Export 7-class Universal BERT from best_model.pt with production ReLU architecture

README.md

@@ -0,0 +1,60 @@
+---
+language: en
+license: other
+tags:
+- text-classification
+- bert
+- insurance
+- universal
+- kinetic
+- riskguru
+pipeline_tag: text-classification
+library_name: transformers
+---
+
+# BERT 7-Class Universal Page Classifier
+
+Fine-tuned BERT model for classifying insurance document pages (Universal / Kinetic / RG / Wrap 7-class model).
+
+Used for document-type signals when OpenAI classification is unavailable (e.g. Kinetic fallback) and for page routing in RG/Wrap pipelines.
+
+## Labels
+
+| ID | Label |
+|----|-------|
+| 0 | acord |
+| 1 | contract |
+| 2 | declaration |
+| 3 | endorsements |
+| 4 | forms |
+| 5 | others |
+| 6 | rating |
+
+## Usage (RunPod / Foundry / any HF runtime)
+
+```python
+import os
+import torch
+from transformers import AutoTokenizer, AutoModel
+
+repo = "injala/bert-universal-classifier-7class"
+token = os.environ.get("HF_TOKEN")
+
+tokenizer = AutoTokenizer.from_pretrained(repo, token=token)
+model = AutoModel.from_pretrained(repo, token=token, trust_remote_code=True)
+model.eval()
+
+text = "ACORD 25 CERTIFICATE OF LIABILITY INSURANCE ..."
+inputs = tokenizer(text, return_tensors="pt", truncation=True, max_length=512)
+with torch.no_grad():
+    logits = model(**inputs)["logits"]
+    probs = torch.softmax(logits, dim=-1)
+    pred_id = probs.argmax(dim=-1).item()
+    label = model.config.id2label[str(pred_id)]
+```
+
+**Note:** Input should be full OCR page text (up to 512 tokens), not short snippets. Production uses ReLU on classifier logits (matches legacy `BERT_Model` inference).
+
+## Source
+
+Exported from `injala/rg_berts_21classes_7classes/best_model.pt`.

bert_universal_classifier_model.py

@@ -0,0 +1,43 @@
+"""Custom 7-class universal BERT page classifier (Kinetic / RG / Wrap production architecture)."""
+
+from transformers import BertConfig, BertModel, BertPreTrainedModel
+import torch.nn as nn
+
+
+class BertUniversalClassifierConfig(BertConfig):
+    model_type = "bert_universal_classifier"
+
+
+class BertUniversalClassifier(BertPreTrainedModel):
+    config_class = BertUniversalClassifierConfig
+
+    def __init__(self, config):
+        super().__init__(config)
+        self.bert = BertModel(config)
+        self.dropout = nn.Dropout(0.2)
+        self.classifier = nn.Linear(config.hidden_size, config.num_labels)
+        self.relu = nn.ReLU()
+        self.post_init()
+
+    def forward(
+        self,
+        input_ids=None,
+        attention_mask=None,
+        token_type_ids=None,
+        labels=None,
+        **kwargs,
+    ):
+        outputs = self.bert(
+            input_ids,
+            attention_mask=attention_mask,
+            token_type_ids=token_type_ids,
+        )
+        pooled_output = self.dropout(outputs.pooler_output)
+        logits = self.relu(self.classifier(pooled_output))
+
+        loss = None
+        if labels is not None:
+            loss_fn = nn.CrossEntropyLoss()
+            loss = loss_fn(logits, labels)
+
+        return {"loss": loss, "logits": logits}

config.json

@@ -0,0 +1,46 @@
+{
+  "architectures": [
+    "BertUniversalClassifier"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "classifier_dropout": null,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 768,
+  "id2label": {
+    "0": "acord",
+    "1": "contract",
+    "2": "declaration",
+    "3": "endorsements",
+    "4": "forms",
+    "5": "others",
+    "6": "rating"
+  },
+  "initializer_range": 0.02,
+  "intermediate_size": 3072,
+  "label2id": {
+    "acord": 0,
+    "contract": 1,
+    "declaration": 2,
+    "endorsements": 3,
+    "forms": 4,
+    "others": 5,
+    "rating": 6
+  },
+  "layer_norm_eps": 1e-12,
+  "max_position_embeddings": 512,
+  "model_type": "bert_universal_classifier",
+  "num_attention_heads": 12,
+  "num_hidden_layers": 12,
+  "pad_token_id": 0,
+  "position_embedding_type": "absolute",
+  "torch_dtype": "float32",
+  "transformers_version": "4.29.2",
+  "type_vocab_size": 2,
+  "use_cache": true,
+  "vocab_size": 30522,
+  "auto_map": {
+    "AutoConfig": "bert_universal_classifier_model.BertUniversalClassifierConfig",
+    "AutoModel": "bert_universal_classifier_model.BertUniversalClassifier"
+  }
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:06c52b0687f9ee9e796d3f710d9d6ab697f113e071ec3816d4d5840f9c04732f
+size 437978212

special_tokens_map.json

@@ -0,0 +1,7 @@
+{
+  "cls_token": "[CLS]",
+  "mask_token": "[MASK]",
+  "pad_token": "[PAD]",
+  "sep_token": "[SEP]",
+  "unk_token": "[UNK]"
+}

tokenizer_config.json

@@ -0,0 +1,15 @@
+{
+  "clean_up_tokenization_spaces": true,
+  "cls_token": "[CLS]",
+  "do_basic_tokenize": true,
+  "do_lower_case": true,
+  "mask_token": "[MASK]",
+  "model_max_length": 512,
+  "never_split": null,
+  "pad_token": "[PAD]",
+  "sep_token": "[SEP]",
+  "strip_accents": null,
+  "tokenize_chinese_chars": true,
+  "tokenizer_class": "BertTokenizer",
+  "unk_token": "[UNK]"
+}