Upload 7 files

NoInsult.py

@@ -0,0 +1,65 @@
+import json
+import os
+import numpy as np
+import torch
+from transformers import AutoTokenizer, AutoModelForSequenceClassification
+from safetensors import safe_open
+
+MODEL_DIR = "./NoInsult/"
+
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+
+with safe_open(f"{MODEL_DIR}/model.safetensors", framework="pt", device="cpu") as f:metadata = f.metadata() or {};name = metadata.get("name", "Unknown Model");author = metadata.get("author", "Unknown Author");print(f"Loading {name} by {author}")
+
+tokenizer = AutoTokenizer.from_pretrained(MODEL_DIR)
+model = AutoModelForSequenceClassification.from_pretrained(MODEL_DIR).to(device)
+model.eval()
+
+with open(os.path.join(MODEL_DIR, "thresholds.json"), "r", encoding="utf-8") as f:config = json.load(f)
+
+LABELS = config["labels"]
+THRESHOLDS = np.array(config["thresholds"], dtype=np.float32)
+
+print("Labels:", LABELS)
+print("Thresholds:", THRESHOLDS.tolist())
+
+
+@torch.no_grad()
+def predict(texts, max_length=256):
+    """texts: str or list[str]. Returns list of dicts per input text."""
+    single = isinstance(texts, str)
+    if single:texts = [texts]
+    enc = tokenizer(
+        texts,
+        truncation=True,
+        max_length=max_length,
+        padding=True,
+        return_tensors="pt",
+    ).to(device)
+
+    logits = model(**enc).logits
+    probs = torch.sigmoid(logits).cpu().numpy()
+
+    results = []
+    for row in probs:
+        flags = (row >= THRESHOLDS)
+        results.append({
+            "scores": {label: float(p) for label, p in zip(LABELS, row)},
+            "flags": {label: bool(f) for label, f in zip(LABELS, flags)},
+            "any_flagged": bool(flags.any()),
+        })
+
+    return results[0] if single else results
+
+
+if __name__ == "__main__":
+    examples = [
+        "Have a great day, thanks for the help!",
+        "You are such an idiot, shut up.",
+    ]
+    for text, result in zip(examples, predict(examples)):
+        print("\nText:", text)
+        for label in LABELS:
+            marker = "⚠️ " if result["flags"][label] else "   "
+            print(f"  {marker}{label:18s} {result['scores'][label]:.4f}")
+        print("     any_flagged:", result["any_flagged"])

NoInsult/config.json

@@ -0,0 +1,48 @@
+{
+  "add_cross_attention": false,
+  "architectures": [
+    "BertForSequenceClassification"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "bos_token_id": null,
+  "classifier_dropout": null,
+  "dtype": "float32",
+  "eos_token_id": null,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 768,
+  "id2label": {
+    "0": "toxicity",
+    "1": "severe_toxicity",
+    "2": "obscene",
+    "3": "threat",
+    "4": "insult",
+    "5": "identity_attack",
+    "6": "sexual_explicit"
+  },
+  "initializer_range": 0.02,
+  "intermediate_size": 3072,
+  "is_decoder": false,
+  "label2id": {
+    "identity_attack": 5,
+    "insult": 4,
+    "obscene": 2,
+    "severe_toxicity": 1,
+    "sexual_explicit": 6,
+    "threat": 3,
+    "toxicity": 0
+  },
+  "layer_norm_eps": 1e-12,
+  "max_position_embeddings": 512,
+  "model_type": "bert",
+  "num_attention_heads": 12,
+  "num_hidden_layers": 12,
+  "pad_token_id": 0,
+  "position_embedding_type": "absolute",
+  "problem_type": "multi_label_classification",
+  "tie_word_embeddings": true,
+  "transformers_version": "5.0.0",
+  "type_vocab_size": 2,
+  "use_cache": false,
+  "vocab_size": 30522
+}

NoInsult/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f494a163eab544d8efc48fb330cf084a15ab8ce45b705e9c140f80384aa1aa62
+size 437974036

NoInsult/thresholds.json

@@ -0,0 +1,20 @@
+{
+  "labels": [
+    "toxicity",
+    "severe_toxicity",
+    "obscene",
+    "threat",
+    "insult",
+    "identity_attack",
+    "sexual_explicit"
+  ],
+  "thresholds": [
+    0.909860372543335,
+    0.5,
+    0.983537495136261,
+    0.9544731974601746,
+    0.9325506091117859,
+    0.9784075617790222,
+    0.9820615649223328
+  ]
+}

NoInsult/tokenizer_config.json

@@ -0,0 +1,16 @@
+{
+  "backend": "tokenizers",
+  "cls_token": "[CLS]",
+  "do_basic_tokenize": true,
+  "do_lower_case": true,
+  "is_local": false,
+  "mask_token": "[MASK]",
+  "model_max_length": 1000000000000000019884624838656,
+  "never_split": null,
+  "pad_token": "[PAD]",
+  "sep_token": "[SEP]",
+  "strip_accents": null,
+  "tokenize_chinese_chars": true,
+  "tokenizer_class": "BertTokenizer",
+  "unk_token": "[UNK]"
+}

use.py

@@ -0,0 +1,5 @@
+from NoInsult import predict
+
+result = predict("I will fuck you up, you piece of shit.")
+print(result["flags"])   # {'toxicity': True, 'severe_toxicity': False, ...}
+print(result["scores"])  # raw sigmoid probabilities per label