Publish onnx-android artifact

.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
+model.ort filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,47 @@
+---
+license: apache-2.0
+pipeline_tag: token-classification
+library_name: openmed
+tags:
+- openmed
+- medical-nlp
+---
+
+# OpenMed-NER-BloodCancerDetect-TinyMed-66M-v1-onnx-android
+
+This model card is rendered from the OpenMed model manifest. Update `models.jsonl` and rerun the publish step instead of editing this file directly.
+
+## Manifest Summary
+
+| Field | Value |
+|---|---|
+| Repository | `OpenMed/OpenMed-NER-BloodCancerDetect-TinyMed-66M-v1-onnx-android` |
+| Family | NER |
+| Task | token-classification |
+| Languages | en |
+| Tier | Tiny |
+| Parameters | 66M (66,000,000) |
+| Architecture | distilbert |
+| Base model | OpenMed/OpenMed-NER-BloodCancerDetect-TinyMed-66M |
+| Formats | onnx-android, int8, ort-android |
+| License | apache-2.0 |
+| arXiv | [arXiv:2508.01630](https://arxiv.org/abs/2508.01630) |
+| Reproducibility hash | `sha256:67c83f0655ea8dcb99e98679d7b6fac978e12f13fd2768200db5aab30676d38d` |
+| Released | 2026-07-09 |
+
+## Benchmark
+
+| Dataset | Micro F1 | Recall |
+|---|---:|---:|
+| Not reported | Not reported | Not reported |
+
+## Canonical Labels
+
+`CL`
+
+## Artifact Format
+
+| Field | Value |
+|---|---|
+| Runtime artifacts | onnx-android, ort-android |
+| Quantization | int8 |

config.json

@@ -0,0 +1,44 @@
+{
+  "transformers_version": "5.12.1",
+  "architectures": [
+    "DistilBertForTokenClassification"
+  ],
+  "output_hidden_states": false,
+  "return_dict": true,
+  "dtype": "bfloat16",
+  "chunk_size_feed_forward": 0,
+  "is_encoder_decoder": false,
+  "id2label": {
+    "0": "B-CL",
+    "1": "I-CL",
+    "2": "O"
+  },
+  "label2id": {
+    "B-CL": 0,
+    "I-CL": 1,
+    "O": 2
+  },
+  "problem_type": null,
+  "vocab_size": 30522,
+  "max_position_embeddings": 512,
+  "sinusoidal_pos_embds": false,
+  "n_layers": 6,
+  "n_heads": 12,
+  "dim": 768,
+  "hidden_dim": 3072,
+  "dropout": 0.1,
+  "attention_dropout": 0.1,
+  "activation": "gelu",
+  "initializer_range": 0.02,
+  "qa_dropout": 0.1,
+  "seq_classif_dropout": 0.2,
+  "pad_token_id": 0,
+  "eos_token_id": null,
+  "bos_token_id": null,
+  "tie_word_embeddings": true,
+  "_name_or_path": "OpenMed/OpenMed-NER-BloodCancerDetect-TinyMed-66M",
+  "model_type": "distilbert",
+  "tie_weights_": true,
+  "output_attentions": false,
+  "max_sequence_length": 512
+}

id2label.json

@@ -0,0 +1,5 @@
+{
+  "0": "B-CL",
+  "1": "I-CL",
+  "2": "O"
+}

model.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e5b731f6df0e1403f66ebf8aac80cd307b14f671e36c3493c36372e82abd4ac1
+size 265598809

model.ort

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9c035e372ea2950d68bc2d5b3c90b59bfe2f43685292adf7ba99055013b2c195
+size 265648400

model.required_operators_and_types.config

@@ -0,0 +1,10 @@
+# Generated from model/s:
+# - /Users/maziyar/Developer/openmed/dist/onnx-android/OpenMed/OpenMed-NER-BloodCancerDetect-TinyMed-66M/model.ort
+ai.onnx;7;And
+ai.onnx;11;Range{"inputs": {"0": ["int64_t"]}}
+ai.onnx;13;Cast{"inputs": {"0": ["int64_t"]}, "outputs": {"0": ["bool"]}},Concat{"inputs": {"0": ["int64_t"]}},Equal{"inputs": {"0": ["int64_t"]}},Expand{"inputs": {"0": ["bool"]}},Flatten,Gather{"inputs": {"0": ["bool", "float", "int64_t"], "1": ["int64_t"]}},IsNaN{"inputs": {"0": ["float"]}},MatMul{"inputs": {"0": ["float"]}},Slice{"inputs": {"0": ["int64_t"], "1": ["int64_t"]}},Softmax{"inputs": {"0": ["float"]}},Transpose{"inputs": {"0": ["float"]}},Unsqueeze
+ai.onnx;14;Add{"inputs": {"0": ["float", "int64_t"]}},Mul{"inputs": {"0": ["int64_t"]}},Reshape
+ai.onnx;15;Shape
+ai.onnx;16;GreaterOrEqual,Where{"inputs": {"1": ["float", "int64_t"]}}
+ai.onnx;17;LayerNormalization
+com.microsoft;1;BiasGelu,FusedMatMul

model_fp16.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:61d8a4f667c7592ecfe83e362595ddbc8f3a55c944caab90b13415299f5f258f
+size 132919522

model_int8.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9980341d5023321cd2066313b35f14dfdf7cd9c818b0ace43766e22621b812ce
+size 138297206

openmed-onnx.json

@@ -0,0 +1,463 @@
+{
+  "format": "openmed-onnx",
+  "format_version": 1,
+  "formats": [
+    "onnx-android",
+    "onnx-int8",
+    "ort-android"
+  ],
+  "task": "token-classification",
+  "family": "distilbert",
+  "source_model_id": "OpenMed/OpenMed-NER-BloodCancerDetect-TinyMed-66M",
+  "config_path": "config.json",
+  "label_map_path": "id2label.json",
+  "artifacts": [
+    {
+      "format": "onnx-android",
+      "path": "model.onnx",
+      "precision": "float32",
+      "metadata": {
+        "profile": "android",
+        "opset": 18,
+        "execution_providers": [
+          "NNAPI",
+          "XNNPACK"
+        ],
+        "inputs": [
+          {
+            "name": "input_ids",
+            "dtype": "int64",
+            "axes": [
+              "batch",
+              "sequence"
+            ],
+            "shape": [
+              {
+                "kind": "dynamic",
+                "name": "batch"
+              },
+              {
+                "kind": "dynamic",
+                "name": "sequence"
+              }
+            ]
+          },
+          {
+            "name": "attention_mask",
+            "dtype": "int64",
+            "axes": [
+              "batch",
+              "sequence"
+            ],
+            "shape": [
+              {
+                "kind": "dynamic",
+                "name": "batch"
+              },
+              {
+                "kind": "dynamic",
+                "name": "sequence"
+              }
+            ]
+          }
+        ],
+        "outputs": [
+          {
+            "name": "logits",
+            "dtype": "float32",
+            "axes": [
+              "batch",
+              "sequence",
+              "labels"
+            ],
+            "shape": [
+              {
+                "kind": "dynamic",
+                "name": "batch"
+              },
+              {
+                "kind": "dynamic",
+                "name": "sequence"
+              },
+              {
+                "kind": "static",
+                "value": 3
+              }
+            ]
+          }
+        ],
+        "operators": [
+          "Gather",
+          "Shape",
+          "Constant",
+          "Unsqueeze",
+          "Slice",
+          "Add",
+          "LayerNormalization",
+          "Cast",
+          "Range",
+          "ConstantOfShape",
+          "GreaterOrEqual",
+          "And",
+          "Flatten",
+          "Mul",
+          "Reshape",
+          "Concat",
+          "Equal",
+          "Where",
+          "Expand",
+          "MatMul",
+          "Transpose",
+          "Softmax",
+          "IsNaN",
+          "Div",
+          "Erf"
+        ],
+        "unsupported_ops": [
+          "GreaterOrEqual",
+          "Flatten",
+          "IsNaN"
+        ],
+        "warnings": [
+          "GreaterOrEqual is not in the Android ONNX Runtime Mobile NNAPI/XNNPACK safe set",
+          "Flatten is not in the Android ONNX Runtime Mobile NNAPI/XNNPACK safe set",
+          "IsNaN is not in the Android ONNX Runtime Mobile NNAPI/XNNPACK safe set"
+        ]
+      }
+    },
+    {
+      "format": "onnx-android",
+      "path": "model_fp16.onnx",
+      "precision": "float16",
+      "metadata": {
+        "profile": "android",
+        "opset": 18,
+        "execution_providers": [
+          "NNAPI",
+          "XNNPACK"
+        ],
+        "inputs": [
+          {
+            "name": "input_ids",
+            "dtype": "int64",
+            "axes": [
+              "batch",
+              "sequence"
+            ],
+            "shape": [
+              {
+                "kind": "dynamic",
+                "name": "batch"
+              },
+              {
+                "kind": "dynamic",
+                "name": "sequence"
+              }
+            ]
+          },
+          {
+            "name": "attention_mask",
+            "dtype": "int64",
+            "axes": [
+              "batch",
+              "sequence"
+            ],
+            "shape": [
+              {
+                "kind": "dynamic",
+                "name": "batch"
+              },
+              {
+                "kind": "dynamic",
+                "name": "sequence"
+              }
+            ]
+          }
+        ],
+        "outputs": [
+          {
+            "name": "logits",
+            "dtype": "float32",
+            "axes": [
+              "batch",
+              "sequence",
+              "labels"
+            ],
+            "shape": [
+              {
+                "kind": "dynamic",
+                "name": "batch"
+              },
+              {
+                "kind": "dynamic",
+                "name": "sequence"
+              },
+              {
+                "kind": "static",
+                "value": 3
+              }
+            ]
+          }
+        ],
+        "operators": [
+          "Gather",
+          "Shape",
+          "Constant",
+          "Unsqueeze",
+          "Slice",
+          "Add",
+          "LayerNormalization",
+          "Cast",
+          "Range",
+          "ConstantOfShape",
+          "GreaterOrEqual",
+          "And",
+          "Flatten",
+          "Mul",
+          "Reshape",
+          "Concat",
+          "Equal",
+          "Where",
+          "Expand",
+          "MatMul",
+          "Transpose",
+          "Softmax",
+          "IsNaN",
+          "Div",
+          "Erf"
+        ],
+        "unsupported_ops": [
+          "GreaterOrEqual",
+          "Flatten",
+          "IsNaN"
+        ],
+        "warnings": [
+          "GreaterOrEqual is not in the Android ONNX Runtime Mobile NNAPI/XNNPACK safe set",
+          "Flatten is not in the Android ONNX Runtime Mobile NNAPI/XNNPACK safe set",
+          "IsNaN is not in the Android ONNX Runtime Mobile NNAPI/XNNPACK safe set"
+        ]
+      }
+    },
+    {
+      "format": "onnx-int8",
+      "path": "model_int8.onnx",
+      "precision": "int8",
+      "metadata": {
+        "format": "onnx-int8",
+        "quantization": {
+          "scheme": "dynamic",
+          "weight_type": "qint8",
+          "op_types": [
+            "MatMul",
+            "Gemm"
+          ],
+          "calibration": "none",
+          "target": "arm64-cpu",
+          "output": "model_int8.onnx"
+        },
+        "profile": "android",
+        "opset": 18,
+        "execution_providers": [
+          "NNAPI",
+          "XNNPACK"
+        ],
+        "inputs": [
+          {
+            "name": "input_ids",
+            "dtype": "int64",
+            "axes": [
+              "batch",
+              "sequence"
+            ],
+            "shape": [
+              {
+                "kind": "dynamic",
+                "name": "batch"
+              },
+              {
+                "kind": "dynamic",
+                "name": "sequence"
+              }
+            ]
+          },
+          {
+            "name": "attention_mask",
+            "dtype": "int64",
+            "axes": [
+              "batch",
+              "sequence"
+            ],
+            "shape": [
+              {
+                "kind": "dynamic",
+                "name": "batch"
+              },
+              {
+                "kind": "dynamic",
+                "name": "sequence"
+              }
+            ]
+          }
+        ],
+        "outputs": [
+          {
+            "name": "logits",
+            "dtype": "float32",
+            "axes": [
+              "batch",
+              "sequence",
+              "labels"
+            ],
+            "shape": [
+              {
+                "kind": "dynamic",
+                "name": "batch"
+              },
+              {
+                "kind": "dynamic",
+                "name": "sequence"
+              },
+              {
+                "kind": "static",
+                "value": 3
+              }
+            ]
+          }
+        ],
+        "operators": [
+          "Constant",
+          "Cast",
+          "Gather",
+          "ConstantOfShape",
+          "Shape",
+          "Flatten",
+          "Unsqueeze",
+          "Range",
+          "Slice",
+          "Add",
+          "LayerNormalization",
+          "DynamicQuantizeLinear",
+          "Mul",
+          "MatMulInteger",
+          "Concat",
+          "Reshape",
+          "Transpose",
+          "GreaterOrEqual",
+          "Equal",
+          "MatMul",
+          "And",
+          "Where",
+          "Expand",
+          "Softmax",
+          "IsNaN",
+          "Div",
+          "Erf"
+        ],
+        "unsupported_ops": [
+          "Flatten",
+          "DynamicQuantizeLinear",
+          "MatMulInteger",
+          "GreaterOrEqual",
+          "IsNaN"
+        ],
+        "warnings": [
+          "Flatten is not in the Android ONNX Runtime Mobile NNAPI/XNNPACK safe set",
+          "DynamicQuantizeLinear is not in the Android ONNX Runtime Mobile NNAPI/XNNPACK safe set",
+          "MatMulInteger is not in the Android ONNX Runtime Mobile NNAPI/XNNPACK safe set",
+          "GreaterOrEqual is not in the Android ONNX Runtime Mobile NNAPI/XNNPACK safe set",
+          "IsNaN is not in the Android ONNX Runtime Mobile NNAPI/XNNPACK safe set"
+        ]
+      }
+    },
+    {
+      "format": "ort-android",
+      "path": "model.ort",
+      "precision": "float32",
+      "metadata": {
+        "format": "ort-android",
+        "profile": "android",
+        "source_onnx_path": "model.onnx",
+        "optimization_style": "Fixed",
+        "target_platform": "arm",
+        "op_config_type": "required_operators_and_types",
+        "enable_type_reduction": true,
+        "ort_path": "model.ort",
+        "op_config_path": "model.required_operators_and_types.config",
+        "operators": [
+          "Gather",
+          "Shape",
+          "Constant",
+          "Unsqueeze",
+          "Slice",
+          "Add",
+          "LayerNormalization",
+          "Cast",
+          "Range",
+          "ConstantOfShape",
+          "GreaterOrEqual",
+          "And",
+          "Flatten",
+          "Mul",
+          "Reshape",
+          "Concat",
+          "Equal",
+          "Where",
+          "Expand",
+          "MatMul",
+          "Transpose",
+          "Softmax",
+          "IsNaN",
+          "Div",
+          "Erf"
+        ],
+        "opset": 18
+      }
+    }
+  ],
+  "minimum_opset": 18,
+  "max_sequence_length": 512,
+  "dynamic_shapes": {
+    "batch_axis": "dynamic",
+    "sequence_axis": "dynamic",
+    "shape_buckets": {
+      "strategy": "next_configured_bucket_or_exact",
+      "axes": {
+        "input": [
+          "batch",
+          "sequence"
+        ],
+        "output": [
+          "batch",
+          "sequence"
+        ]
+      },
+      "buckets": [
+        8,
+        16,
+        32,
+        64,
+        128,
+        256,
+        512,
+        1024,
+        2048
+      ],
+      "min_length": 8,
+      "max_length": 2048,
+      "overflow": "exact_length"
+    }
+  },
+  "optimization": {
+    "enabled": false,
+    "minimum_opset": 18
+  },
+  "validation": {},
+  "operator_fallbacks": [],
+  "tokenizer": {
+    "path": ".",
+    "files": [
+      "tokenizer.json",
+      "tokenizer_config.json"
+    ]
+  }
+}

tokenizer_config.json

@@ -0,0 +1,23 @@
+{
+  "backend": "tokenizers",
+  "clean_up_tokenization_spaces": false,
+  "cls_token": "[CLS]",
+  "do_lower_case": true,
+  "is_local": false,
+  "local_files_only": false,
+  "mask_token": "[MASK]",
+  "max_length": 512,
+  "model_max_length": 512,
+  "pad_to_multiple_of": null,
+  "pad_token": "[PAD]",
+  "pad_token_type_id": 0,
+  "padding_side": "right",
+  "sep_token": "[SEP]",
+  "stride": 0,
+  "strip_accents": null,
+  "tokenize_chinese_chars": true,
+  "tokenizer_class": "DistilBertTokenizer",
+  "truncation_side": "right",
+  "truncation_strategy": "longest_first",
+  "unk_token": "[UNK]"
+}