits14",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 9496)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 9496]",
+        "name" : "logits15",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 9496)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 9496]",
+        "name" : "logits16",
+        "type" : "MultiArray"
+      }
+    ],
+    "modelParameters" : [
+
+    ],
+    "specificationVersion" : 9,
+    "mlProgramOperationTypeHistogram" : {
+      "Ios18.transpose" : 17,
+      "Ios18.constexprLutToDense" : 16,
+      "Ios18.expandDims" : 1,
+      "Ios18.conv" : 16,
+      "Ios18.squeeze" : 16
+    },
+    "computePrecision" : "Mixed (Float16, Int32)",
+    "isUpdatable" : "0",
+    "stateSchema" : [
+
+    ],
+    "availability" : {
+      "macOS" : "15.0",
+      "tvOS" : "18.0",
+      "visionOS" : "2.0",
+      "watchOS" : "11.0",
+      "iOS" : "18.0",
+      "macCatalyst" : "18.0"
+    },
+    "modelType" : {
+      "name" : "MLModelType_mlProgram"
+    },
+    "userDefinedMetadata" : {
+      "com.github.apple.coremltools.conversion_date" : "2026-02-09",
+      "com.github.apple.coremltools.source" : "torch==2.5.0",
+      "com.github.apple.coremltools.version" : "9.0",
+      "com.github.apple.coremltools.source_dialect" : "TorchScript"
+    },
+    "inputSchema" : [
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 1536)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 1536]",
+        "name" : "hidden_states",
+        "type" : "MultiArray"
+      }
+    ],
+    "generatedClassName" : "qwen25_lm_head_lut6",
+    "method" : "predict"
+  }
+]

qwen25_lm_head_lut6.mlmodelc/model.mil

@@ -0,0 +1,186 @@
+program(1.3)
+[buildInfo = dict<string, string>({{"coremlc-component-MIL", "3510.2.1"}, {"coremlc-version", "3500.32.1"}})]
+{
+    func main<ios18>(tensor<fp16, [1, 1, 1536]> hidden_states) {
+            tensor<int32, [3]> var_5 = const()[name = string("op_5"), val = tensor<int32, [3]>([0, 2, 1])];
+            tensor<int32, [1]> input_axes_0 = const()[name = string("input_axes_0"), val = tensor<int32, [1]>([2])];
+            tensor<fp16, [1, 1536, 1]> var_6_cast_fp16 = transpose(perm = var_5, x = hidden_states)[name = string("transpose_16")];
+            tensor<fp16, [1, 1536, 1, 1]> input_cast_fp16 = expand_dims(axes = input_axes_0, x = var_6_cast_fp16)[name = string("input_cast_fp16")];
+            string var_29_pad_type_0 = const()[name = string("op_29_pad_type_0"), val = string("valid")];
+            tensor<int32, [2]> var_29_strides_0 = const()[name = string("op_29_strides_0"), val = tensor<int32, [2]>([1, 1])];
+            tensor<int32, [4]> var_29_pad_0 = const()[name = string("op_29_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
+            tensor<int32, [2]> var_29_dilations_0 = const()[name = string("op_29_dilations_0"), val = tensor<int32, [2]>([1, 1])];
+            int32 var_29_groups_0 = const()[name = string("op_29_groups_0"), val = int32(1)];
+            tensor<fp16, [9496, 1536, 1, 1]> op_9_promoted_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [9496, 1536, 1, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(64))), lut = tensor<fp16, [2374, 1, 1, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(10939520))))[name = string("op_9_promoted_to_fp16_palettized")];
+            tensor<fp16, [1, 9496, 1, 1]> var_29_cast_fp16 = conv(dilations = var_29_dilations_0, groups = var_29_groups_0, pad = var_29_pad_0, pad_type = var_29_pad_type_0, strides = var_29_strides_0, weight = op_9_promoted_to_fp16_palettized, x = input_cast_fp16)[name = string("op_29_cast_fp16")];
+            tensor<int32, [1]> var_31_axes_0 = const()[name = string("op_31_axes_0"), val = tensor<int32, [1]>([2])];
+            tensor<fp16, [1, 9496, 1]> var_31_cast_fp16 = squeeze(axes = var_31_axes_0, x = var_29_cast_fp16)[name = string("op_31_cast_fp16")];
+            tensor<int32, [3]> var_34_perm_0 = const()[name = string("op_34_perm_0"), val = tensor<int32, [3]>([0, 2, 1])];
+            string var_55_pad_type_0 = const()[name = string("op_55_pad_type_0"), val = string("valid")];
+            tensor<int32, [2]> var_55_strides_0 = const()[name = string("op_55_strides_0"), val = tensor<int32, [2]>([1, 1])];
+            tensor<int32, [4]> var_55_pad_0 = const()[name = string("op_55_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
+            tensor<int32, [2]> var_55_dilations_0 = const()[name = string("op_55_dilations_0"), val = tensor<int32, [2]>([1, 1])];
+            int32 var_55_groups_0 = const()[name = string("op_55_groups_0"), val = int32(1)];
+            tensor<fp16, [9496, 1536, 1, 1]> op_35_promoted_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [9496, 1536, 1, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(11243456))), lut = tensor<fp16, [2374, 1, 1, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(22182912))))[name = string("op_35_promoted_to_fp16_palettized")];
+            tensor<fp16, [1, 9496, 1, 1]> var_55_cast_fp16 = conv(dilations = var_55_dilations_0, groups = var_55_groups_0, pad = var_55_pad_0, pad_type = var_55_pad_type_0, strides = var_55_strides_0, weight = op_35_promoted_to_fp16_palettized, x = input_cast_fp16)[name = string("op_55_cast_fp16")];
+            tensor<int32, [1]> var_57_axes_0 = const()[name = string("op_57_axes_0"), val = tensor<int32, [1]>([2])];
+            tensor<fp16, [1, 9496, 1]> var_57_cast_fp16 = squeeze(axes = var_57_axes_0, x = var_55_cast_fp16)[name = string("op_57_cast_fp16")];
+            tensor<int32, [3]> var_60_perm_0 = const()[name = string("op_60_perm_0"), val = tensor<int32, [3]>([0, 2, 1])];
+            string var_81_pad_type_0 = const()[name = string("op_81_pad_type_0"), val = string("valid")];
+            tensor<int32, [2]> var_81_strides_0 = const()[name = string("op_81_strides_0"), val = tensor<int32, [2]>([1, 1])];
+            tensor<int32, [4]> var_81_pad_0 = const()[name = string("op_81_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
+            tensor<int32, [2]> var_81_dilations_0 = const()[name = string("op_81_dilations_0"), val = tensor<int32, [2]>([1, 1])];
+            int32 var_81_groups_0 = const()[name = string("op_81_groups_0"), val = int32(1)];
+            tensor<fp16, [9496, 1536, 1, 1]> op_61_promoted_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [9496, 1536, 1, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(22486848))), lut = tensor<fp16, [2374, 1, 1, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(33426304))))[name = string("op_61_promoted_to_fp16_palettized")];
+            tensor<fp16, [1, 9496, 1, 1]> var_81_cast_fp16 = conv(dilations = var_81_dilations_0, groups = var_81_groups_0, pad = var_81_pad_0, pad_type = var_81_pad_type_0, strides = var_81_strides_0, weight = op_61_promoted_to_fp16_palettized, x = input_cast_fp16)[name = string("op_81_cast_fp16")];
+            tensor<int32, [1]> var_83_axes_0 = const()[name = string("op_83_axes_0"), val = tensor<int32, [1]>([2])];
+            tensor<fp16, [1, 9496, 1]> var_83_cast_fp16 = squeeze(axes = var_83_axes_0, x = var_81_cast_fp16)[name = string("op_83_cast_fp16")];
+            tensor<int32, [3]> var_86_perm_0 = const()[name = string("op_86_perm_0"), val = tensor<int32, [3]>([0, 2, 1])];
+            string var_107_pad_type_0 = const()[name = string("op_107_pad_type_0"), val = string("valid")];
+            tensor<int32, [2]> var_107_strides_0 = const()[name = string("op_107_strides_0"), val = tensor<int32, [2]>([1, 1])];
+            tensor<int32, [4]> var_107_pad_0 = const()[name = string("op_107_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
+            tensor<int32, [2]> var_107_dilations_0 = const()[name = string("op_107_dilations_0"), val = tensor<int32, [2]>([1, 1])];
+            int32 var_107_groups_0 = const()[name = string("op_107_groups_0"), val = int32(1)];
+            tensor<fp16, [9496, 1536, 1, 1]> op_87_promoted_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [9496, 1536, 1, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(33730240))), lut = tensor<fp16, [2374, 1, 1, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(44669696))))[name = string("op_87_promoted_to_fp16_palettized")];
+            tensor<fp16, [1, 9496, 1, 1]> var_107_cast_fp16 = conv(dilations = var_107_dilations_0, groups = var_107_groups_0, pad = var_107_pad_0, pad_type = var_107_pad_type_0, strides = var_107_strides_0, weight = op_87_promoted_to_fp16_palettized, x = input_cast_fp16)[name = string("op_107_cast_fp16")];
+            tensor<int32, [1]> var_109_axes_0 = const()[name = string("op_109_axes_0"), val = tensor<int32, [1]>([2])];
+            tensor<fp16, [1, 9496, 1]> var_109_cast_fp16 = squeeze(axes = var_109_axes_0, x = var_107_cast_fp16)[name = string("op_109_cast_fp16")];
+            tensor<int32, [3]> var_112_perm_0 = const()[name = string("op_112_perm_0"), val = tensor<int32, [3]>([0, 2, 1])];
+            string var_133_pad_type_0 = const()[name = string("op_133_pad_type_0"), val = string("valid")];
+            tensor<int32, [2]> var_133_strides_0 = const()[name = string("op_133_strides_0"), val = tensor<int32, [2]>([1, 1])];
+            tensor<int32, [4]> var_133_pad_0 = const()[name = string("op_133_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
+            tensor<int32, [2]> var_133_dilations_0 = const()[name = string("op_133_dilations_0"), val = tensor<int32, [2]>([1, 1])];
+            int32 var_133_groups_0 = const()[name = string("op_133_groups_0"), val = int32(1)];
+            tensor<fp16, [9496, 1536, 1, 1]> op_113_promoted_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [9496, 1536, 1, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(44973632))), lut = tensor<fp16, [2374, 1, 1, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(55913088))))[name = string("op_113_promoted_to_fp16_palettized")];
+            tensor<fp16, [1, 9496, 1, 1]> var_133_cast_fp16 = conv(dilations = var_133_dilations_0, groups = var_133_groups_0, pad = var_133_pad_0, pad_type = var_133_pad_type_0, strides = var_133_strides_0, weight = op_113_promoted_to_fp16_palettized, x = input_cast_fp16)[name = string("op_133_cast_fp16")];
+            tensor<int32, [1]> var_135_axes_0 = const()[name = string("op_135_axes_0"), val = tensor<int32, [1]>([2])];
+            tensor<fp16, [1, 9496, 1]> var_135_cast_fp16 = squeeze(axes = var_135_axes_0, x = var_133_cast_fp16)[name = string("op_135_cast_fp16")];
+            tensor<int32, [3]> var_138_perm_0 = const()[name = string("op_138_perm_0"), val = tensor<int32, [3]>([0, 2, 1])];
+            string var_159_pad_type_0 = const()[name = string("op_159_pad_type_0"), val = string("valid")];
+            tensor<int32, [2]> var_159_strides_0 = const()[name = string("op_159_strides_0"), val = tensor<int32, [2]>([1, 1])];
+            tensor<int32, [4]> var_159_pad_0 = const()[name = string("op_159_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
+            tensor<int32, [2]> var_159_dilations_0 = const()[name = string("op_159_dilations_0"), val = tensor<int32, [2]>([1, 1])];
+            int32 var_159_groups_0 = const()[name = string("op_159_groups_0"), val = int32(1)];
+            tensor<fp16, [9496, 1536, 1, 1]> op_139_promoted_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [9496, 1536, 1, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(56217024))), lut = tensor<fp16, [2374, 1, 1, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(67156480))))[name = string("op_139_promoted_to_fp16_palettized")];
+            tensor<fp16, [1, 9496, 1, 1]> var_159_cast_fp16 = conv(dilations = var_159_dilations_0, groups = var_159_groups_0, pad = var_159_pad_0, pad_type = var_159_pad_type_0, strides = var_159_strides_0, weight = op_139_promoted_to_fp16_palettized, x = input_cast_fp16)[name = string("op_159_cast_fp16")];
+            tensor<int32, [1]> var_161_axes_0 = const()[name = string("op_161_axes_0"), val = tensor<int32, [1]>([2])];
+            tensor<fp16, [1, 9496, 1]> var_161_cast_fp16 = squeeze(axes = var_161_axes_0, x = var_159_cast_fp16)[name = string("op_161_cast_fp16")];
+            tensor<int32, [3]> var_164_perm_0 = const()[name = string("op_164_perm_0"), val = tensor<int32, [3]>([0, 2, 1])];
+            string var_185_pad_type_0 = const()[name = string("op_185_pad_type_0"), val = string("valid")];
+            tensor<int32, [2]> var_185_strides_0 = const()[name = string("op_185_strides_0"), val = tensor<int32, [2]>([1, 1])];
+            tensor<int32, [4]> var_185_pad_0 = const()[name = string("op_185_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
+            tensor<int32, [2]> var_185_dilations_0 = const()[name = string("op_185_dilations_0"), val = tensor<int32, [2]>([1, 1])];
+            int32 var_185_groups_0 = const()[name = string("op_185_groups_0"), val = int32(1)];
+            tensor<fp16, [9496, 1536, 1, 1]> op_165_promoted_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [9496, 1536, 1, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(67460416))), lut = tensor<fp16, [2374, 1, 1, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(78399872))))[name = string("op_165_promoted_to_fp16_palettized")];
+            tensor<fp16, [1, 9496, 1, 1]> var_185_cast_fp16 = conv(dilations = var_185_dilations_0, groups = var_185_groups_0, pad = var_185_pad_0, pad_type = var_185_pad_type_0, strides = var_185_strides_0, weight = op_165_promoted_to_fp16_palettized, x = input_cast_fp16)[name = string("op_185_cast_fp16")];
+            tensor<int32, [1]> var_187_axes_0 = const()[name = string("op_187_axes_0"), val = tensor<int32, [1]>([2])];
+            tensor<fp16, [1, 9496, 1]> var_187_cast_fp16 = squeeze(axes = var_187_axes_0, x = var_185_cast_fp16)[name = string("op_187_cast_fp16")];
+            tensor<int32, [3]> var_190_perm_0 = const()[name = string("op_190_perm_0"), val = tensor<int32, [3]>([0, 2, 1])];
+            string var_211_pad_type_0 = const()[name = string("op_211_pad_type_0"), val = string("valid")];
+            tensor<int32, [2]> var_211_strides_0 = const()[name = string("op_211_strides_0"), val = tensor<int32, [2]>([1, 1])];
+            tensor<int32, [4]> var_211_pad_0 = const()[name = string("op_211_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
+            tensor<int32, [2]> var_211_dilations_0 = const()[name = string("op_211_dilations_0"), val = tensor<int32, [2]>([1, 1])];
+            int32 var_211_groups_0 = const()[name = string("op_211_groups_0"), val = int32(1)];
+            tensor<fp16, [9496, 1536, 1, 1]> op_191_promoted_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [9496, 1536, 1, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(78703808))), lut = tensor<fp16, [2374, 1, 1, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(89643264))))[name = string("op_191_promoted_to_fp16_palettized")];
+            tensor<fp16, [1, 9496, 1, 1]> var_211_cast_fp16 = conv(dilations = var_211_dilations_0, groups = var_211_groups_0, pad = var_211_pad_0, pad_type = var_211_pad_type_0, strides = var_211_strides_0, weight = op_191_promoted_to_fp16_palettized, x = input_cast_fp16)[name = string("op_211_cast_fp16")];
+            tensor<int32, [1]> var_213_axes_0 = const()[name = string("op_213_axes_0"), val = tensor<int32, [1]>([2])];
+            tensor<fp16, [1, 9496, 1]> var_213_cast_fp16 = squeeze(axes = var_213_axes_0, x = var_211_cast_fp16)[name = string("op_213_cast_fp16")];
+            tensor<int32, [3]> var_216_perm_0 = const()[name = string("op_216_perm_0"), val = tensor<int32, [3]>([0, 2, 1])];
+            string var_237_pad_type_0 = const()[name = string("op_237_pad_type_0"), val = string("valid")];
+            tensor<int32, [2]> var_237_strides_0 = const()[name = string("op_237_strides_0"), val = tensor<int32, [2]>([1, 1])];
+            tensor<int32, [4]> var_237_pad_0 = const()[name = string("op_237_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
+            tensor<int32, [2]> var_237_dilations_0 = const()[name = string("op_237_dilations_0"), val = tensor<int32, [2]>([1, 1])];
+            int32 var_237_groups_0 = const()[name = string("op_237_groups_0"), val = int32(1)];
+            tensor<fp16, [9496, 1536, 1, 1]> op_217_promoted_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [9496, 1536, 1, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(89947200))), lut = tensor<fp16, [2374, 1, 1, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(100886656))))[name = string("op_217_promoted_to_fp16_palettized")];
+            tensor<fp16, [1, 9496, 1, 1]> var_237_cast_fp16 = conv(dilations = var_237_dilations_0, groups = var_237_groups_0, pad = var_237_pad_0, pad_type = var_237_pad_type_0, strides = var_237_strides_0, weight = op_217_promoted_to_fp16_palettized, x = input_cast_fp16)[name = string("op_237_cast_fp16")];
+            tensor<int32, [1]> var_239_axes_0 = const()[name = string("op_239_axes_0"), val = tensor<int32, [1]>([2])];
+            tensor<fp16, [1, 9496, 1]> var_239_cast_fp16 = squeeze(axes = var_239_axes_0, x = var_237_cast_fp16)[name = string("op_239_cast_fp16")];
+            tensor<int32, [3]> var_242_perm_0 = const()[name = string("op_242_perm_0"), val = tensor<int32, [3]>([0, 2, 1])];
+            string var_263_pad_type_0 = const()[name = string("op_263_pad_type_0"), val = string("valid")];
+            tensor<int32, [2]> var_263_strides_0 = const()[name = string("op_263_strides_0"), val = tensor<int32, [2]>([1, 1])];
+            tensor<int32, [4]> var_263_pad_0 = const()[name = string("op_263_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
+            tensor<int32, [2]> var_263_dilations_0 = const()[name = string("op_263_dilations_0"), val = tensor<int32, [2]>([1, 1])];
+            int32 var_263_groups_0 = const()[name = string("op_263_groups_0"), val = int32(1)];
+            tensor<fp16, [9496, 1536, 1, 1]> op_243_promoted_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [9496, 1536, 1, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(101190592))), lut = tensor<fp16, [2374, 1, 1, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(112130048))))[name = string("op_243_promoted_to_fp16_palettized")];
+            tensor<fp16, [1, 9496, 1, 1]> var_263_cast_fp16 = conv(dilations = var_263_dilations_0, groups = var_263_groups_0, pad = var_263_pad_0, pad_type = var_263_pad_type_0, strides = var_263_strides_0, weight = op_243_promoted_to_fp16_palettized, x = input_cast_fp16)[name = string("op_263_cast_fp16")];
+            tensor<int32, [1]> var_265_axes_0 = const()[name = string("op_265_axes_0"), val = tensor<int32, [1]>([2])];
+            tensor<fp16, [1, 9496, 1]> var_265_cast_fp16 = squeeze(axes = var_265_axes_0, x = var_263_cast_fp16)[name = string("op_265_cast_fp16")];
+            tensor<int32, [3]> var_268_perm_0 = const()[name = string("op_268_perm_0"), val = tensor<int32, [3]>([0, 2, 1])];
+            string var_289_pad_type_0 = const()[name = string("op_289_pad_type_0"), val = string("valid")];
+            tensor<int32, [2]> var_289_strides_0 = const()[name = string("op_289_strides_0"), val = tensor<int32, [2]>([1, 1])];
+            tensor<int32, [4]> var_289_pad_0 = const()[name = string("op_289_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
+            tensor<int32, [2]> var_289_dilations_0 = const()[name = string("op_289_dilations_0"), val = tensor<int32, [2]>([1, 1])];
+            int32 var_289_groups_0 = const()[name = string("op_289_groups_0"), val = int32(1)];
+            tensor<fp16, [9496, 1536, 1, 1]> op_269_promoted_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [9496, 1536, 1, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(112433984))), lut = tensor<fp16, [2374, 1, 1, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(123373440))))[name = string("op_269_promoted_to_fp16_palettized")];
+            tensor<fp16, [1, 9496, 1, 1]> var_289_cast_fp16 = conv(dilations = var_289_dilations_0, groups = var_289_groups_0, pad = var_289_pad_0, pad_type = var_289_pad_type_0, strides = var_289_strides_0, weight = op_269_promoted_to_fp16_palettized, x = input_cast_fp16)[name = string("op_289_cast_fp16")];
+            tensor<int32, [1]> var_291_axes_0 = const()[name = string("op_291_axes_0"), val = tensor<int32, [1]>([2])];
+            tensor<fp16, [1, 9496, 1]> var_291_cast_fp16 = squeeze(axes = var_291_axes_0, x = var_289_cast_fp16)[name = string("op_291_cast_fp16")];
+            tensor<int32, [3]> var_294_perm_0 = const()[name = string("op_294_perm_0"), val = tensor<int32, [3]>([0, 2, 1])];
+            string var_315_pad_type_0 = const()[name = string("op_315_pad_type_0"), val = string("valid")];
+            tensor<int32, [2]> var_315_strides_0 = const()[name = string("op_315_strides_0"), val = tensor<int32, [2]>([1, 1])];
+            tensor<int32, [4]> var_315_pad_0 = const()[name = string("op_315_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
+            tensor<int32, [2]> var_315_dilations_0 = const()[name = string("op_315_dilations_0"), val = tensor<int32, [2]>([1, 1])];
+            int32 var_315_groups_0 = const()[name = string("op_315_groups_0"), val = int32(1)];
+            tensor<fp16, [9496, 1536, 1, 1]> op_295_promoted_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [9496, 1536, 1, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(123677376))), lut = tensor<fp16, [2374, 1, 1, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(134616832))))[name = string("op_295_promoted_to_fp16_palettized")];
+            tensor<fp16, [1, 9496, 1, 1]> var_315_cast_fp16 = conv(dilations = var_315_dilations_0, groups = var_315_groups_0, pad = var_315_pad_0, pad_type = var_315_pad_type_0, strides = var_315_strides_0, weight = op_295_promoted_to_fp16_palettized, x = input_cast_fp16)[name = string("op_315_cast_fp16")];
+            tensor<int32, [1]> var_317_axes_0 = const()[name = string("op_317_axes_0"), val = tensor<int32, [1]>([2])];
+            tensor<fp16, [1, 9496, 1]> var_317_cast_fp16 = squeeze(axes = var_317_axes_0, x = var_315_cast_fp16)[name = string("op_317_cast_fp16")];
+            tensor<int32, [3]> var_320_perm_0 = const()[name = string("op_320_perm_0"), val = tensor<int32, [3]>([0, 2, 1])];
+            string var_341_pad_type_0 = const()[name = string("op_341_pad_type_0"), val = string("valid")];
+            tensor<int32, [2]> var_341_strides_0 = const()[name = string("op_341_strides_0"), val = tensor<int32, [2]>([1, 1])];
+            tensor<int32, [4]> var_341_pad_0 = const()[name = string("op_341_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
+            tensor<int32, [2]> var_341_dilations_0 = const()[name = string("op_341_dilations_0"), val = tensor<int32, [2]>([1, 1])];
+            int32 var_341_groups_0 = const()[name = string("op_341_groups_0"), val = int32(1)];
+            tensor<fp16, [9496, 1536, 1, 1]> op_321_promoted_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [9496, 1536, 1, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(134920768))), lut = tensor<fp16, [2374, 1, 1, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(145860224))))[name = string("op_321_promoted_to_fp16_palettized")];
+            tensor<fp16, [1, 9496, 1, 1]> var_341_cast_fp16 = conv(dilations = var_341_dilations_0, groups = var_341_groups_0, pad = var_341_pad_0, pad_type = var_341_pad_type_0, strides = var_341_strides_0, weight = op_321_promoted_to_fp16_palettized, x = input_cast_fp16)[name = string("op_341_cast_fp16")];
+            tensor<int32, [1]> var_343_axes_0 = const()[name = string("op_343_axes_0"), val = tensor<int32, [1]>([2])];
+            tensor<fp16, [1, 9496, 1]> var_343_cast_fp16 = squeeze(axes = var_343_axes_0, x = var_341_cast_fp16)[name = string("op_343_cast_fp16")];
+            tensor<int32, [3]> var_346_perm_0 = const()[name = string("op_346_perm_0"), val = tensor<int32, [3]>([0, 2, 1])];
+            string var_367_pad_type_0 = const()[name = string("op_367_pad_type_0"), val = string("valid")];
+            tensor<int32, [2]> var_367_strides_0 = const()[name = string("op_367_strides_0"), val = tensor<int32, [2]>([1, 1])];
+            tensor<int32, [4]> var_367_pad_0 = const()[name = string("op_367_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
+            tensor<int32, [2]> var_367_dilations_0 = const()[name = string("op_367_dilations_0"), val = tensor<int32, [2]>([1, 1])];
+            int32 var_367_groups_0 = const()[name = string("op_367_groups_0"), val = int32(1)];
+            tensor<fp16, [9496, 1536, 1, 1]> op_347_promoted_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [9496, 1536, 1, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(146164160))), lut = tensor<fp16, [2374, 1, 1, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(157103616))))[name = string("op_347_promoted_to_fp16_palettized")];
+            tensor<fp16, [1, 9496, 1, 1]> var_367_cast_fp16 = conv(dilations = var_367_dilations_0, groups = var_367_groups_0, pad = var_367_pad_0, pad_type = var_367_pad_type_0, strides = var_367_strides_0, weight = op_347_promoted_to_fp16_palettized, x = input_cast_fp16)[name = string("op_367_cast_fp16")];
+            tensor<int32, [1]> var_369_axes_0 = const()[name = string("op_369_axes_0"), val = tensor<int32, [1]>([2])];
+            tensor<fp16, [1, 9496, 1]> var_369_cast_fp16 = squeeze(axes = var_369_axes_0, x = var_367_cast_fp16)[name = string("op_369_cast_fp16")];
+            tensor<int32, [3]> var_372_perm_0 = const()[name = string("op_372_perm_0"), val = tensor<int32, [3]>([0, 2, 1])];
+            string var_393_pad_type_0 = const()[name = string("op_393_pad_type_0"), val = string("valid")];
+            tensor<int32, [2]> var_393_strides_0 = const()[name = string("op_393_strides_0"), val = tensor<int32, [2]>([1, 1])];
+            tensor<int32, [4]> var_393_pad_0 = const()[name = string("op_393_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
+            tensor<int32, [2]> var_393_dilations_0 = const()[name = string("op_393_dilations_0"), val = tensor<int32, [2]>([1, 1])];
+            int32 var_393_groups_0 = const()[name = string("op_393_groups_0"), val = int32(1)];
+            tensor<fp16, [9496, 1536, 1, 1]> op_373_promoted_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [9496, 1536, 1, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(157407552))), lut = tensor<fp16, [2374, 1, 1, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(168347008))))[name = string("op_373_promoted_to_fp16_palettized")];
+            tensor<fp16, [1, 9496, 1, 1]> var_393_cast_fp16 = conv(dilations = var_393_dilations_0, groups = var_393_groups_0, pad = var_393_pad_0, pad_type = var_393_pad_type_0, strides = var_393_strides_0, weight = op_373_promoted_to_fp16_palettized, x = input_cast_fp16)[name = string("op_393_cast_fp16")];
+            tensor<int32, [1]> var_395_axes_0 = const()[name = string("op_395_axes_0"), val = tensor<int32, [1]>([2])];
+            tensor<fp16, [1, 9496, 1]> var_395_cast_fp16 = squeeze(axes = var_395_axes_0, x = var_393_cast_fp16)[name = string("op_395_cast_fp16")];
+            tensor<int32, [3]> var_398_perm_0 = const()[name = string("op_398_perm_0"), val = tensor<int32, [3]>([0, 2, 1])];
+            string var_419_pad_type_0 = const()[name = string("op_419_pad_type_0"), val = string("valid")];
+            tensor<int32, [2]> var_419_strides_0 = const()[name = string("op_419_strides_0"), val = tensor<int32, [2]>([1, 1])];
+            tensor<int32, [4]> var_419_pad_0 = const()[name = string("op_419_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
+            tensor<int32, [2]> var_419_dilations_0 = const()[name = string("op_419_dilations_0"), val = tensor<int32, [2]>([1, 1])];
+            int32 var_419_groups_0 = const()[name = string("op_419_groups_0"), val = int32(1)];
+            tensor<fp16, [9496, 1536, 1, 1]> op_399_promoted_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [9496, 1536, 1, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(168650944))), lut = tensor<fp16, [2374, 1, 1, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(179590400))))[name = string("op_399_promoted_to_fp16_palettized")];
+            tensor<fp16, [1, 9496, 1, 1]> var_419_cast_fp16 = conv(dilations = var_419_dilations_0, groups = var_419_groups_0, pad = var_419_pad_0, pad_type = var_419_pad_type_0, strides = var_419_strides_0, weight = op_399_promoted_to_fp16_palettized, x = input_cast_fp16)[name = string("op_419_cast_fp16")];
+            tensor<int32, [1]> var_421_axes_0 = const()[name = string("op_421_axes_0"), val = tensor<int32, [1]>([2])];
+            tensor<fp16, [1, 9496, 1]> var_421_cast_fp16 = squeeze(axes = var_421_axes_0, x = var_419_cast_fp16)[name = string("op_421_cast_fp16")];
+            tensor<int32, [3]> var_424_perm_0 = const()[name = string("op_424_perm_0"), val = tensor<int32, [3]>([0, 2, 1])];
+            tensor<fp16, [1, 1, 9496]> logits1 = transpose(perm = var_34_perm_0, x = var_31_cast_fp16)[name = string("transpose_0")];
+            tensor<fp16, [1, 1, 9496]> logits2 = transpose(perm = var_60_perm_0, x = var_57_cast_fp16)[name = string("transpose_1")];
+            tensor<fp16, [1, 1, 9496]> logits3 = transpose(perm = var_86_perm_0, x = var_83_cast_fp16)[name = string("transpose_2")];
+            tensor<fp16, [1, 1, 9496]> logits4 = transpose(perm = var_112_perm_0, x = var_109_cast_fp16)[name = string("transpose_3")];
+            tensor<fp16, [1, 1, 9496]> logits5 = transpose(perm = var_138_perm_0, x = var_135_cast_fp16)[name = string("transpose_4")];
+            tensor<fp16, [1, 1, 9496]> logits6 = transpose(perm = var_164_perm_0, x = var_161_cast_fp16)[name = string("transpose_5")];
+            tensor<fp16, [1, 1, 9496]> logits7 = transpose(perm = var_190_perm_0, x = var_187_cast_fp16)[name = string("transpose_6")];
+            tensor<fp16, [1, 1, 9496]> logits8 = transpose(perm = var_216_perm_0, x = var_213_cast_fp16)[name = string("transpose_7")];
+            tensor<fp16, [1, 1, 9496]> logits9 = transpose(perm = var_242_perm_0, x = var_239_cast_fp16)[name = string("transpose_8")];
+            tensor<fp16, [1, 1, 9496]> logits10 = transpose(perm = var_268_perm_0, x = var_265_cast_fp16)[name = string("transpose_9")];
+            tensor<fp16, [1, 1, 9496]> logits11 = transpose(perm = var_294_perm_0, x = var_291_cast_fp16)[name = string("transpose_10")];
+            tensor<fp16, [1, 1, 9496]> logits12 = transpose(perm = var_320_perm_0, x = var_317_cast_fp16)[name = string("transpose_11")];
+            tensor<fp16, [1, 1, 9496]> logits13 = transpose(perm = var_346_perm_0, x = var_343_cast_fp16)[name = string("transpose_12")];
+            tensor<fp16, [1, 1, 9496]> logits14 = transpose(perm = var_372_perm_0, x = var_369_cast_fp16)[name = string("transpose_13")];
+            tensor<fp16, [1, 1, 9496]> logits15 = transpose(perm = var_398_perm_0, x = var_395_cast_fp16)[name = string("transpose_14")];
+            tensor<fp16, [1, 1, 9496]> logits16 = transpose(perm = var_424_perm_0, x = var_421_cast_fp16)[name = string("transpose_15")];
+        } -> (logits1, logits2, logits3, logits4, logits5, logits6, logits7, logits8, logits9, logits10, logits11, logits12, logits13, logits14, logits15, logits16);
+}

qwen25_lm_head_lut6.mlmodelc/weights/weight.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ce6267d9cb0bfaa14fea5432d2adeb2846d3af0524f1543ed8c2e80a85e507e5
+size 179894336

tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:296e081e2f5ecf9d87814aa9b0f4b12d670ed2b2e2be6c84e01a9466c953afb7
+size 11422267

tokenizer_config.json

@@ -0,0 +1,224 @@
+{
+  "add_bos_token": false,
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "151643": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151644": {
+      "content": "<|im_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151645": {
+      "content": "<|im_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151646": {
+      "content": "<|object_ref_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151647": {
+      "content": "<|object_ref_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151648": {
+      "content": "<|box_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151649": {
+      "content": "<|box_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151650": {
+      "content": "<|quad_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151651": {
+      "content": "<|quad_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151652": {
+      "content": "<|vision_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151653": {
+      "content": "<|vision_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151654": {
+      "content": "<|vision_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151655": {
+      "content": "<|image_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151656": {
+      "content": "<|video_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151657": {
+      "content": "<tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151658": {
+      "content": "</tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151659": {
+      "content": "<|fim_prefix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151660": {
+      "content": "<|fim_middle|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151661": {
+      "content": "<|fim_suffix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151662": {
+      "content": "<|fim_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151663": {
+      "content": "<|repo_name|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151664": {
+      "content": "<|file_sep|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151665": {
+      "content": "<think>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151666": {
+      "content": "</think>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    }
+  },
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": {},
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "padding_side": "right",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}