update

flexflow · Feb 13, 2025 · 4d75ff6 · 4d75ff6
1 parent be5edcf
commit 4d75ff6
Showing 1 changed file with 119 additions and 0 deletions.
diff --git a/suffix_decoding/helper.ipynb b/suffix_decoding/helper.ipynb
@@ -0,0 +1,119 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": 22,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import json\n",
+    "from collections import OrderedDict\n",
+    "original_fp=\"/home/jovyan/pscratch/suffix-tree-decoding/trace/llama70b/cortex-llama3.1-70b.json\"\n",
+    "new_fp=\"/home/jovyan/pscratch/suffix-tree-decoding/trace/llama70b/cortex-llama3.1-70b_debug.json\"\n",
+    "\n",
+    "with open(original_fp, \"r\") as f:\n",
+    "    original_data = json.load(f, object_pairs_hook=OrderedDict)\n",
+    "partitions = original_data[\"partitions\"]\n",
+    "# remove partitions whose partition_name does not match \"FEATURE_EXTRACTION\"\n",
+    "partitions = [p for p in partitions if p[\"partition_name\"] == \"FEATURE_EXTRACTION\"]\n",
+    "assert len(partitions) == 1\n",
+    "partitions[0][\"eval_entries\"] = [e for e in partitions[0][\"eval_entries\"] if e[\"prompt_length\"] == 967]\n",
+    "partitions[0][\"eval_entries\"] = partitions[0][\"eval_entries\"][:1]\n",
+    "\n",
+    "new_json = OrderedDict({\n",
+    "    \"partitions\": partitions,\n",
+    "    \"metadata\": original_data[\"metadata\"],\n",
+    "})\n",
+    "\n",
+    "# save data back to json\n",
+    "with open(new_fp, \"w\") as f:\n",
+    "    json.dump(new_json, f, indent=2)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 23,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "[720, 220, 330, 46458, 30015, 794, 905, 11, 720, 220, 330, 20693, 291, 794, 837, 345, 220, 330, 16608, 42394, 30015, 794, 905, 345, 220, 330, 13430, 39341, 2987, 794, 905, 198, 534, 74694]\n"
+     ]
+    }
+   ],
+   "source": [
+    "import os\n",
+    "os.environ[\"HF_HOME\"] = \"/home/jovyan/pscratch/.cache/huggingface\"\n",
+    "# tokenize a sentence with llama 3.1 70b\n",
+    "from transformers import AutoTokenizer\n",
+    "\n",
+    "tokenizer = AutoTokenizer.from_pretrained(\"meta-llama/Llama-3.1-70B-Instruct\")\n",
+    "sentence = \" \\n  \\\"ratio_question\\\": false, \\n  \\\"timed\\\": true,\\n  \\\"rolling_metric_question\\\": false,\\n  \\\"rank_calculation\\\": false\\n}\\n```\"\n",
+    "correct_output=\" \\n  \\\"ratio_question\\\": true, \\n  \\\"last_period\\\": false, \\n  \\\"timedimension_aggregation\\\": true, \\n  \\\"missing_time_period\\\": false, \\n  \\\"period_over_period_question\\\": true, \\n  \\\"rolling_metric_question\\\": false, \\n  \\\"consecutive_trend_question\\\": false, \\n  \\\"rank_calculation\\\": false \\n}\\n```\"\n",
+    "tokens = tokenizer.encode(sentence, add_special_tokens=False)\n",
+    "print(tokens)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 24,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "' \\n  \"ratio_question\": false, \\n  \"tim'"
+      ]
+     },
+     "execution_count": 24,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "tokenizer.decode([720, 220, 330, 46458, 30015, 794, 905, 11, 720, 220, 330, 20693])"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 26,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "[720, 220, 330, 46458, 30015, 794, 837, 11, 720, 220, 330, 4354, 21485, 794, 905, 11, 720, 220, 330, 20693, 291, 18658, 21233, 35542, 794, 837, 11, 720, 220, 330, 31716, 3084, 21485, 794, 905, 11, 720, 220, 330, 19862, 15793, 21485, 30015, 794, 837, 11, 720, 220, 330, 16608, 42394, 30015, 794, 905, 11, 720, 220, 330, 444, 86880, 530, 9484, 30015, 794, 905, 11, 720, 220, 330, 13430, 39341, 2987, 794, 905, 720, 534, 74694]\n"
+     ]
+    }
+   ],
+   "source": [
+    "tokens=tokenizer.encode(correct_output, add_special_tokens=False)\n",
+    "print(tokens)"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "base",
+   "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": 2
+}