Upload test_guardrail.ipynb with huggingface_hub

test_guardrail.ipynb

@@ -3,13 +3,7 @@
   {
    "cell_type": "markdown",
    "metadata": {},
-   "source": [
-    "# Guardrail Korean E2B — Test Notebook\n",
-    "\n",
-    "Test the fine-tuned `FrameByFrame/gemma4-guardrail-ko` model for Korean + English content safety classification.\n",
-    "\n",
-    "**Capabilities:** Moderation, PII detection, Prompt injection, Blocklist, Output validation"
-   ]
+   "source": "# Guardrail Korean Gemma 4 E2B — Test Notebook\n\nTest the fine-tuned `FrameByFrame/guardrail-korean-gemma-4-E2B` model for Korean + English content safety classification.\n\n**Capabilities:** Moderation, PII detection, Prompt injection, Blocklist, Output validation, Adversarial obfuscation resistance"
   },
   {
    "cell_type": "markdown",
@@ -37,7 +31,7 @@
    "execution_count": null,
    "metadata": {},
    "outputs": [],
-   "source": "import torch\nfrom transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig\n\nMODEL_ID = \"FrameByFrame/gemma4-guardrail-ko\"\n\ntokenizer = AutoTokenizer.from_pretrained(MODEL_ID)\n\n# Load in 4-bit for free Colab GPUs. Drop the quantization_config for full precision.\nbnb_config = BitsAndBytesConfig(\n    load_in_4bit=True,\n    bnb_4bit_quant_type=\"nf4\",\n    bnb_4bit_compute_dtype=torch.bfloat16,\n)\n\nmodel = AutoModelForCausalLM.from_pretrained(\n    MODEL_ID,\n    quantization_config=bnb_config,\n    device_map=\"auto\",\n    trust_remote_code=True,\n)\nmodel.eval()\nprint(\"Model loaded.\")\n"
+   "source": "import torch\nfrom transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig\n\nMODEL_ID = \"FrameByFrame/guardrail-korean-gemma-4-E2B\"\n\ntokenizer = AutoTokenizer.from_pretrained(MODEL_ID)\n\n# Load in 4-bit for free Colab GPUs. Drop the quantization_config for full precision.\nbnb_config = BitsAndBytesConfig(\n    load_in_4bit=True,\n    bnb_4bit_quant_type=\"nf4\",\n    bnb_4bit_compute_dtype=torch.bfloat16,\n)\n\nmodel = AutoModelForCausalLM.from_pretrained(\n    MODEL_ID,\n    quantization_config=bnb_config,\n    device_map=\"auto\",\n    trust_remote_code=True,\n)\nmodel.eval()\nprint(\"Model loaded.\")\n"
   },
   {
    "cell_type": "markdown",