examples/notebooks/HyenaDNA-Fine-Tuning.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Fine-Tune HyenaDNA On A Downstream Classification Task"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "In this notebook, we will fine-tune the HynaDNA model on a downstream binary classification task involving cis-regulatory elements.\n",
    "\n",
    "[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/helicalAI/helical/blob/main/examples/notebooks/HyenaDNA-Fine-Tuning.ipynb)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Install the package"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# !pip install helical"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Imports"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from helical.models.hyena_dna import HyenaDNA, HyenaDNAConfig, HyenaDNAFineTuningModel\n",
    "from datasets import load_dataset\n",
    "from sklearn.metrics import classification_report,confusion_matrix,ConfusionMatrixDisplay\n",
    "import matplotlib.pyplot as plt\n",
    "import torch\n",
    "import logging, warnings\n",
    "import numpy as np\n",
    "\n",
    "logging.getLogger().setLevel(logging.ERROR)\n",
    "\n",
    "warnings.filterwarnings(\"ignore\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Make use of GPU if there is a GPU present, otherwise default to the CPU"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "device = \"cuda\" if torch.cuda.is_available() else \"cpu\""
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Dataset Download"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- We now use the Hugging Face API to download the [dataset](https://huggingface.co/datasets/InstaDeepAI/nucleotide_transformer_downstream_tasks_revised)\n",
    "- We select the task from the list of tasks available for this dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "label = \"promoter_tata\"\n",
    "dataset_train = load_dataset(\"InstaDeepAI/nucleotide_transformer_downstream_tasks_revised\", label, split=\"train\", trust_remote_code=True)\n",
    "dataset_test = load_dataset(\"InstaDeepAI/nucleotide_transformer_downstream_tasks_revised\", label, split=\"test\", trust_remote_code=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{0: 2531, 1: 2531}"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "unique, counts = np.unique(dataset_train[\"label\"], return_counts=True)\n",
    "dict(zip(unique, counts))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{0: 106, 1: 106}"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "unique, counts = np.unique(dataset_test[\"label\"], return_counts=True)\n",
    "dict(zip(unique, counts))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Since this is a binary classification task, only 2 labels are present below"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Define our HyenaDNA fine-tuning model and configs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "hyena_config = HyenaDNAConfig(model_name=\"hyenadna-tiny-1k-seqlen-d256\", batch_size=10, device=device)\n",
    "hyena_fine_tune = HyenaDNAFineTuningModel(hyena_config, fine_tuning_head=\"classification\", output_size=len(np.unique(dataset_train[\"label\"])))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Process the dataset for both training and testing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Processing sequences: 100%|██████████| 5062/5062 [00:00<00:00, 10775.16it/s]\n",
      "Processing sequences: 100%|██████████| 212/212 [00:00<00:00, 9641.03it/s]\n"
     ]
    }
   ],
   "source": [
    "train_dataset = hyena_fine_tune.process_data(dataset_train[\"sequence\"])\n",
    "test_dataset = hyena_fine_tune.process_data(dataset_test[\"sequence\"])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Fine-tune the model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "For this we set our training and validation datasets along with their labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Fine-Tuning: epoch 1/10: 100%|██████████| 507/507 [00:02<00:00, 191.25it/s, loss=0.654]\n",
      "Fine-Tuning Validation: 100%|██████████| 22/22 [00:00<00:00, 729.64it/s, val_loss=0.602]\n",
      "Fine-Tuning: epoch 2/10: 100%|██████████| 507/507 [00:02<00:00, 213.34it/s, loss=0.441]\n",
      "Fine-Tuning Validation: 100%|██████████| 22/22 [00:00<00:00, 717.45it/s, val_loss=0.796]\n",
      "Fine-Tuning: epoch 3/10: 100%|██████████| 507/507 [00:02<00:00, 211.52it/s, loss=0.453]\n",
      "Fine-Tuning Validation: 100%|██████████| 22/22 [00:00<00:00, 716.82it/s, val_loss=0.597]\n",
      "Fine-Tuning: epoch 4/10: 100%|██████████| 507/507 [00:02<00:00, 206.29it/s, loss=0.444]\n",
      "Fine-Tuning Validation: 100%|██████████| 22/22 [00:00<00:00, 730.15it/s, val_loss=0.492]\n",
      "Fine-Tuning: epoch 5/10: 100%|██████████| 507/507 [00:02<00:00, 232.99it/s, loss=0.439]\n",
      "Fine-Tuning Validation: 100%|██████████| 22/22 [00:00<00:00, 724.66it/s, val_loss=0.424]\n",
      "Fine-Tuning: epoch 6/10: 100%|██████████| 507/507 [00:01<00:00, 256.32it/s, loss=0.438]\n",
      "Fine-Tuning Validation: 100%|██████████| 22/22 [00:00<00:00, 724.62it/s, val_loss=0.382]\n",
      "Fine-Tuning: epoch 7/10: 100%|██████████| 507/507 [00:01<00:00, 258.73it/s, loss=0.436]\n",
      "Fine-Tuning Validation: 100%|██████████| 22/22 [00:00<00:00, 715.60it/s, val_loss=0.351]\n",
      "Fine-Tuning: epoch 8/10: 100%|██████████| 507/507 [00:02<00:00, 233.66it/s, loss=0.434]\n",
      "Fine-Tuning Validation: 100%|██████████| 22/22 [00:00<00:00, 709.78it/s, val_loss=0.336]\n",
      "Fine-Tuning: epoch 9/10: 100%|██████████| 507/507 [00:02<00:00, 240.10it/s, loss=0.432]\n",
      "Fine-Tuning Validation: 100%|██████████| 22/22 [00:00<00:00, 714.25it/s, val_loss=0.328]\n",
      "Fine-Tuning: epoch 10/10: 100%|██████████| 507/507 [00:02<00:00, 234.96it/s, loss=0.428]\n",
      "Fine-Tuning Validation: 100%|██████████| 22/22 [00:00<00:00, 716.93it/s, val_loss=0.324]\n"
     ]
    }
   ],
   "source": [
    "hyena_fine_tune.train(train_dataset=train_dataset, train_labels=dataset_train[\"label\"], validation_dataset=test_dataset, validation_labels=dataset_test[\"label\"], epochs=10, optimizer_params={\"lr\": 2e-6}, lr_scheduler_params={\"name\": \"linear\", \"num_warmup_steps\": 0})"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Evaluate the fine-tuned model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We now get the outputs of the model on our test dataset and display some results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 22/22 [00:00<00:00, 782.08it/s]\n"
     ]
    }
   ],
   "source": [
    "outputs = hyena_fine_tune.get_outputs(test_dataset)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.87      0.88      0.87       106\n",
      "           1       0.88      0.87      0.87       106\n",
      "\n",
      "    accuracy                           0.87       212\n",
      "   macro avg       0.87      0.87      0.87       212\n",
      "weighted avg       0.87      0.87      0.87       212\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(classification_report(dataset_test[\"label\"], outputs.argmax(axis=1)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1000x1000 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Compute the confusion matrix\n",
    "cm = confusion_matrix(dataset_test[\"label\"], outputs.argmax(axis=1))\n",
    "\n",
    "# Perform row-wise normalization\n",
    "cm_normalized = cm.astype('float') / cm.sum(axis=1)[:, np.newaxis]\n",
    "\n",
    "# Get unique labels in the order they appear in the confusion matrix\n",
    "unique_labels = np.unique(np.concatenate((dataset_test[\"label\"], outputs.argmax(axis=1))))\n",
    "\n",
    "# Create and plot the normalized confusion matrix\n",
    "fig, ax = plt.subplots(figsize=(10, 10))\n",
    "disp = ConfusionMatrixDisplay(confusion_matrix=cm_normalized, display_labels=unique_labels)\n",
    "disp.plot(ax=ax, xticks_rotation='vertical', values_format='.2f', cmap='coolwarm')\n",
    "\n",
    "# Customize the plot\n",
    "ax.set_title('Normalized Confusion Matrix (Row-wise)')\n",
    "fig.set_facecolor(\"none\")\n",
    "\n",
    "plt.tight_layout()\n",
    "plt.show()"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "helical-dev",
   "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": "3.10.15"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}