{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# scRep embedding example\n",
    "\n",
    "Run this notebook from the root of the downloaded scRep model repository."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "!pip install -r requirements.txt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from pathlib import Path\n",
    "from types import SimpleNamespace\n",
    "import numpy as np\n",
    "from scRep_inference import encode_embeddings, load_examples, load_scRep_bundle\n",
    "from scRep_pretrain.vocab import gene_vocab_to_map\n",
    "\n",
    "checkpoint = Path('checkpoints/scRep_20260625_30M')\n",
    "input_h5ad = Path('/path/to/input.h5ad')  # change this\n",
    "args = SimpleNamespace(\n",
    "    model_dir=str(checkpoint), asset_dir='', device='cuda',\n",
    "    batch_size=128, n_bins=50, max_input_genes=2048, backbone='auto',\n",
    ")\n",
    "model, gene_vocab, *_ = load_scRep_bundle(checkpoint, args)\n",
    "gene_name_to_id = gene_vocab_to_map(gene_vocab)\n",
    "examples, label_key = load_examples(input_h5ad, gene_name_to_id)\n",
    "embeddings = encode_embeddings(model, examples, args)\n",
    "np.save('embeddings.npy', embeddings)\n",
    "print(embeddings.shape, label_key)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {"display_name": "Python 3", "language": "python", "name": "python3"},
  "language_info": {"name": "python", "version": "3.10"}
 },
 "nbformat": 4,
 "nbformat_minor": 5
}