Release Werea KVKK PII 150M research preview

README.md

@@ -0,0 +1,22 @@
+---
+language: [tr]
+license: apache-2.0
+pipeline_tag: token-classification
+base_model: ytu-ce-cosmos/modernbert-tr-base
+library_name: transformers
+tags: [kvkk, pii, ner, privacy, turkish, werea]
+datasets: [Werea-co/Werea-KVKK-Bench]
+---
+# Werea-KVKK-PII-150M
+
+Research-preview Turkish PII recogniser for the Werea KVKK PrivacyOps project.
+It detects PERSON, ADDRESS, EMAIL, PHONE_TR and HEALTH_DATA spans. The public
+training set is deterministic and entirely synthetic; it contains no customer
+records. Use the rules layer for checksummed identifiers and evaluate on your
+own lawyer-reviewed corpus before production deployment.
+
+Validation F1 on the synthetic split: **1.0000**.
+This score must not be interpreted as real-world compliance performance.
+
+The model cannot determine legal compliance or choose a legal basis. Human
+review is mandatory. See the dataset card and Werea PrivacyOps documentation.

config.json

@@ -0,0 +1,71 @@
+{
+  "architectures": [
+    "ModernBertForTokenClassification"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "bos_token_id": null,
+  "classifier_activation": "silu",
+  "classifier_bias": false,
+  "classifier_dropout": 0.0,
+  "classifier_pooling": "mean",
+  "cls_token_id": 2,
+  "decoder_bias": true,
+  "deterministic_flash_attn": false,
+  "dtype": "float32",
+  "embedding_dropout": 0.0,
+  "eos_token_id": 50000,
+  "global_attn_every_n_layers": 3,
+  "global_rope_theta": 160000.0,
+  "gradient_checkpointing": false,
+  "hidden_activation": "gelu",
+  "hidden_size": 768,
+  "id2label": {
+    "0": "O",
+    "1": "B-ADDRESS",
+    "2": "I-ADDRESS",
+    "3": "B-EMAIL",
+    "4": "I-EMAIL",
+    "5": "B-HEALTH_DATA",
+    "6": "I-HEALTH_DATA",
+    "7": "B-PERSON",
+    "8": "I-PERSON",
+    "9": "B-PHONE_TR",
+    "10": "I-PHONE_TR"
+  },
+  "initializer_cutoff_factor": 2.0,
+  "initializer_range": 0.02,
+  "intermediate_size": 1152,
+  "label2id": {
+    "B-ADDRESS": 1,
+    "B-EMAIL": 3,
+    "B-HEALTH_DATA": 5,
+    "B-PERSON": 7,
+    "B-PHONE_TR": 9,
+    "I-ADDRESS": 2,
+    "I-EMAIL": 4,
+    "I-HEALTH_DATA": 6,
+    "I-PERSON": 8,
+    "I-PHONE_TR": 10,
+    "O": 0
+  },
+  "layer_norm_eps": 1e-05,
+  "local_attention": 128,
+  "local_rope_theta": 10000.0,
+  "max_position_embeddings": 8192,
+  "mlp_bias": false,
+  "mlp_dropout": 0.0,
+  "model_type": "modernbert",
+  "norm_bias": false,
+  "norm_eps": 1e-05,
+  "num_attention_heads": 12,
+  "num_hidden_layers": 22,
+  "pad_token_id": 0,
+  "position_embedding_type": "absolute",
+  "repad_logits_with_grad": false,
+  "sep_token_id": 3,
+  "sparse_pred_ignore_index": -100,
+  "sparse_prediction": false,
+  "transformers_version": "4.57.1",
+  "vocab_size": 50008
+}

metrics.json

@@ -0,0 +1,11 @@
+{
+  "epoch": 3.0,
+  "eval_accuracy": 1.0,
+  "eval_f1": 1.0,
+  "eval_loss": 0.00014840990479569882,
+  "eval_precision": 1.0,
+  "eval_recall": 1.0,
+  "eval_runtime": 0.153,
+  "eval_samples_per_second": 784.567,
+  "eval_steps_per_second": 26.152
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3b1a47a777cf4b33f9bbc675f7f972abeab866fd9192ec6dc534241dd5d092d9
+size 597361556

special_tokens_map.json

@@ -0,0 +1,44 @@
+{
+  "cls_token": {
+    "content": "[CLS]",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "[EOS]",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "mask_token": {
+    "content": "[MASK]",
+    "lstrip": true,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "[PAD]",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "sep_token": {
+    "content": "[SEP]",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "unk_token": {
+    "content": "[UNK]",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

tokenizer_config.json

@@ -0,0 +1,71 @@
+{
+  "added_tokens_decoder": {
+    "0": {
+      "content": "[PAD]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "[UNK]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "[CLS]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "3": {
+      "content": "[SEP]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "4": {
+      "content": "[MASK]",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "50000": {
+      "content": "[EOS]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "clean_up_tokenization_spaces": true,
+  "cls_token": "[CLS]",
+  "do_basic_tokenize": true,
+  "do_lower_case": false,
+  "eos_token": "[EOS]",
+  "extra_special_tokens": {},
+  "mask_token": "[MASK]",
+  "model_input_names": [
+    "input_ids",
+    "attention_mask"
+  ],
+  "model_max_length": 8192,
+  "never_split": null,
+  "pad_token": "[PAD]",
+  "sep_token": "[SEP]",
+  "strip_accents": false,
+  "tokenize_chinese_chars": false,
+  "tokenizer_class": "PreTrainedTokenizerFast",
+  "unk_token": "[UNK]"
+}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7015a13568bb8c5411788167520bb74d2678063f1fd3e831c315fe785a377048
+size 5905