From 3f4fe8c7c1dfede63fcf6a22de52b9c3251b5dd1 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=D0=9C=D0=B0=D0=B7=D1=83=D1=80=20=D0=93=D1=80=D0=B5=D1=82?=
=?UTF-8?q?=D0=B0=20=D0=95=D0=B2=D0=B3=D0=B5=D0=BD=D1=8C=D0=B5=D0=B2=D0=BD?=
=?UTF-8?q?=D0=B0?= <gemazur_1@edu.hse.ru>
Date: Mon, 24 Mar 2025 15:46:45 +0300
Subject: [PATCH] proverka obuchenya
---
.ipynb_checkpoints/proverkabert-checkpoint.py | 6 ++++--
proverkabert.py | 6 ++++--
2 files changed, 8 insertions(+), 4 deletions(-)
diff --git a/.ipynb_checkpoints/proverkabert-checkpoint.py b/.ipynb_checkpoints/proverkabert-checkpoint.py
index 61952b9..32036d5 100644
--- a/.ipynb_checkpoints/proverkabert-checkpoint.py
+++ b/.ipynb_checkpoints/proverkabert-checkpoint.py
@@ -183,9 +183,11 @@ def classify_prompt(prompt, tokenizer, model):
"safety": safety_label,
"attack_type": attack_label,
"safety_confidence": round(probs_safety[0, pred_safety].item(), 4),
- "attack_confidence": round(probs_attack[0, pred_attack].item(), 4) if safety_label == "unsafe" else 0.0
+ "attack_confidence": round(probs_attack[0, pred_attack].item(), 4) if safety_label == "unsafe" else 0.0,
+ "safety_probs": probs_safety.tolist(),
+ "attack_probs": probs_attack.tolist()
}
-
+
def main():
MODEL_PATH = "./fine-tuned-bert-lora-multi-task"
diff --git a/proverkabert.py b/proverkabert.py
index 61952b9..32036d5 100644
--- a/proverkabert.py
+++ b/proverkabert.py
@@ -183,9 +183,11 @@ def classify_prompt(prompt, tokenizer, model):
"safety": safety_label,
"attack_type": attack_label,
"safety_confidence": round(probs_safety[0, pred_safety].item(), 4),
- "attack_confidence": round(probs_attack[0, pred_attack].item(), 4) if safety_label == "unsafe" else 0.0
+ "attack_confidence": round(probs_attack[0, pred_attack].item(), 4) if safety_label == "unsafe" else 0.0,
+ "safety_probs": probs_safety.tolist(),
+ "attack_probs": probs_attack.tolist()
}
-
+
def main():
MODEL_PATH = "./fine-tuned-bert-lora-multi-task"
--
GitLab