Upload meta_v2.json with huggingface_hub
Browse files- meta_v2.json +54 -0
meta_v2.json
ADDED
|
@@ -0,0 +1,54 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"model": "cortyx_v2_final.pt",
|
| 3 |
+
"base": "microsoft/deberta-v3-small",
|
| 4 |
+
"version": "v2.0",
|
| 5 |
+
"epochs": 10,
|
| 6 |
+
"best_f1_macro": 0.6129,
|
| 7 |
+
"labels": [
|
| 8 |
+
"safe",
|
| 9 |
+
"mild_toxicity",
|
| 10 |
+
"severe_toxicity",
|
| 11 |
+
"harassment",
|
| 12 |
+
"hate_speech",
|
| 13 |
+
"threat",
|
| 14 |
+
"insult",
|
| 15 |
+
"profanity",
|
| 16 |
+
"sexual_content",
|
| 17 |
+
"violence",
|
| 18 |
+
"self_harm",
|
| 19 |
+
"extremism",
|
| 20 |
+
"illegal_instruction",
|
| 21 |
+
"jailbreak_attempt",
|
| 22 |
+
"prompt_injection",
|
| 23 |
+
"obfuscated_toxicity",
|
| 24 |
+
"misinformation_risk"
|
| 25 |
+
],
|
| 26 |
+
"thresholds": {
|
| 27 |
+
"safe": 0.5,
|
| 28 |
+
"mild_toxicity": 0.7,
|
| 29 |
+
"severe_toxicity": 0.4,
|
| 30 |
+
"harassment": 0.5,
|
| 31 |
+
"hate_speech": 0.45,
|
| 32 |
+
"threat": 0.4,
|
| 33 |
+
"insult": 0.55,
|
| 34 |
+
"profanity": 0.6,
|
| 35 |
+
"sexual_content": 0.45,
|
| 36 |
+
"violence": 0.4,
|
| 37 |
+
"self_harm": 0.35,
|
| 38 |
+
"extremism": 0.4,
|
| 39 |
+
"illegal_instruction": 0.45,
|
| 40 |
+
"jailbreak_attempt": 0.45,
|
| 41 |
+
"prompt_injection": 0.45,
|
| 42 |
+
"obfuscated_toxicity": 0.5,
|
| 43 |
+
"misinformation_risk": 0.5
|
| 44 |
+
},
|
| 45 |
+
"datasets": [
|
| 46 |
+
"QuantaSparkLabs/cortyx-safety-dataset",
|
| 47 |
+
"google/civil_comments",
|
| 48 |
+
"lmsys/toxic-chat",
|
| 49 |
+
"lmsys/lmsys-chat-1m",
|
| 50 |
+
"cardiffnlp/tweet_eval"
|
| 51 |
+
],
|
| 52 |
+
"training_samples": 1844,
|
| 53 |
+
"val_samples": 462
|
| 54 |
+
}
|