Add RLVR math tutor notebook

rlvr_math_tutor.ipynb

@@ -0,0 +1,191 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": 1,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "!pip install label-studio-sdk\n",
+    "\n",
+    "from label_studio_sdk import Client\n",
+    "from dotenv import load_dotenv\n",
+    "import os\n",
+    "import json"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 2,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "load_dotenv()\n",
+    "\n",
+    "LS_URL = \"http://localhost:8080\"  \n",
+    "API_KEY = os.environ.get(\"API_KEY\")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 3,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from label_studio_sdk.client import LabelStudio\n",
+    "\n",
+    "ls = LabelStudio(\n",
+    "    base_url=LS_URL,  \n",
+    "    api_key=API_KEY,\n",
+    ")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 4,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "label_config = \"\"\"\n",
+    "<View>\n",
+    "  <Text name=\"input\" value=\"$prompt\"/>\n",
+    "  <Text name=\"output\" value=\"$response\"/>\n",
+    "  <Choices name=\"reward\" toName=\"output\">\n",
+    "    <Choice value=\"1\">Good</Choice>\n",
+    "    <Choice value=\"0\">Bad</Choice>\n",
+    "  </Choices>\n",
+    "</View>\n",
+    "\"\"\""
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 5,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "\n",
+      "verify:True\n",
+      "Created project 53\n"
+     ]
+    }
+   ],
+   "source": [
+    "proj = ls.projects.create(\n",
+    "    title=\"RLVR Scoring\",\n",
+    "    description=\"Annotate responses for RLVR training.\",\n",
+    "    label_config=label_config,  \n",
+    "    color=\"#FF8800\"\n",
+    ")\n",
+    "pid = proj.id\n",
+    "print(f\"Created project {pid}\")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 6,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "Uploaded tasks\n"
+     ]
+    }
+   ],
+   "source": [
+    "tasks = [\n",
+    "    {\n",
+    "        \"prompt\": \"Solve for x: 2x + 3 = 7\",\n",
+    "        \"response\": \"Subtract 3 from both sides: 2x = 4, then divide by 2. So, x = 2.\"\n",
+    "    },\n",
+    "    {\n",
+    "        \"prompt\": \"What is the derivative of x^2?\",\n",
+    "        \"response\": \"2x\"\n",
+    "    },\n",
+    "    {\n",
+    "        \"prompt\": \"Calculate the area of a circle with radius 3\",\n",
+    "        \"response\": \"To calculate the area of a circle, use the formula Area = π * r^2. With r = 3, Area = π * 3 * 3 = 9π.\"\n",
+    "    }\n",
+    "]\n",
+    "\n",
+    "ls.projects.import_tasks(id=pid, request=tasks)\n",
+    "print(\"Uploaded tasks\")\n"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "(Human annotation happens in UI)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 7,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "Exported 3 tasks in RLVR format (prompt, response, reward)\n"
+     ]
+    }
+   ],
+   "source": [
+    "tasks = ls.tasks.list(project=proj.id)\n",
+    "\n",
+    "dataset = []\n",
+    "for task in tasks:\n",
+    "    obj = task.model_dump()\n",
+    "    item = {\n",
+    "        \"prompt\": obj[\"data\"].get(\"prompt\"),\n",
+    "        \"response\": obj[\"data\"].get(\"response\"),\n",
+    "        \"reward\": None   \n",
+    "    }\n",
+    "    \n",
+    "    for ann in obj.get(\"annotations\", []):\n",
+    "        for r in ann.get(\"result\", []):\n",
+    "            if \"value\" in r and \"choices\" in r[\"value\"]:\n",
+    "                choice = r[\"value\"][\"choices\"][0]\n",
+    "                if choice == \"Good\":\n",
+    "                    item[\"reward\"] = 1\n",
+    "                elif choice == \"Bad\":\n",
+    "                    item[\"reward\"] = 0\n",
+    "    dataset.append(item)\n",
+    "\n",
+    "\n",
+    "with open(\"rlvr_data.json\", \"w\") as f:\n",
+    "    json.dump(dataset, f, indent=2)\n",
+    "\n",
+    "print(f\"Exported {len(dataset)} tasks in RLVR format (prompt, response, reward)\")"
+   ]
+  }
+ ],
+ "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": "3.12.2"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 2
+}