Created
August 1, 2019 15:49
-
-
Save stuhlmueller/b3b9157f587f93b6e79c6b104b10a0d6 to your computer and use it in GitHub Desktop.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| { | |
| "cells": [ | |
| { | |
| "cell_type": "markdown", | |
| "metadata": {}, | |
| "source": [ | |
| "# Predicting How Performance Scales with Dataset Size on Natural Language Tasks\n", | |
| "\n", | |
| "## Summary\n", | |
| "This notebook presents experiments performed at Ought to test the relationship between dataset size and task performance. This was informed by previous work that saw a power law relationship between dataset size and error rate. We show that this relationship holds for tasks similar to those that would be involved in automating the Judge in Factored Evaluation. We think this relationship will allow us to test whether ideas will work on small datasets and predict whether they will work if scaled up.\n", | |
| "\n", | |
| "## Hypothesis\n", | |
| "- Power law relationship between loss/error rate and dataset size\n", | |
| " - For L (loss or error rate), n (dataset size):\n", | |
| " \n", | |
| " $L =\\exp(b) * n^m$ or equivalently $\\log(L) = m * \\log(n) + b$\n", | |
| "- We can predict performance on language modeling tasks using small datasets\n", | |
| "- Dataset design (underlying task, how clean the data is, what format it is in) are more important than small changes in architecture\n", | |
| "- This lets us iterate the design of small datasets until we have a version that we think will perform well when scaled up\n", | |
| "\n", | |
| "### How could this fail?\n", | |
| "- Predicted relationship could fail to hold for small or large dataset sizes\n", | |
| "- Different design choices in datasets don't lead to a significant difference in data scaling performance\n", | |
| "- Hyperparameter choices can change which datasets are easier to learn\n", | |
| "\n", | |
| "## Background\n", | |
| "### Why is this important for Ought?\n", | |
| "- We want to eventually train language models to take on the role of Experts and Judges in Factored Evaluation systems\n", | |
| "- These are generative tasks (eg. generating subquestions to ask to help make progress on a larger question), rather than classification tasks\n", | |
| "- We are comparatively more interested in dataset collection than the rest of the ML community (Factored Cognition/Factored Evaluation will require a lot of specialized human data to work, whereas other ML research can share datasets between many different projects)\n", | |
| "- We have limited data generation capacity, so we want to be able to predict whether a task can be performed before gathering large datasets\n", | |
| "\n", | |
| "\n", | |
| "### Previous Work\n", | |
| "[Hestness, J., Narang, S., Ardalani, N., Diamos, G., Jun, H., Kianinejad, H., ... & Zhou, Y. (2017). Deep learning scaling is predictable, empirically. arXiv preprint arXiv:1712.00409.](https://arxiv.org/abs/1712.00409)\n", | |
| "- Showed power-law scaling with dataset size in a variety of domains (image classification, language modelling, speech recognition), for both error and loss.\n", | |
| "- Used large dataset sizes and models, up to the limit of available compute\n", | |
| "\n", | |
| "### Differences from Previous Work\n", | |
| "We are interested in tasks that differ from previous work in the following ways:\n", | |
| "- Transformer\n", | |
| " - Hestness et al. used LSTMs as the Transformer had not yet been widely adopted\n", | |
| "- Small dataset sizes\n", | |
| " - The size of dataset that Ought can gather is smaller than most (though not all) experiments in Hestness et al.\n", | |
| "- GPT-2 finetuning (\n", | |
| " - We want to start from pretrained models, whereas Hestness et al. trained models from scratch\n", | |
| "- Generative tasks\n", | |
| " - We want to generate questions using a conditional language model, whereas Hestness et al. used classification and standard language modelling tasks\n", | |
| " \n", | |
| "## Process\n", | |
| "- Started with publicly available 117M parameter version of GPT-2 language model\n", | |
| "- Finetuned on available data for 1 epoch, learning rate 1e-5, Adam optimizer" | |
| ] | |
| }, | |
| { | |
| "cell_type": "code", | |
| "execution_count": 21, | |
| "metadata": {}, | |
| "outputs": [ | |
| { | |
| "name": "stdout", | |
| "output_type": "stream", | |
| "text": [ | |
| "The autoreload extension is already loaded. To reload it, use:\n", | |
| " %reload_ext autoreload\n" | |
| ] | |
| } | |
| ], | |
| "source": [ | |
| "%load_ext autoreload\n", | |
| "%autoreload 2\n", | |
| "from data_scaling import *\n", | |
| "%matplotlib inline" | |
| ] | |
| }, | |
| { | |
| "cell_type": "markdown", | |
| "metadata": {}, | |
| "source": [ | |
| "## Complex Web Questions\n", | |
| "\n", | |
| "- Complex Web Questions tasks are questions that can be answered by using Google to find the answers to 2 simpler questions\n", | |
| "- Decomposition is the task of taking an original question, and outputing the two simpler questions to search with Google" | |
| ] | |
| }, | |
| { | |
| "cell_type": "markdown", | |
| "metadata": {}, | |
| "source": [ | |
| "### SPARQL Decomposition\n", | |
| "- Decomposition using templates for original SPARQL queries\n", | |
| "- Usually make sense, but worded differently than original question\n", | |
| "- When fully trained, the ML algorithm sometimes still makes a non sequitur (ie. substituting the name of the wrong person)\n", | |
| "\n", | |
| "```\n", | |
| "Examples\n", | |
| "Question: When did the champion of the 1999 World Series win their first World Series?\n", | |
| "Decomposition: 1. the team won the 1999 World Series championship? 2. what year did %composition win their first World Series?\n", | |
| "\n", | |
| "Question: Who is the current president of where the tv show named The Bride with White Hair was filmed in 2010?\n", | |
| "Decomposition: 1. the tv show The Bride with White Hair was filmed there? 2. who is the current president of %composition 2010?\n", | |
| "```" | |
| ] | |
| }, | |
| { | |
| "cell_type": "code", | |
| "execution_count": 22, | |
| "metadata": {}, | |
| "outputs": [ | |
| { | |
| "name": "stdout", | |
| "output_type": "stream", | |
| "text": [ | |
| "Validation Loss = 3.90*n^-0.23\n", | |
| "Halving Validation Loss requires ~ 10^1.3 times more data\n" | |
| ] | |
| }, | |
| { | |
| "data": { | |
| "text/html": [ | |
| "<div>\n", | |
| "<style scoped>\n", | |
| " .dataframe tbody tr th:only-of-type {\n", | |
| " vertical-align: middle;\n", | |
| " }\n", | |
| "\n", | |
| " .dataframe tbody tr th {\n", | |
| " vertical-align: top;\n", | |
| " }\n", | |
| "\n", | |
| " .dataframe thead th {\n", | |
| " text-align: right;\n", | |
| " }\n", | |
| "</style>\n", | |
| "<table border=\"1\" class=\"dataframe\">\n", | |
| " <thead>\n", | |
| " <tr style=\"text-align: right;\">\n", | |
| " <th></th>\n", | |
| " <th>Target Validation Loss</th>\n", | |
| " <th>Log10 (Dataset Size)</th>\n", | |
| " </tr>\n", | |
| " </thead>\n", | |
| " <tbody>\n", | |
| " <tr>\n", | |
| " <th>0</th>\n", | |
| " <td>1.0</td>\n", | |
| " <td>2.6</td>\n", | |
| " </tr>\n", | |
| " <tr>\n", | |
| " <th>1</th>\n", | |
| " <td>0.5</td>\n", | |
| " <td>3.9</td>\n", | |
| " </tr>\n", | |
| " <tr>\n", | |
| " <th>2</th>\n", | |
| " <td>0.2</td>\n", | |
| " <td>5.6</td>\n", | |
| " </tr>\n", | |
| " <tr>\n", | |
| " <th>3</th>\n", | |
| " <td>0.1</td>\n", | |
| " <td>7.0</td>\n", | |
| " </tr>\n", | |
| " </tbody>\n", | |
| "</table>\n", | |
| "</div>" | |
| ], | |
| "text/plain": [ | |
| " Target Validation Loss Log10 (Dataset Size)\n", | |
| "0 1.0 2.6\n", | |
| "1 0.5 3.9\n", | |
| "2 0.2 5.6\n", | |
| "3 0.1 7.0" | |
| ] | |
| }, | |
| "metadata": {}, | |
| "output_type": "display_data" | |
| }, | |
| { | |
| "data": { | |
| "image/png": "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\n", | |
| "text/plain": [ | |
| "<Figure size 1152x432 with 2 Axes>" | |
| ] | |
| }, | |
| "metadata": { | |
| "needs_background": "light" | |
| }, | |
| "output_type": "display_data" | |
| } | |
| ], | |
| "source": [ | |
| "model_plots(\"local/cwq/decomposition_machine/best\")" | |
| ] | |
| }, | |
| { | |
| "cell_type": "markdown", | |
| "metadata": {}, | |
| "source": [ | |
| "### Heuristic Decompositions\n", | |
| "\n", | |
| "- Decomposition using heuristics that build the decomposition from sections of the original question\n", | |
| "- More closely resemble original question, but don't always make sense\n", | |
| "\n", | |
| "Examples\n", | |
| "```\n", | |
| "Question: When did the champion of the 1999 World Series win their first World Series?\n", | |
| "Decomposition: 1. the 1999 World Series? 2. When did the champion of %composition win their first World Series?\n", | |
| "\n", | |
| "Question: Who is the current president of where the tv show named The Bride with White Hair was filmed in 2010?\n", | |
| "Decomposition: 1. the tv show named The Bride with White Hair was filmed in? 2. Who is the current president of where %composition 2010?\n", | |
| "```" | |
| ] | |
| }, | |
| { | |
| "cell_type": "code", | |
| "execution_count": 23, | |
| "metadata": {}, | |
| "outputs": [ | |
| { | |
| "name": "stdout", | |
| "output_type": "stream", | |
| "text": [ | |
| "Validation Loss = 2.85*n^-0.38\n", | |
| "Halving Validation Loss requires ~ 10^0.8 times more data\n" | |
| ] | |
| }, | |
| { | |
| "data": { | |
| "text/html": [ | |
| "<div>\n", | |
| "<style scoped>\n", | |
| " .dataframe tbody tr th:only-of-type {\n", | |
| " vertical-align: middle;\n", | |
| " }\n", | |
| "\n", | |
| " .dataframe tbody tr th {\n", | |
| " vertical-align: top;\n", | |
| " }\n", | |
| "\n", | |
| " .dataframe thead th {\n", | |
| " text-align: right;\n", | |
| " }\n", | |
| "</style>\n", | |
| "<table border=\"1\" class=\"dataframe\">\n", | |
| " <thead>\n", | |
| " <tr style=\"text-align: right;\">\n", | |
| " <th></th>\n", | |
| " <th>Target Validation Loss</th>\n", | |
| " <th>Log10 (Dataset Size)</th>\n", | |
| " </tr>\n", | |
| " </thead>\n", | |
| " <tbody>\n", | |
| " <tr>\n", | |
| " <th>0</th>\n", | |
| " <td>1.0</td>\n", | |
| " <td>1.2</td>\n", | |
| " </tr>\n", | |
| " <tr>\n", | |
| " <th>1</th>\n", | |
| " <td>0.5</td>\n", | |
| " <td>2.0</td>\n", | |
| " </tr>\n", | |
| " <tr>\n", | |
| " <th>2</th>\n", | |
| " <td>0.2</td>\n", | |
| " <td>3.1</td>\n", | |
| " </tr>\n", | |
| " <tr>\n", | |
| " <th>3</th>\n", | |
| " <td>0.1</td>\n", | |
| " <td>3.9</td>\n", | |
| " </tr>\n", | |
| " </tbody>\n", | |
| "</table>\n", | |
| "</div>" | |
| ], | |
| "text/plain": [ | |
| " Target Validation Loss Log10 (Dataset Size)\n", | |
| "0 1.0 1.2\n", | |
| "1 0.5 2.0\n", | |
| "2 0.2 3.1\n", | |
| "3 0.1 3.9" | |
| ] | |
| }, | |
| "metadata": {}, | |
| "output_type": "display_data" | |
| }, | |
| { | |
| "data": { | |
| "image/png": "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\n", | |
| "text/plain": [ | |
| "<Figure size 1152x432 with 2 Axes>" | |
| ] | |
| }, | |
| "metadata": { | |
| "needs_background": "light" | |
| }, | |
| "output_type": "display_data" | |
| } | |
| ], | |
| "source": [ | |
| "model_plots(\"cwq/decomposition/latest\")" | |
| ] | |
| }, | |
| { | |
| "cell_type": "markdown", | |
| "metadata": {}, | |
| "source": [ | |
| "## MAWPS\n", | |
| "\n", | |
| "- Math word problems, filtered for simpler ones, seem to be pretty formulaic/model seems to pick up on cues in how things are worded\n", | |
| "\n", | |
| "Examples\n", | |
| "```\n", | |
| "Question: April's discount flowers was having a sale where each rose was 9 dollars. If April started with 11 roses and had 8 roses left, how much money did she earn?\n", | |
| "Output: X=(9.0*(11.0-8.0))\n", | |
| "\n", | |
| "Question: A pet store had 56 puppies. In one day they sold 24 of them and put the rest into cages with 4 in each cage. How many cages did they use? ?\n", | |
| "X=((56.0-24.0)/4.0)\n", | |
| "\n", | |
| "Question: There are 6 candies in a pile on the desk. Each candy comes in a package of 15. 4 candies are added to the pile. How many candies are there in the pile??\n", | |
| "X=(6.0+4.0)\n", | |
| "```" | |
| ] | |
| }, | |
| { | |
| "cell_type": "code", | |
| "execution_count": 24, | |
| "metadata": {}, | |
| "outputs": [ | |
| { | |
| "name": "stdout", | |
| "output_type": "stream", | |
| "text": [ | |
| "Validation Loss = 4.59*n^-0.44\n", | |
| "Halving Validation Loss requires ~ 10^0.7 times more data\n" | |
| ] | |
| }, | |
| { | |
| "data": { | |
| "text/html": [ | |
| "<div>\n", | |
| "<style scoped>\n", | |
| " .dataframe tbody tr th:only-of-type {\n", | |
| " vertical-align: middle;\n", | |
| " }\n", | |
| "\n", | |
| " .dataframe tbody tr th {\n", | |
| " vertical-align: top;\n", | |
| " }\n", | |
| "\n", | |
| " .dataframe thead th {\n", | |
| " text-align: right;\n", | |
| " }\n", | |
| "</style>\n", | |
| "<table border=\"1\" class=\"dataframe\">\n", | |
| " <thead>\n", | |
| " <tr style=\"text-align: right;\">\n", | |
| " <th></th>\n", | |
| " <th>Target Validation Loss</th>\n", | |
| " <th>Log10 (Dataset Size)</th>\n", | |
| " </tr>\n", | |
| " </thead>\n", | |
| " <tbody>\n", | |
| " <tr>\n", | |
| " <th>0</th>\n", | |
| " <td>1.0</td>\n", | |
| " <td>1.5</td>\n", | |
| " </tr>\n", | |
| " <tr>\n", | |
| " <th>1</th>\n", | |
| " <td>0.5</td>\n", | |
| " <td>2.2</td>\n", | |
| " </tr>\n", | |
| " <tr>\n", | |
| " <th>2</th>\n", | |
| " <td>0.2</td>\n", | |
| " <td>3.1</td>\n", | |
| " </tr>\n", | |
| " <tr>\n", | |
| " <th>3</th>\n", | |
| " <td>0.1</td>\n", | |
| " <td>3.7</td>\n", | |
| " </tr>\n", | |
| " </tbody>\n", | |
| "</table>\n", | |
| "</div>" | |
| ], | |
| "text/plain": [ | |
| " Target Validation Loss Log10 (Dataset Size)\n", | |
| "0 1.0 1.5\n", | |
| "1 0.5 2.2\n", | |
| "2 0.2 3.1\n", | |
| "3 0.1 3.7" | |
| ] | |
| }, | |
| "metadata": {}, | |
| "output_type": "display_data" | |
| }, | |
| { | |
| "data": { | |
| "image/png": "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\n", | |
| "text/plain": [ | |
| "<Figure size 1152x432 with 2 Axes>" | |
| ] | |
| }, | |
| "metadata": { | |
| "needs_background": "light" | |
| }, | |
| "output_type": "display_data" | |
| } | |
| ], | |
| "source": [ | |
| "model_plots(\"mawps/latest\")" | |
| ] | |
| }, | |
| { | |
| "cell_type": "markdown", | |
| "metadata": { | |
| "toc-hr-collapsed": false | |
| }, | |
| "source": [ | |
| "## Quora Similar Questions\n", | |
| "\n", | |
| "- Given quora question, generate a question with similar meaning\n", | |
| "- For each pair, two examples are created (given A output B, given B output A)\n", | |
| "\n", | |
| "Examples\n", | |
| "```\n", | |
| "Question: How is it to be audience in a Kapil Sharma show?\n", | |
| "Rewritten: What is it like watching 'The Kapil Sharma Show' live in the studio?\n", | |
| "Question: Nobody is answering my question in Quora. Does it mean my questions are stupid or people around me are stupid?\n", | |
| "How come nobody is answering my questions in Quora?\n", | |
| "```" | |
| ] | |
| }, | |
| { | |
| "cell_type": "code", | |
| "execution_count": 25, | |
| "metadata": {}, | |
| "outputs": [ | |
| { | |
| "name": "stdout", | |
| "output_type": "stream", | |
| "text": [ | |
| "Validation Loss = 2.06*n^-0.04\n", | |
| "Halving Validation Loss requires ~ 10^6.8 times more data\n" | |
| ] | |
| }, | |
| { | |
| "data": { | |
| "text/html": [ | |
| "<div>\n", | |
| "<style scoped>\n", | |
| " .dataframe tbody tr th:only-of-type {\n", | |
| " vertical-align: middle;\n", | |
| " }\n", | |
| "\n", | |
| " .dataframe tbody tr th {\n", | |
| " vertical-align: top;\n", | |
| " }\n", | |
| "\n", | |
| " .dataframe thead th {\n", | |
| " text-align: right;\n", | |
| " }\n", | |
| "</style>\n", | |
| "<table border=\"1\" class=\"dataframe\">\n", | |
| " <thead>\n", | |
| " <tr style=\"text-align: right;\">\n", | |
| " <th></th>\n", | |
| " <th>Target Validation Loss</th>\n", | |
| " <th>Log10 (Dataset Size)</th>\n", | |
| " </tr>\n", | |
| " </thead>\n", | |
| " <tbody>\n", | |
| " <tr>\n", | |
| " <th>0</th>\n", | |
| " <td>1.0</td>\n", | |
| " <td>7.1</td>\n", | |
| " </tr>\n", | |
| " <tr>\n", | |
| " <th>1</th>\n", | |
| " <td>0.5</td>\n", | |
| " <td>13.9</td>\n", | |
| " </tr>\n", | |
| " <tr>\n", | |
| " <th>2</th>\n", | |
| " <td>0.2</td>\n", | |
| " <td>22.9</td>\n", | |
| " </tr>\n", | |
| " <tr>\n", | |
| " <th>3</th>\n", | |
| " <td>0.1</td>\n", | |
| " <td>29.7</td>\n", | |
| " </tr>\n", | |
| " </tbody>\n", | |
| "</table>\n", | |
| "</div>" | |
| ], | |
| "text/plain": [ | |
| " Target Validation Loss Log10 (Dataset Size)\n", | |
| "0 1.0 7.1\n", | |
| "1 0.5 13.9\n", | |
| "2 0.2 22.9\n", | |
| "3 0.1 29.7" | |
| ] | |
| }, | |
| "metadata": {}, | |
| "output_type": "display_data" | |
| }, | |
| { | |
| "data": { | |
| "image/png": "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\n", | |
| "text/plain": [ | |
| "<Figure size 1152x432 with 2 Axes>" | |
| ] | |
| }, | |
| "metadata": { | |
| "needs_background": "light" | |
| }, | |
| "output_type": "display_data" | |
| } | |
| ], | |
| "source": [ | |
| "model_plots(\"quora/1\")" | |
| ] | |
| }, | |
| { | |
| "cell_type": "code", | |
| "execution_count": 26, | |
| "metadata": {}, | |
| "outputs": [ | |
| { | |
| "name": "stdout", | |
| "output_type": "stream", | |
| "text": [ | |
| "quora_smaller/latest []\n" | |
| ] | |
| } | |
| ], | |
| "source": [ | |
| "model_plots(\"quora_smaller/latest\")" | |
| ] | |
| }, | |
| { | |
| "cell_type": "markdown", | |
| "metadata": {}, | |
| "source": [ | |
| "## Conclusion\n", | |
| "- Scaling relationship in Hestness et al. holds for tasks similar to those of interest for Ought\n", | |
| "- Changes to the dataset construction (for Complex Web Questions) change the power law exponent of the data scaling curve\n", | |
| "\n", | |
| "## Further Work\n", | |
| "- Improve metrics\n", | |
| "- Test models/datasets in the original paper to see if relationship still holds at small dataset size" | |
| ] | |
| } | |
| ], | |
| "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": "3.7.3" | |
| } | |
| }, | |
| "nbformat": 4, | |
| "nbformat_minor": 2 | |
| } |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment