Upload 10 files

Updated Version of Risc scorer and Clause Classifier models added

README.md

@@ -0,0 +1,110 @@
+---
+language: en
+license: mit
+tags:
+  - legal
+  - bert
+  - text-classification
+  - multi-task
+datasets:
+  - coastalcph/lex_glue
+---
+
+# Legal AI Risk Analyzer — V7
+
+Multi-task Legal-BERT for clause classification and risk scoring.
+
+## Performance
+| Metric | Score |
+|--------|-------|
+| Classification Accuracy | 0.8793 (87.93%) |
+| F1 Score (weighted) | 0.8750 |
+| R² Score (risk scoring) | 0.8692 |
+| MAE (risk scoring) | 3.57 points |
+| Risk Category Accuracy | 0.9336 (93.36%) |
+
+## Architecture
+- **Base model**: `nlpaueb/legal-bert-base-uncased`
+- **Task 1**: 100-class clause classification (LEDGAR)
+- **Task 2**: Risk score regression (0–100)
+- **Heads**: Linear(768→384→N) + ReLU + Dropout(0.25)
+
+## Usage
+```python
+import torch, json
+import torch.nn as nn
+from transformers import AutoTokenizer, AutoModel
+
+class MultiTaskLegalModel(nn.Module):
+    def __init__(self, model_name, num_labels,
+                 hidden_size=768, dropout_rate=0.25):
+        super().__init__()
+        self.bert       = AutoModel.from_pretrained(model_name)
+        self.dropout    = nn.Dropout(dropout_rate)
+        self.classifier = nn.Sequential(
+            nn.Linear(hidden_size, hidden_size // 2),
+            nn.ReLU(), nn.Dropout(dropout_rate),
+            nn.Linear(hidden_size // 2, num_labels)
+        )
+        self.regressor  = nn.Sequential(
+            nn.Linear(hidden_size, hidden_size // 2),
+            nn.ReLU(), nn.Dropout(dropout_rate),
+            nn.Linear(hidden_size // 2, 1),
+            nn.Sigmoid()
+        )
+    def forward(self, input_ids, attention_mask):
+        out = self.bert(input_ids=input_ids,
+                        attention_mask=attention_mask,
+                        return_dict=True)
+        cls = self.dropout(out.last_hidden_state[:, 0, :])
+        return self.classifier(cls), self.regressor(cls).squeeze(-1)
+
+
+# ── Load ──────────────────────────────────────────────────────────────────
+model_dir = "path/to/legal_ai_hf_upload"   # or HF repo id after upload
+
+with open(f"{model_dir}/metadata.json") as f:
+    meta = json.load(f)
+
+tokenizer = AutoTokenizer.from_pretrained(model_dir)
+model     = MultiTaskLegalModel(model_dir, meta['num_labels'])
+model.load_state_dict(
+    torch.load(f"{model_dir}/full_model.pt", map_location="cpu")
+)
+model.eval()
+
+# ── Inference ─────────────────────────────────────────────────────────────
+def analyse_clause(text: str) -> dict:
+    inputs = tokenizer(text, return_tensors="pt",
+                       truncation=True, max_length=256, padding=True)
+    with torch.no_grad():
+        logits, risk = model(**inputs)
+    label    = meta["label_names"][logits.argmax().item()]
+    score    = round(float(risk.item()) * 100, 1)
+    category = "Low" if score < 40 else ("Medium" if score < 70 else "High")
+    confidence = round(float(torch.softmax(logits, dim=1).max()), 3)
+    return {
+        "clause_type": label,
+        "confidence":  confidence,
+        "risk_score":  score,
+        "category":    category
+    }
+
+# Test
+result = analyse_clause(
+    "The Company shall indemnify and hold harmless from all claims "
+    "without limitation whatsoever."
+)
+print(result)
+# {'clause_type': 'Indemnifications', 'confidence': 0.91,
+#   'risk_score': 85.0, 'category': 'High'}
+```
+
+## Training details
+- Dataset: LEDGAR — 60k train / 10k val / 10k test
+- Hardware: 2× NVIDIA T4 (DataParallel)
+- Batch size: 256 (128 per GPU) | FP16 mixed precision
+- Optimizer: AdamW + Layer-wise LR Decay (γ=0.88)
+- Scheduler: Cosine annealing + warmup (6%)
+- Loss: CrossEntropyLoss (label_smoothing=0.1) + HuberLoss (δ=0.1)
+- Epochs: 15

classifier_head.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:12af285bc91b425e73f27ab5883a41e8d594df89e43f612c1d5990de929d7e05
+size 1337773

config.json

@@ -0,0 +1,34 @@
+{
+  "add_cross_attention": false,
+  "architectures": [
+    "BertModel"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "bos_token_id": 0,
+  "classifier_dropout": null,
+  "dtype": "float32",
+  "eos_token_id": null,
+  "eos_token_ids": 0,
+  "finetuning_task": null,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 768,
+  "initializer_range": 0.02,
+  "intermediate_size": 3072,
+  "is_decoder": false,
+  "layer_norm_eps": 1e-12,
+  "max_position_embeddings": 512,
+  "model_type": "bert",
+  "num_attention_heads": 12,
+  "num_hidden_layers": 12,
+  "output_past": true,
+  "pad_token_id": 0,
+  "pruned_heads": {},
+  "tie_word_embeddings": true,
+  "torchscript": false,
+  "transformers_version": "5.2.0",
+  "type_vocab_size": 2,
+  "use_bfloat16": false,
+  "use_cache": true,
+  "vocab_size": 30522
+}

full_model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9c30e9fd19d08ea0d840657146a4457349252f0505439f66c62ae9c49344013c
+size 440531744

metadata.json

@@ -0,0 +1,145 @@
+{
+  "model_version": "V7",
+  "model_name": "nlpaueb/legal-bert-base-uncased",
+  "num_labels": 100,
+  "label_names": [
+    "Adjustments",
+    "Agreements",
+    "Amendments",
+    "Anti-Corruption Laws",
+    "Applicable Laws",
+    "Approvals",
+    "Arbitration",
+    "Assignments",
+    "Assigns",
+    "Authority",
+    "Authorizations",
+    "Base Salary",
+    "Benefits",
+    "Binding Effects",
+    "Books",
+    "Brokers",
+    "Capitalization",
+    "Change In Control",
+    "Closings",
+    "Compliance With Laws",
+    "Confidentiality",
+    "Consent To Jurisdiction",
+    "Consents",
+    "Construction",
+    "Cooperation",
+    "Costs",
+    "Counterparts",
+    "Death",
+    "Defined Terms",
+    "Definitions",
+    "Disability",
+    "Disclosures",
+    "Duties",
+    "Effective Dates",
+    "Effectiveness",
+    "Employment",
+    "Enforceability",
+    "Enforcements",
+    "Entire Agreements",
+    "Erisa",
+    "Existence",
+    "Expenses",
+    "Fees",
+    "Financial Statements",
+    "Forfeitures",
+    "Further Assurances",
+    "General",
+    "Governing Laws",
+    "Headings",
+    "Indemnifications",
+    "Indemnity",
+    "Insurances",
+    "Integration",
+    "Intellectual Property",
+    "Interests",
+    "Interpretations",
+    "Jurisdictions",
+    "Liens",
+    "Litigations",
+    "Miscellaneous",
+    "Modifications",
+    "No Conflicts",
+    "No Defaults",
+    "No Waivers",
+    "Non-Disparagement",
+    "Notices",
+    "Organizations",
+    "Participations",
+    "Payments",
+    "Positions",
+    "Powers",
+    "Publicity",
+    "Qualifications",
+    "Records",
+    "Releases",
+    "Remedies",
+    "Representations",
+    "Sales",
+    "Sanctions",
+    "Severability",
+    "Solvency",
+    "Specific Performance",
+    "Submission To Jurisdiction",
+    "Subsidiaries",
+    "Successors",
+    "Survival",
+    "Tax Withholdings",
+    "Taxes",
+    "Terminations",
+    "Terms",
+    "Titles",
+    "Transactions With Affiliates",
+    "Use Of Proceeds",
+    "Vacations",
+    "Venues",
+    "Vesting",
+    "Waiver Of Jury Trials",
+    "Waivers",
+    "Warranties",
+    "Withholdings"
+  ],
+  "max_sequence_length": 256,
+  "hidden_size": 768,
+  "dropout_rate": 0.25,
+  "training_gpus": 2,
+  "test_metrics": {
+    "classification": {
+      "accuracy": 0.8793,
+      "precision": 0.8757207303564448,
+      "recall": 0.8793,
+      "f1_score": 0.8749556050721212
+    },
+    "regression": {
+      "r2_score": 0.8692134972569636,
+      "mae": 3.5734300209999086,
+      "rmse": 5.846707991168016
+    },
+    "risk_categories": {
+      "accuracy": 0.9336
+    }
+  },
+  "training_config": {
+    "epochs": 15,
+    "batch_size": 256,
+    "batch_per_gpu": 128,
+    "base_learning_rate": 4e-05,
+    "warmup_ratio": 0.06,
+    "loss_functions": {
+      "classification": "CrossEntropyLoss (label_smoothing=0.1)",
+      "regression": "Huber Loss (delta=0.1)"
+    },
+    "loss_weights": {
+      "classification": 1.0,
+      "regression": 1.0
+    },
+    "optimizer": "AdamW with LLRD (gamma=0.88)",
+    "scheduler": "Cosine Annealing with Warmup",
+    "risk_label_coverage": "100/100 classes"
+  }
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2ad1a5bca49725d153c49cef476c5a747f9a8a6f57c29d4bd89aca3bfdcc3df8
+size 437951304

regressor_head.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a51bb8a7c3f8876b7bf5590c57a11a229733e06a341e846f55c7418fa768a57a
+size 1185315

tokenizer_config.json

@@ -0,0 +1,14 @@
+{
+  "backend": "tokenizers",
+  "cls_token": "[CLS]",
+  "do_lower_case": true,
+  "is_local": false,
+  "mask_token": "[MASK]",
+  "model_max_length": 512,
+  "pad_token": "[PAD]",
+  "sep_token": "[SEP]",
+  "strip_accents": null,
+  "tokenize_chinese_chars": true,
+  "tokenizer_class": "BertTokenizer",
+  "unk_token": "[UNK]"
+}

training_history.csv

@@ -0,0 +1,16 @@
+epoch,train_loss,train_loss_cls,train_loss_reg,train_acc,val_loss,val_accuracy,val_f1,val_precision,val_recall,val_r2,val_mae,val_rmse
+1,3.6710602258114107,3.664399197253775,0.0066610286676106936,0.33708333333333335,0.5214192204177379,0.7572,0.7074415922932084,0.7036618344786799,0.7572,0.7661761656092274,5.079775939941406,7.737287896207631
+2,1.5659432710485255,1.5635212751145058,0.0024219956506598504,0.7995333333333333,0.24467970952391624,0.833,0.8135713267477054,0.8135959772541416,0.833,0.8374696367492676,3.942265531730652,6.450770938813874
+3,1.2965205502002797,1.294595095451842,0.0019254570311688362,0.8434833333333334,0.21539378352463245,0.848,0.8345476464553864,0.8447049536660243,0.848,0.8555408712095498,3.7188382574081422,6.081586153227106
+4,1.2116904644255941,1.2100293194994014,0.0016611464165567243,0.8661,0.20681057162582875,0.8584,0.8485770548769197,0.8520636039272422,0.8584,0.8619396128537031,3.690098307609558,5.945370326210087
+5,1.1546311690452251,1.153167647250155,0.0014635232742875814,0.8831333333333333,0.196885877661407,0.8648,0.8553786209185408,0.8567124767369673,0.8648,0.8661651999474788,3.606699145126343,5.853678888960011
+6,1.1107255012431043,1.1093843274928155,0.0013411752805311946,0.898,0.19397251792252063,0.8702,0.862556230588007,0.8626834743488424,0.8702,0.868818861297368,3.5736558856964113,5.795355289996768
+7,1.0735298057819935,1.0722795833932592,0.0012502233292213938,0.911,0.19419998489320278,0.8729,0.8662488873408196,0.8668921113165466,0.8729,0.8677959834232898,3.5752074138641357,5.817905901660457
+8,1.0450228313182264,1.0438580972083071,0.0011647329100982306,0.9198666666666667,0.1971759855747223,0.8745,0.8682543336719728,0.868860550038201,0.8745,0.8690140446623786,3.5184344007492063,5.791042253720064
+9,1.0214721765924006,1.0203666260901918,0.001105548023305675,0.9289,0.19810124449431896,0.8759,0.8694744478579515,0.8721128694836355,0.8759,0.8700229420868484,3.55117027759552,5.768696875215731
+10,0.9992985306902135,0.9982370878787751,0.0010614459296351575,0.9360833333333334,0.19851093888282775,0.8787,0.8738019174469928,0.8775553207188174,0.8787,0.8699872963650783,3.574739803504944,5.769487842755515
+11,0.9855700401549644,0.9845429326625581,0.0010271075374862933,0.9424,0.19890217967331408,0.8797,0.875550782111539,0.8792660052234149,0.8797,0.8693493539105841,3.5854038719177246,5.78362529705738
+12,0.9731350028768498,0.9721321176975332,0.001002886459885284,0.9464833333333333,0.2014853686094284,0.8793,0.8748119788168973,0.8774139782397148,0.8793,0.8677888459116989,3.5980608905792235,5.818062949878145
+13,0.9651041398657129,0.9641242927693306,0.0009798480299341077,0.9500333333333333,0.20219865441322327,0.8797,0.8754897760360602,0.8785927062807161,0.8797,0.8688536145318553,3.566714022064209,5.794587570320839
+14,0.9606386002073897,0.959660526539417,0.0009780740605647418,0.95155,0.20259030126035213,0.8811,0.8768527318659198,0.8807566329654987,0.8811,0.869047870671788,3.5660733938217164,5.790294461724023
+15,0.959062233630647,0.9580984607655951,0.0009637737220668412,0.95235,0.20248149074614047,0.8813,0.8771001359798906,0.8810218945645605,0.8813,0.8691144230475137,3.565014412307739,5.78882290555858