{ "cells": [ { "cell_type": "code", "execution_count": 1, "execution_state": "idle", "id": "1ddfc692-bda7-4d38-a549-2fb0d40d437d", "metadata": {}, "outputs": [], "source": [ "import math\n", "import os\n", "import re\n", "import sys\n", "\n", "sys.path.append(\"..\")\n", "\n", "import torch\n", "import transformers\n", "from safetensors import safe_open\n", "\n", "from utils import add_attn_hooks\n", "\n", "# text_dir = \"rtl\"\n", "text_dir = \"ltr\"\n", "# tokenizer = transformers.AutoTokenizer.from_pretrained(\"bert-base-uncased\")\n", "# model = transformers.AutoModelForMaskedLM.from_pretrained(f\"/home/sipb/nlp-class-project/checkpoints/bert_base_{text_dir}/epoch_3_checkpt\", ignore_mismatched_sizes=True)\n", "\n", "tokenizer = transformers.AutoTokenizer.from_pretrained(\"distilbert/distilbert-base-uncased\")" ] }, { "cell_type": "code", "execution_count": 6, "execution_state": "idle", "id": "5bd236ae-119c-4ea6-9a0f-03272f528caf", "metadata": {}, "outputs": [], "source": [] }, { "cell_type": "code", "execution_count": 19, "execution_state": "idle", "id": "eaf99031-9141-43dd-89ba-be9b8e63a1ba", "metadata": {}, "outputs": [], "source": [ "with open(\"/home/sipb/nlp-class-project/data/riddles.txt\", \"r\") as f:\n", " riddles = [line.rstrip() for line in f.readlines()]\n", "# with open(\"/home/sipb/nlp-class-project/data/ltr_riddles.txt\", \"r\") as f:\n", "# riddles_aq = [line.rstrip() for line in f.readlines()]" ] }, { "cell_type": "code", "execution_count": 22, "execution_state": "idle", "id": "9652550f-61b9-4b9b-ad10-9d9873a9e80b", "metadata": {}, "outputs": [], "source": [ "pattern = r\"^(.*)(What am I\\?\\s*)(.+)$\"\n", "\n", "riddles_qa = []\n", "riddles_aq = []\n", "for riddle in riddles:\n", " riddles_qa.append(re.sub(pattern, r\"Q: \\1\\2A: \\3\", riddle))\n", " riddles_aq.append(re.sub(pattern, r\"A: \\3 Q: \\1\\2\", riddle).rstrip())" ] }, { "cell_type": "code", "execution_count": 28, "execution_state": "idle", "id": "94da0be0-d6ef-46be-9fff-4ebf022e4fed", "metadata": {}, "outputs": [ { "name": "stderr", "output_type": "stream", "text": [ "Some weights of the model checkpoint at /home/sipb/nlp-class-project/checkpoints/distilbert_base_ltr/epoch_3_checkpt were not used when initializing DistilBertForMaskedLM: ['attention_mask']\n", "- This IS expected if you are initializing DistilBertForMaskedLM from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).\n", "- This IS NOT expected if you are initializing DistilBertForMaskedLM from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).\n", "Some weights of the model checkpoint at /home/sipb/nlp-class-project/checkpoints/distilbert_base_rtl/epoch_3_checkpt were not used when initializing DistilBertForMaskedLM: ['attention_mask']\n", "- This IS expected if you are initializing DistilBertForMaskedLM from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).\n", "- This IS NOT expected if you are initializing DistilBertForMaskedLM from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).\n", "Some weights of the model checkpoint at /home/sipb/nlp-class-project/checkpoints/bert_base_ltr/epoch_3_checkpt were not used when initializing BertForMaskedLM: ['attention_mask']\n", "- This IS expected if you are initializing BertForMaskedLM from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).\n", "- This IS NOT expected if you are initializing BertForMaskedLM from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).\n", "Some weights of the model checkpoint at /home/sipb/nlp-class-project/checkpoints/bert_base_rtl/epoch_3_checkpt were not used when initializing BertForMaskedLM: ['attention_mask']\n", "- This IS expected if you are initializing BertForMaskedLM from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).\n", "- This IS NOT expected if you are initializing BertForMaskedLM from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).\n", "Some weights of the model checkpoint at /home/sipb/nlp-class-project/checkpoints/bert_large_ltr/epoch_3_checkpt were not used when initializing BertForMaskedLM: ['attention_mask']\n", "- This IS expected if you are initializing BertForMaskedLM from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).\n", "- This IS NOT expected if you are initializing BertForMaskedLM from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).\n", "Some weights of the model checkpoint at /home/sipb/nlp-class-project/checkpoints/bert_large_rtl/epoch_3_checkpt were not used when initializing BertForMaskedLM: ['attention_mask']\n", "- This IS expected if you are initializing BertForMaskedLM from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).\n", "- This IS NOT expected if you are initializing BertForMaskedLM from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).\n" ] } ], "source": [ "results = []\n", "# path_prefixes = [f\"bert_{size}\" for size in (6, 11, 19, 35)] + [\"distilbert_base\"]\n", "# for path_prefix in path_prefixes:\n", "# for text_dir in (\"ltr\", \"rtl\"):\n", "# checkpt_dir = f\"/home/sipb/nlp-class-project/checkpoints/{path_prefix}_{text_dir}_scratch/epoch_3_checkpt\"\n", "\n", "path_prefixes = [\"distilbert_base\", \"bert_base\", \"bert_large\"]\n", "for path_prefix in path_prefixes:\n", " for text_dir in (\"ltr\", \"rtl\"):\n", " checkpt_dir = f\"/home/sipb/nlp-class-project/checkpoints/{path_prefix}_{text_dir}/epoch_3_checkpt\"\n", " try:\n", " model = transformers.AutoModelForMaskedLM.from_pretrained(checkpt_dir)\n", " except:\n", " config = transformers.AutoConfig.from_pretrained(os.path.join(checkpt_dir, \"config.json\"))\n", " config.max_position_embeddings = 512\n", " model = transformers.AutoModelForMaskedLM.from_pretrained(checkpt_dir, config=config)\n", " \n", " add_attn_hooks(model, text_dir)\n", " model.eval();\n", "\n", " for dataset_type, dataset in [\n", " (\"qa\", riddles_qa),\n", " (\"aq\", riddles_aq),\n", " ]:\n", " ppls = []\n", " for riddle in dataset:\n", " batch = tokenizer([riddle], return_tensors=\"pt\", padding_side=\"left\" if text_dir == \"rtl\" else \"right\", padding=\"max_length\", max_length=128)\n", " batch[\"labels\"] = batch[\"input_ids\"].clone()\n", " batch[\"labels\"][batch[\"attention_mask\"] == 0] = -100\n", " with torch.inference_mode():\n", " output = model(**batch)\n", " ppls.append(math.e ** output.loss.item())\n", "\n", " results.append((sum(ppls) / len(ppls), dataset_type, text_dir, path_prefix))" ] }, { "cell_type": "code", "execution_count": 26, "execution_state": "idle", "id": "bdee66ad-65ad-40c7-ac86-9a2d6b8fba02", "metadata": {}, "outputs": [], "source": [ "to_params = {\n", " \"bert_6\": 6,\n", " \"bert_11\": 11,\n", " \"bert_19\": 19,\n", " \"bert_35\": 35,\n", " \"distilbert_base\": 67,\n", " \"bert_base\": 110,\n", " \"bert_large\": 335,\n", "}" ] }, { "cell_type": "code", "execution_count": 27, "execution_state": "idle", "id": "d1668465-fe85-4310-8d88-031d4b8d361f", "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "LTR & 6M & AQ & 1980 \\\\\n", "RTL & 6M & AQ & 1440 \\\\\n", "LTR & 11M & AQ & 1600 \\\\\n", "RTL & 11M & AQ & 1490 \\\\\n", "LTR & 19M & AQ & 2310 \\\\\n", "RTL & 19M & AQ & 2740 \\\\\n", "LTR & 35M & AQ & 3650 \\\\\n", "RTL & 35M & AQ & 4090 \\\\\n", "LTR & 67M & AQ & 6360 \\\\\n", "RTL & 67M & AQ & 22900 \\\\\n", "LTR & 6M & QA & 1790 \\\\\n", "RTL & 6M & QA & 1850 \\\\\n", "LTR & 11M & QA & 1430 \\\\\n", "RTL & 11M & QA & 1820 \\\\\n", "LTR & 19M & QA & 2280 \\\\\n", "RTL & 19M & QA & 3740 \\\\\n", "LTR & 35M & QA & 3690 \\\\\n", "RTL & 35M & QA & 4650 \\\\\n", "LTR & 67M & QA & 6340 \\\\\n", "RTL & 67M & QA & 22900 \\\\\n" ] } ], "source": [ "for ppl, task, text_dir, path_prefix in sorted(results, key=lambda x: (x[1], to_params[x[3]], x[2])):\n", " ppl = int(float(f\"{ppl:.3g}\"))\n", " print(rf\"{text_dir.upper()} & {to_params[path_prefix]}M & {task.upper()} & {ppl} \\\\\")" ] }, { "cell_type": "code", "execution_count": 29, "execution_state": "idle", "id": "e9ac0af8-2638-4076-b0cc-9ec9355c2c01", "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "LTR & 67M & AQ & 969 \\\\\n", "RTL & 67M & AQ & 675 \\\\\n", "LTR & 110M & AQ & 1880 \\\\\n", "RTL & 110M & AQ & 483 \\\\\n", "LTR & 335M & AQ & 2610 \\\\\n", "RTL & 335M & AQ & 695 \\\\\n", "LTR & 67M & QA & 952 \\\\\n", "RTL & 67M & QA & 781 \\\\\n", "LTR & 110M & QA & 2020 \\\\\n", "RTL & 110M & QA & 689 \\\\\n", "LTR & 335M & QA & 2950 \\\\\n", "RTL & 335M & QA & 782 \\\\\n" ] } ], "source": [ "for ppl, task, text_dir, path_prefix in sorted(results, key=lambda x: (x[1], to_params[x[3]], x[2])):\n", " ppl = int(float(f\"{ppl:.3g}\"))\n", " print(rf\"{text_dir.upper()} & {to_params[path_prefix]}M & {task.upper()} & {ppl} \\\\\")" ] }, { "cell_type": "code", "execution_count": 48, "execution_state": "idle", "id": "8894ca16-58e3-4448-bec8-c962f5135737", "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "the more you take, the more you leave behind. what am i? @ @ @ @ @ @ @ @ @ @ @ @ @ the @ @ @ @ ( the the the. @ the @ @ ( @ @ ( @ @ @ @ ( the.. @ ( @ ) @ the @ the the\n" ] } ], "source": [ "# input_text = [\"The more you take, the more you leave behind. What am I?\"]\n", "# batch = tokenizer(input_text, return_tensors=\"pt\", padding_side=\"right\", padding=\"max_length\", max_length=64)\n", "# output_ids = model.generate(batch['input_ids'], max_length=128, do_sample=False) # do_sample=False ensures greedy decoding\n", "# decoded_output = tokenizer.decode(output_ids[0], skip_special_tokens=True)\n", "# print(decoded_output)" ] }, { "cell_type": "code", "execution_count": 47, "execution_state": "idle", "id": "a4098975-2df6-4435-bc93-1a5afd6d7e68", "metadata": {}, "outputs": [], "source": [ "# CPU is fast enough\n", "\n", "ppls = []\n", "for riddle in riddles:\n", " batch = tokenizer([riddle], return_tensors=\"pt\", padding_side=\"left\" if text_dir == \"rtl\" else \"right\", padding=\"max_length\", max_length=128)\n", " batch[\"labels\"] = batch[\"input_ids\"].clone()\n", " batch[\"labels\"][batch[\"attention_mask\"] == 0] = -100\n", " # batch = tokenizer([riddle], return_tensors=\"pt\")#, padding_side=\"left\" if text_dir == \"rtl\" else \"right\", padding=\"longest\", max_length=128)\n", " # batch[\"labels\"] = batch[\"input_ids\"]\n", " with torch.inference_mode():\n", " output = model(**batch)\n", " ppls.append(math.e ** output.loss.item())" ] }, { "cell_type": "code", "execution_count": 16, "execution_state": "idle", "id": "c68b5235-a4a7-4f38-9acb-f5072e546a96", "metadata": {}, "outputs": [ { "data": { "text/plain": [ "(array([ 4., 6., 11., 6., 5., 2., 1., 1., 2., 2.]),\n", " array([ 613.56297843, 829.36555779, 1045.16813716, 1260.97071653,\n", " 1476.77329589, 1692.57587526, 1908.37845463, 2124.18103399,\n", " 2339.98361336, 2555.78619272, 2771.58877209]),\n", " )" ] }, "execution_count": 16, "metadata": {}, "output_type": "execute_result" }, { "data": { "image/png": "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", "text/plain": [ "
" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "import matplotlib.pyplot as plt\n", "\n", "plt.hist(ppls)" ] }, { "cell_type": "code", "execution_count": null, "id": "86daa05b-5784-457b-b65e-8b8395128d6f", "metadata": {}, "outputs": [], "source": [] } ], "metadata": { "kernelspec": { "display_name": "Python 3 (ipykernel)", "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": "3.12.7" } }, "nbformat": 4, "nbformat_minor": 5 }