Best checkpoint - DistilBERT prompt injection detector, F1=1.0 val, 0.959 test

config.json

@@ -0,0 +1,36 @@
+{
+  "activation": "gelu",
+  "architectures": [
+    "DistilBertForSequenceClassification"
+  ],
+  "attention_dropout": 0.1,
+  "bos_token_id": null,
+  "dim": 768,
+  "dropout": 0.1,
+  "dtype": "float32",
+  "eos_token_id": null,
+  "hidden_dim": 3072,
+  "id2label": {
+    "0": "safe",
+    "1": "injection"
+  },
+  "initializer_range": 0.02,
+  "label2id": {
+    "injection": 1,
+    "safe": 0
+  },
+  "max_position_embeddings": 512,
+  "model_type": "distilbert",
+  "n_heads": 12,
+  "n_layers": 6,
+  "pad_token_id": 0,
+  "problem_type": "single_label_classification",
+  "qa_dropout": 0.1,
+  "seq_classif_dropout": 0.2,
+  "sinusoidal_pos_embds": false,
+  "tie_weights_": true,
+  "tie_word_embeddings": true,
+  "transformers_version": "5.8.1",
+  "use_cache": false,
+  "vocab_size": 30522
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f3ec844b481b829bfd89701701cf86ed6be606573f76e8ec3fa27b135dfc2387
+size 267832560

tokenizer_config.json

@@ -0,0 +1,15 @@
+{
+  "backend": "tokenizers",
+  "cls_token": "[CLS]",
+  "do_lower_case": true,
+  "is_local": false,
+  "local_files_only": 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]"
+}

trainer_state.json

@@ -0,0 +1,116 @@
+{
+  "best_global_step": 99,
+  "best_metric": 1.0,
+  "best_model_checkpoint": "./pi-detector/checkpoint-99",
+  "epoch": 1.0,
+  "eval_steps": 500,
+  "global_step": 99,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.010101010101010102,
+      "grad_norm": 1.3391671180725098,
+      "learning_rate": 0.0,
+      "loss": 0.6838793754577637,
+      "step": 1
+    },
+    {
+      "epoch": 0.10101010101010101,
+      "grad_norm": 1.2506626844406128,
+      "learning_rate": 1.8000000000000001e-06,
+      "loss": 0.6808025042215983,
+      "step": 10
+    },
+    {
+      "epoch": 0.20202020202020202,
+      "grad_norm": 1.1518597602844238,
+      "learning_rate": 3.8000000000000005e-06,
+      "loss": 0.6861461639404297,
+      "step": 20
+    },
+    {
+      "epoch": 0.30303030303030304,
+      "grad_norm": 1.91855788230896,
+      "learning_rate": 5.8e-06,
+      "loss": 0.6577313423156739,
+      "step": 30
+    },
+    {
+      "epoch": 0.40404040404040403,
+      "grad_norm": 1.835654854774475,
+      "learning_rate": 7.800000000000002e-06,
+      "loss": 0.5954501628875732,
+      "step": 40
+    },
+    {
+      "epoch": 0.5050505050505051,
+      "grad_norm": 1.999822735786438,
+      "learning_rate": 9.800000000000001e-06,
+      "loss": 0.5008408069610596,
+      "step": 50
+    },
+    {
+      "epoch": 0.6060606060606061,
+      "grad_norm": 2.83133864402771,
+      "learning_rate": 1.18e-05,
+      "loss": 0.37516820430755615,
+      "step": 60
+    },
+    {
+      "epoch": 0.7070707070707071,
+      "grad_norm": 1.6260528564453125,
+      "learning_rate": 1.38e-05,
+      "loss": 0.29813175201416015,
+      "step": 70
+    },
+    {
+      "epoch": 0.8080808080808081,
+      "grad_norm": 1.7019435167312622,
+      "learning_rate": 1.58e-05,
+      "loss": 0.2266446352005005,
+      "step": 80
+    },
+    {
+      "epoch": 0.9090909090909091,
+      "grad_norm": 1.1480830907821655,
+      "learning_rate": 1.7800000000000002e-05,
+      "loss": 0.17041007280349732,
+      "step": 90
+    },
+    {
+      "epoch": 1.0,
+      "eval_accuracy": 1.0,
+      "eval_f1": 1.0,
+      "eval_loss": 0.032340291887521744,
+      "eval_precision": 1.0,
+      "eval_recall": 1.0,
+      "eval_runtime": 0.2543,
+      "eval_samples_per_second": 503.339,
+      "eval_steps_per_second": 15.729,
+      "step": 99
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 495,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 5,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 32576631053856.0,
+  "train_batch_size": 16,
+  "trial_name": null,
+  "trial_params": null
+}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ae22f7875c9b5a884d3c1d28b340e1a4ddf3dd21a23f7ab1eece96a498a9a384
+size 5329