waterfall 0.2.3__tar.gz → 0.2.5__tar.gz

{waterfall-0.2.3 → waterfall-0.2.5}/.gitignore

@@ -3,5 +3,4 @@
 *.egg-info
 /.python-version
 /.idea
-dist
-media
+dist

{waterfall-0.2.3 → waterfall-0.2.5}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: waterfall
-Version: 0.2.3
+Version: 0.2.5
 Summary: Scalable Framework for Robust Text Watermarking and Provenance for LLMs
 Project-URL: Homepage, https://github.com/aoi3142/Waterfall
 Project-URL: Issues, https://github.com/aoi3142/Waterfall/issues
@@ -11,7 +11,7 @@ Classifier: Operating System :: OS Independent
 Classifier: Programming Language :: Python :: 3
 Requires-Python: >=3.10
 Requires-Dist: accelerate>=0.29.0
-Requires-Dist: numpy>=2.0.0
+Requires-Dist: numpy>=1.25.0
 Requires-Dist: scipy>=1.13.0
 Requires-Dist: sentence-transformers>=3.0.0
 Requires-Dist: torch>=2.3.0

waterfall-0.2.5/pypi.txt

@@ -0,0 +1,2 @@
+python3 -m build
+python3 -m twine upload dist/*

{waterfall-0.2.3 → waterfall-0.2.5}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "waterfall"
-version = "0.2.3"
+version = "0.2.5"
 authors = [
   { name = "Xinyuan Niu", email="aperture@outlook.sg" }
 ]
@@ -20,7 +20,7 @@ license-files = ["LICEN[CS]E*"]
 
 dependencies = [
   "accelerate>=0.29.0",
-  "numpy>=2.0.0",
+  "numpy>=1.25.0",
   "scipy>=1.13.0",
   "sentence-transformers>=3.0.0",
   "torch>=2.3.0",

{waterfall-0.2.3 → waterfall-0.2.5}/requirements.txt

@@ -1,5 +1,5 @@
 accelerate>=0.29.0
-numpy>=2.0.0
+numpy>=1.25.0
 scipy>=1.13.0
 sentence-transformers>=3.0.0
 torch>=2.3.0

waterfall-0.2.5/test.ipynb

@@ -0,0 +1,129 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": 1,
+   "id": "ddca610d",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from waterfall.watermark import *\n",
+    "import torch"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 2,
+   "id": "f0b0a02a",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "model_name_or_path = 'meta-llama/Llama-3.1-8B-Instruct'"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 3,
+   "id": "ca5f7d55",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "2ef1e58174204e879a462a2a829d74fa",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Loading checkpoint shards:   0%|          | 0/4 [00:00<?, ?it/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    }
+   ],
+   "source": [
+    "\n",
+    "tokenizer = AutoTokenizer.from_pretrained(model_name_or_path)\n",
+    "model = AutoModelForCausalLM.from_pretrained(\n",
+    "    model_name_or_path,\n",
+    "    torch_dtype=torch.bfloat16,\n",
+    "    device_map=\"cuda:5\",\n",
+    "    )"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 4,
+   "id": "2c475de5",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "watermarker = Watermarker(tokenizer=tokenizer, model=model, id=1)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 5,
+   "id": "e74a58b5",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "texts = [\"James is a scary man who loves to scare children\", \"There were 3 birds that ate the whole bowl\"]"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 8,
+   "id": "fd0af01e",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "watermarked_text = watermark_texts(texts, id, watermarker=watermarker)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 10,
+   "id": "4d58ec82",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "['James is an intimidating individual known for taking great pleasure in frightening kids, and this tendency often sends them running in terror.',\n",
+       " \"A small flock comprised of three birds consumed the entirety of the offered bowl's contents within a single meal session.\"]"
+      ]
+     },
+     "execution_count": 10,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "watermarked_text"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.12.9"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}

waterfall-0.2.5/test_.ipynb

@@ -0,0 +1,118 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": 49,
+   "id": "116f61a2",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import numpy as np\n",
+    "from tqdm import tqdm\n",
+    "from itertools import product, combinations\n",
+    "from scipy.stats import binom"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 50,
+   "id": "53c66973",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "np.float64(0.38742048899999965)"
+      ]
+     },
+     "execution_count": 50,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "binom.pmf(1, 10, 0.1)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 56,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "100%|██████████| 32/32 [00:00<00:00, 10845.88it/s]\n",
+      "100%|██████████| 32/32 [00:00<00:00, 3035.09it/s]\n",
+      "100%|██████████| 32/32 [00:00<00:00, 1622.42it/s]\n",
+      "100%|██████████| 32/32 [00:00<00:00, 2020.93it/s]\n",
+      "100%|██████████| 32/32 [00:00<00:00, 4749.72it/s]"
+     ]
+    },
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "0.6582226332676197\n"
+     ]
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "\n"
+     ]
+    }
+   ],
+   "source": [
+    "n = 5   # n-gram\n",
+    "# k = 1   # no of bits modified\n",
+    "l = 8   # bits of watermark\n",
+    "p = 0.1 # probability of modification\n",
+    "\n",
+    "result = 0\n",
+    "for k in range(0, n + 1):\n",
+    "    binomial_prob = binom.pmf(k, n, p)\n",
+    "    if k == 0:\n",
+    "        result += binomial_prob * 1\n",
+    "        continue\n",
+    "    results = []\n",
+    "\n",
+    "    original = np.zeros(n, dtype=bool)\n",
+    "    for original in tqdm(product([True, False], repeat=n), total=2**n):\n",
+    "        original = np.array(original, dtype=bool)\n",
+    "        original_agreement = original.mean() > 0.5\n",
+    "        for modified_indices in combinations(range(n), k):\n",
+    "            modified = original.copy()\n",
+    "            for i in product([True, False], repeat=k):\n",
+    "                modified[np.array(modified_indices)] = i\n",
+    "                modified_agreement = modified.mean() > 0.5\n",
+    "                results.append(original_agreement == modified_agreement)\n",
+    "    result += binomial_prob * np.mean(results) ** l\n",
+    "print(result)"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.12.9"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}

{waterfall-0.2.3 → waterfall-0.2.5}/waterfall/WatermarkerBase.py

@@ -116,7 +116,7 @@ class Watermarker:
         self.N = self.tokenizer.vocab_size
 
         if isinstance(model, str):
-            self.load_model(model, device_map=device, dtype=dtype)
+            self.load_model(model, device_map=device if device is not None else "auto", dtype=dtype)
         else:
             self.model = model