{ "cells": [ { "cell_type": "code", "execution_count": 1, "id": "33df4373-a37b-4fd0-bc67-c297812871e4", "metadata": {}, "outputs": [], "source": [ "# Use the trained astroBERT model with the fill-mask pipeline" ] }, { "cell_type": "markdown", "id": "164ee9bd-27f9-40a4-8461-3ce12fc928b0", "metadata": {}, "source": [ "# Tutorial 1: using astroBERT with the fill-mask pipeline" ] }, { "cell_type": "code", "execution_count": 2, "id": "59429414-f07e-45e5-8825-6fc6a8d26653", "metadata": {}, "outputs": [], "source": [ "# 1 - load models and tokenizer" ] }, { "cell_type": "code", "execution_count": 3, "id": "db8ee724-6a2a-4ea5-820e-5e2aa0a0f622", "metadata": {}, "outputs": [ { "name": "stderr", "output_type": "stream", "text": [ "2022-10-31 11:29:29.185816: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcudart.so.11.0\n" ] } ], "source": [ "from transformers import AutoTokenizer, BertForMaskedLM" ] }, { "cell_type": "code", "execution_count": 4, "id": "9a98fb63-0793-4684-a202-931cad17c7ca", "metadata": {}, "outputs": [], "source": [ "# the model path can either be the name of the Huggingface repository\n", "remote_model_path = 'adsabs/astroBERT'\n", "# or the local path to the directory containing model weight and tokenizer vocab\n", "local_model_path = '../'" ] }, { "cell_type": "code", "execution_count": 5, "id": "25fedd16-283b-4817-9b19-2a5ff1c5ba88", "metadata": {}, "outputs": [], "source": [ "# make sure you load the tokenier with do_lower_case=False\n", "astroBERT_tokenizer = AutoTokenizer.from_pretrained(pretrained_model_name_or_path=remote_model_path,\n", " add_special_tokens=False,\n", " do_lower_case=False,\n", " )" ] }, { "cell_type": "code", "execution_count": 6, "id": "fb10db03-a5f0-44f7-8d41-0285f898a90d", "metadata": {}, "outputs": [ { "name": "stderr", "output_type": "stream", "text": [ "Some weights of the model checkpoint at adsabs/astroBERT were not used when initializing BertForMaskedLM: ['cls.seq_relationship.bias', 'cls.seq_relationship.weight']\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": [ "astroBERT_automodel_for_mlm = BertForMaskedLM.from_pretrained(pretrained_model_name_or_path=remote_model_path,\n", " )" ] }, { "cell_type": "code", "execution_count": 7, "id": "e8b9b073-3876-4d0b-b8b2-e46fa25c76f0", "metadata": {}, "outputs": [], "source": [ "# for pipeline to work you have to ensure that the model returns a dict\n", "astroBERT_automodel_for_mlm.config.return_dict=True" ] }, { "cell_type": "code", "execution_count": 8, "id": "94338f6f-3467-4696-bf7d-f41a12eb889d", "metadata": {}, "outputs": [], "source": [ "from transformers import FillMaskPipeline" ] }, { "cell_type": "code", "execution_count": 9, "id": "7b980d9f-4d86-4b54-9324-d57dd9b4b64f", "metadata": {}, "outputs": [], "source": [ "astroBERT_pipeline = FillMaskPipeline(model=astroBERT_automodel_for_mlm,\n", " tokenizer=astroBERT_tokenizer,\n", " task='fill-mask',\n", " )" ] }, { "cell_type": "code", "execution_count": 10, "id": "5cb4d27b-ee3c-4ac7-ace2-4cc57ea9ce7a", "metadata": {}, "outputs": [], "source": [ "clean_sentences = ['M67 is one of the most studied open clusters .',\n", "'A solar twin is a star with atmospheric parameters and chemical composition very similar to our Sun .',\n", "'The dynamical evolution of planets close to their star is affected by tidal effects .',\n", "'The Kepler satellite collected high-precision long-term and continuous light curves for more than 100,000 solar-type stars .',\n", "'The Local Group is composed of the Milky Way, the Andromeda Galaxy, and numerous smaller satellite galaxies .',\n", "'Cepheid variables are used to determine the distances to galaxies in the local universe .',\n", "'Jets are created and sustained by accretion of matter onto a compact massive object .',\n", "'A single star of one solar mass will evolve into a white dwarf .',\n", "'The Very Large Array observes the sky at radio wavelengths .',\n", "'Elements heavier than iron are generated in supernovae explosions .',\n", "'Spitzer was the first spacecraft to fly in an Earth-trailing orbit .',\n", "'Galaxy mergers can occur when two (or more) galaxies collide .',\n", "'Dark matter is a hypothetical form of matter thought to account for approximately 85% of the matter in the universe .',\n", "'The Local Group of galaxies is pulled toward The Great Attractor .',\n", "'The Moon is the only satellite of the Earth .',\n", "'Galaxies are categorized according to their visual morphology as elliptical, spiral, or irregular .',\n", "'Stars are made mostly of hydrogen .',\n", "'Comet tails are created as comets approach the Sun .',\n", "'Pluto is a dwarf planet in the Kuiper Belt .',\n", "'The Milky Way has a supermassive black hole, Sagittarius A*, at its center .',\n", "'Andromeda is the nearest large galaxy to the Milky Way and is roughly its equal in mass .',\n", "'The interstellar medium is the gas and dust between stars .',\n", "'The cosmic microwave background (CMB, CMBR), in Big Bang cosmology, is electromagnetic radiation which is a remnant from an early stage of the universe .',\n", "'The Large and Small Magellanic Clouds are irregular dwarf galaxies and are two satellite galaxies of the Milky Way .']" ] }, { "cell_type": "code", "execution_count": 11, "id": "9f3a6fdc-182f-4edb-8ef4-7e4253c2d4db", "metadata": {}, "outputs": [], "source": [ "masked_sentences = ['M67 is one of the most studied [MASK] clusters .',\n", "'A solar twin is a star with [MASK] parameters and chemical composition very similar to our Sun .',\n", "'The dynamical evolution of planets close to their star is affected by [MASK] effects .',\n", "'The Kepler satellite collected high-precision long-term and continuous light [MASK] for more than 100,000 solar-type stars .',\n", "'The Local Group is composed of the Milky Way, the [MASK] Galaxy, and numerous smaller satellite galaxies .',\n", "'Cepheid variables are used to determine the [MASK] to galaxies in the local universe .',\n", "'Jets are created and sustained by [MASK] of matter onto a compact massive object .',\n", "'A single star of one solar mass will evolve into a [MASK] dwarf .',\n", "'The Very Large Array observes the sky at [MASK] wavelengths .',\n", "'Elements heavier than [MASK] are generated in supernovae explosions .',\n", "'Spitzer was the first [MASK] to fly in an Earth-trailing orbit .',\n", "'Galaxy [MASK] can occur when two (or more) galaxies collide .',\n", "'Dark [MASK] is a hypothetical form of matter thought to account for approximately 85% of the matter in the universe .',\n", "'The Local Group of galaxies is pulled toward The Great [MASK] .',\n", "'The Moon is the only [MASK] of the Earth .',\n", "'Galaxies are categorized according to their visual morphology as [MASK] , spiral, or irregular .',\n", "'Stars are made mostly of [MASK] .',\n", "'Comet tails are created as comets approach the [MASK] .',\n", "'Pluto is a dwarf [MASK] in the Kuiper Belt .',\n", "'The Milky Way has a [MASK] black hole, Sagittarius A*, at its center .',\n", "'Andromeda is the nearest large [MASK] to the Milky Way and is roughly its equal in mass .',\n", "'The [MASK] medium is the gas and dust between stars .',\n", "'The cosmic microwave background (CMB, CMBR), in Big Bang cosmology, is electromagnetic radiation which is a remnant from an early stage of the [MASK] .',\n", "'The Large and Small Magellanic Clouds are irregular [MASK] galaxies and are two satellite galaxies of the Milky Way .',\n", "]" ] }, { "cell_type": "code", "execution_count": 12, "id": "d4c729ad-89f4-4e70-b433-a65b6035c10b", "metadata": {}, "outputs": [], "source": [ "masked_words = [x for s1,s2 in zip(clean_sentences, masked_sentences) \n", " for x,y in zip(s1.split(), s2.split()) if y=='[MASK]']" ] }, { "cell_type": "code", "execution_count": 13, "id": "2a07a641-61a7-42dd-b70e-62eb97ad4e4b", "metadata": {}, "outputs": [], "source": [ "results = astroBERT_pipeline(inputs=masked_sentences, \n", " top_k=3\n", " )" ] }, { "cell_type": "code", "execution_count": 14, "id": "ec2880d9-a8ad-4919-ab5b-732f3bcc21ae", "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "M67 is one of the most studied [MASK] clusters .\n", "original: open\n", "\t open 0.87\n", "\t globular 0.07\n", "\t star 0.03\n", "\n", "A solar twin is a star with [MASK] parameters and chemical composition very similar to our Sun .\n", "original: atmospheric\n", "\t fundamental 0.56\n", "\t physical 0.25\n", "\t stellar 0.05\n", "\n", "The dynamical evolution of planets close to their star is affected by [MASK] effects .\n", "original: tidal\n", "\t tidal 0.38\n", "\t secular 0.09\n", "\t relativistic 0.05\n", "\n", "The Kepler satellite collected high-precision long-term and continuous light [MASK] for more than 100,000 solar-type stars .\n", "original: curves\n", "\t curves 0.99\n", "\t curve 0.00\n", "\t variations 0.00\n", "\n", "The Local Group is composed of the Milky Way, the [MASK] Galaxy, and numerous smaller satellite galaxies .\n", "original: Andromeda\n", "\t Andromeda 0.99\n", "\t M31 0.00\n", "\t Sagittarius 0.00\n", "\n", "Cepheid variables are used to determine the [MASK] to galaxies in the local universe .\n", "original: distances\n", "\t distances 0.79\n", "\t distance 0.21\n", "\t redshifts 0.00\n", "\n", "Jets are created and sustained by [MASK] of matter onto a compact massive object .\n", "original: accretion\n", "\t accretion 0.79\n", "\t infall 0.13\n", "\t fall 0.02\n", "\n", "A single star of one solar mass will evolve into a [MASK] dwarf .\n", "original: white\n", "\t white 0.77\n", "\t brown 0.19\n", "\t red 0.02\n", "\n", "The Very Large Array observes the sky at [MASK] wavelengths .\n", "original: radio\n", "\t radio 0.29\n", "\t centimeter 0.10\n", "\t all 0.09\n", "\n", "Elements heavier than [MASK] are generated in supernovae explosions .\n", "original: iron\n", "\t iron 0.34\n", "\t helium 0.16\n", "\t oxygen 0.07\n", "\n", "Spitzer was the first [MASK] to fly in an Earth-trailing orbit .\n", "original: spacecraft\n", "\t satellite 0.42\n", "\t spacecraft 0.20\n", "\t observatory 0.16\n", "\n", "Galaxy [MASK] can occur when two (or more) galaxies collide .\n", "original: mergers\n", "\t collisions 0.25\n", "\t collision 0.22\n", "\t merging 0.21\n", "\n", "Dark [MASK] is a hypothetical form of matter thought to account for approximately 85% of the matter in the universe .\n", "original: matter\n", "\t energy 0.64\n", "\t Energy 0.24\n", "\t matter 0.10\n", "\n", "The Local Group of galaxies is pulled toward The Great [MASK] .\n", "original: Attractor\n", "\t Wall 0.96\n", "\t East 0.01\n", "\t Planet 0.00\n", "\n", "The Moon is the only [MASK] of the Earth .\n", "original: satellite\n", "\t satellite 0.38\n", "\t moon 0.31\n", "\t constituent 0.07\n", "\n", "Galaxies are categorized according to their visual morphology as [MASK] , spiral, or irregular .\n", "original: elliptical,\n", "\t elliptical 0.92\n", "\t spheroidal 0.02\n", "\t irregular 0.01\n", "\n", "Stars are made mostly of [MASK] .\n", "original: hydrogen\n", "\t hydrogen 0.20\n", "\t helium 0.14\n", "\t carbon 0.12\n", "\n", "Comet tails are created as comets approach the [MASK] .\n", "original: Sun\n", "\t Sun 0.45\n", "\t sun 0.23\n", "\t Earth 0.19\n", "\n", "Pluto is a dwarf [MASK] in the Kuiper Belt .\n", "original: planet\n", "\t planet 0.96\n", "\t satellite 0.02\n", "\t nova 0.00\n", "\n", "The Milky Way has a [MASK] black hole, Sagittarius A*, at its center .\n", "original: supermassive\n", "\t supermassive 0.80\n", "\t massive 0.17\n", "\t stellar 0.00\n", "\n", "Andromeda is the nearest large [MASK] to the Milky Way and is roughly its equal in mass .\n", "original: galaxy\n", "\t galaxy 0.68\n", "\t spiral 0.12\n", "\t satellite 0.09\n", "\n", "The [MASK] medium is the gas and dust between stars .\n", "original: interstellar\n", "\t interstellar 0.87\n", "\t interplanetary 0.05\n", "\t intracluster 0.03\n", "\n", "The cosmic microwave background (CMB, CMBR), in Big Bang cosmology, is electromagnetic radiation which is a remnant from an early stage of the [MASK] .\n", "original: universe\n", "\t universe 0.45\n", "\t Universe 0.26\n", "\t expansion 0.09\n", "\n", "The Large and Small Magellanic Clouds are irregular [MASK] galaxies and are two satellite galaxies of the Milky Way .\n", "original: dwarf\n", "\t dwarf 0.68\n", "\t satellite 0.13\n", "\t Magellanic 0.08\n", "\n" ] } ], "source": [ "for w, s, rs in zip(masked_words, masked_sentences, results):\n", " print(s)\n", " print('original: {}'.format(w))\n", " for r in rs:\n", " print('\\t {} {:0.2f}'.format(r['token_str'], r['score']))\n", " print()" ] } ], "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": "3.8.5" } }, "nbformat": 4, "nbformat_minor": 5 }