update

app.py

@@ -1,36 +1,79 @@
+import os
+import json
+import re
 from flask import Flask, request, jsonify
 from transformers import AutoTokenizer, AutoModelForSequenceClassification
 import torch
 
 app = Flask(__name__)
 
-# Load model dan tokenizer
-tokenizer = AutoTokenizer.from_pretrained("./hate_speech_model")
-model = AutoModelForSequenceClassification.from_pretrained("./hate_speech_model")
+MODEL_PATH = "./hate_speech_model"
+
+tokenizer = AutoTokenizer.from_pretrained(MODEL_PATH)
+model = AutoModelForSequenceClassification.from_pretrained(MODEL_PATH)
+
+threshold = 0.5
+if os.path.exists(os.path.join(MODEL_PATH, "training_config.json")):
+    try:
+        with open(os.path.join(MODEL_PATH, "training_config.json"), "r") as f:
+            config = json.load(f)
+            threshold = config.get("optimal_threshold", 0.5)
+            print(f"Loaded optimal threshold from config: {threshold:.4f}")
+    except Exception as e:
+        print(f"Gagal memuat training_config.json: {e}. Menggunakan threshold default 0.5")
+
+def preprocess_text(text):
+
+    text = re.sub(r'@[^\s]+', '@USER', text)
+    text = re.sub(r'https?://[^\s]+', 'HTTPURL', text)
+    text = re.sub(r'\s+', ' ', text).strip()
+    return text
 
 @app.route("/")
 def home():
-    return jsonify({"message": "Hate Speech API is running!"})
+    return jsonify({
+        "message": "Hate Speech Detection API",
+        "model_path": MODEL_PATH,
+        "optimal_threshold": threshold
+    })
 
 @app.route("/predict", methods=["POST"])
 def predict():
+    # Verifikasi custom API key untuk mencegah penyalahgunaan API publik
+    api_key = request.headers.get("X-API-Key")
+    if api_key != "sh1eld-hate-speech-key-2026":
+        return jsonify({"error": "Akses ditolak: API Key tidak valid atau tidak disediakan"}), 401
+
     data = request.get_json()
 
     if not data or "text" not in data:
-        return jsonify({"error": "Please provide 'text'"}), 400
+        return jsonify({"error": "Silakan kirimkan parameter 'text'"}), 400
 
     text = data["text"]
+    
+    # Preprocessing teks sebelum diklasifikasikan
+    text_clean = preprocess_text(text)
 
-    inputs = tokenizer(text, return_tensors="pt", truncation=True, padding=True)
+    inputs = tokenizer(text_clean, return_tensors="pt", truncation=True, padding=True)
     with torch.no_grad():
         outputs = model(**inputs)
         probs = torch.nn.functional.softmax(outputs.logits, dim=1)
-        label = "Hate" if torch.argmax(probs).item() == 1 else "Non-Hate"
+        hate_prob = probs[0][1].item()
+        
+        # Terapkan threshold optimal untuk deteksi kelas Hate
+        pred_label = 1 if hate_prob >= threshold else 0
+        label = "Hate" if pred_label == 1 else "Non-Hate"
+        confidence = hate_prob if pred_label == 1 else (1 - hate_prob)
 
     return jsonify({
         "label": label,
-        "confidence": float(probs.max())
+        "confidence": confidence,
+        "probabilities": {
+            "Non-Hate": float(probs[0][0]),
+            "Hate": float(probs[0][1])
+        }
     })
 
 if __name__ == "__main__":
     app.run(host="0.0.0.0", port=7860)
+

hate_speech_model/config.json

@@ -1,17 +1,22 @@
 {
-  "_num_labels": 5,
+  "add_cross_attention": false,
   "architectures": [
     "BertForSequenceClassification"
   ],
   "attention_probs_dropout_prob": 0.1,
+  "bos_token_id": 0,
   "classifier_dropout": null,
-  "directionality": "bidi",
   "dtype": "float32",
+  "eos_token_id": null,
+  "eos_token_ids": 0,
+  "finetuning_task": null,
+  "gradient_checkpointing": false,
   "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",
@@ -19,15 +24,12 @@
   "num_hidden_layers": 12,
   "output_past": true,
   "pad_token_id": 0,
-  "pooler_fc_size": 768,
-  "pooler_num_attention_heads": 12,
-  "pooler_num_fc_layers": 3,
-  "pooler_size_per_head": 128,
-  "pooler_type": "first_token_transform",
-  "position_embedding_type": "absolute",
-  "problem_type": "single_label_classification",
-  "transformers_version": "4.57.1",
+  "pruned_heads": {},
+  "tie_word_embeddings": true,
+  "torchscript": false,
+  "transformers_version": "5.0.0",
   "type_vocab_size": 2,
-  "use_cache": true,
-  "vocab_size": 50000
+  "use_bfloat16": false,
+  "use_cache": false,
+  "vocab_size": 31923
 }

hate_speech_model/model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ad083ce4110a1a1ff4a3210ff6ba26fce90e369318f66eb70e62657e211eed46
-size 497795072
+oid sha256:f3bce66292ecf2ad3a0cc46a18cdd4dd144c2cf8fa539dea7807951a7ad1bf88
+size 442262496

hate_speech_model/special_tokens_map.json

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

hate_speech_model/tokenizer_config.json

@@ -1,54 +1,10 @@
 {
-  "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": false,
-      "normalized": false,
-      "rstrip": false,
-      "single_word": false,
-      "special": true
-    }
-  },
-  "clean_up_tokenization_spaces": true,
+  "backend": "tokenizers",
   "cls_token": "[CLS]",
-  "do_basic_tokenize": true,
-  "do_lower_case": true,
-  "extra_special_tokens": {},
+  "do_lower_case": false,
+  "is_local": false,
   "mask_token": "[MASK]",
   "model_max_length": 1000000000000000019884624838656,
-  "never_split": null,
   "pad_token": "[PAD]",
   "sep_token": "[SEP]",
   "strip_accents": null,

hate_speech_model/training_config.json

@@ -0,0 +1,20 @@
+{
+  "optimal_threshold": 0.4299999999999998,
+  "class_weights": {
+    "non_hate": 0.5966592554586397,
+    "hate": 3.0864051902093776
+  },
+  "label_mapping": {
+    "0": "Non-Hate",
+    "1": "Hate"
+  },
+  "model_name": "indolem/indobertweet-base-uncased",
+  "max_length": 128,
+  "dataset": "Exqrch/IndoDiscourse",
+  "metrics": {
+    "accuracy": 0.7902178066488346,
+    "precision": 0.3990306946688207,
+    "recall": 0.5825471698113207,
+    "f1_score": 0.473633748801534
+  }
+}