{
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "# X-Pert for Drug + Gene Co-Training (Multi-Perturbation)\n",
        "\n",
        "This notebook adapts the drug-only workflow to support joint training with drug and gene perturbations.\n",
        "\n",
        "Key points:\n",
        "- Unified `pert_mode = 'drug_gene'` for mixed perturbations\n",
        "- Gene embeddings (e.g., scGPT token vectors) and drug embeddings (e.g., ECFP/GROVER) are both loaded\n",
        "- Batch-level `drug_mask` splits perturbations by type to build two modality-specific embedding tensors\n",
        "- Optional dosage modeling for drugs only\n",
        "- Compatible with scGPT-based encoder and scFlamingo cross-modality design\n",
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "## 1. Imports\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 1,
      "metadata": {},
      "outputs": [],
      "source": [
        "# Core\n",
        "import os, json, time, copy, warnings\n",
        "from pathlib import Path\n",
        "from typing import Dict\n",
        "\n",
        "import numpy as np\n",
        "import pandas as pd\n",
        "import torch\n",
        "import torch.nn.functional as F\n",
        "from torch import nn\n",
        "from tqdm import tqdm\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "# X-Pert package\n",
        "import xpert\n",
        "from xpert.models import TransformerGenerator\n",
        "from xpert.loss import masked_relative_error, masked_mse_loss, masked_huber_loss\n",
        "from xpert.external_model.scgpt.gene_tokenizer import GeneVocab\n",
        "from xpert.external_model.gears.inference import compute_metrics\n",
        "# from xpert.data import Byte_Pert_Data\n",
        "\n",
        "# Utils\n",
        "from xpert.utils import fix_seed, merge_plot\n",
        "from xpert.external_model.scgpt.util import set_seed, map_raw_id_to_vocab_id, add_file_handler\n",
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "## 2. Global Settings and Logger\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 2,
      "metadata": {},
      "outputs": [
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "INFO:xpert:Running on 2025-11-03 17:05:57\n"
          ]
        }
      ],
      "source": [
        "# Reproducibility\n",
        "set_seed(42)\n",
        "fix_seed(2024)\n",
        "\n",
        "# Logger to file\n",
        "save_root = Path(\"./L1000_phase1_cotrain/model_mode_scFlamingo_drug_gene_v1\")\n",
        "save_root.mkdir(parents=True, exist_ok=True)\n",
        "logger = xpert.logger\n",
        "add_file_handler(logger, save_root / \"run.log\")\n",
        "logger.info(f\"Running on {time.strftime('%Y-%m-%d %H:%M:%S')}\")\n",
        "\n",
        "# Plot default\n",
        "plt.rcdefaults()\n",
        "warnings.filterwarnings(\"ignore\")\n",
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "## 3. Data Configuration\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 3,
      "metadata": {},
      "outputs": [],
      "source": [
        "# Dataset and split\n",
        "prefix = 'L1000_phase1_cotrain'  # or 'sciplex_3', etc.\n",
        "add_control = False\n",
        "\n",
        "# Split strategies (for L1000 datasets)\n",
        "# data_split_0: random perts\n",
        "# data_split_1: split drugs\n",
        "# data_split_2: split cell types\n",
        "split_col = 'data_split_1'\n",
        "\n",
        "# Paths\n",
        "# Base data directory\n",
        "data_dir = Path('../../data') / prefix\n",
        "pert_data_version = 'pert_data.pkl'\n",
        "\n",
        "# Drug embedding directory (local repo data)\n",
        "drug_embed_dir = data_dir\n",
        "\n",
        "# Gene embedding directory (external scGPT-based embeddings), absolute path\n",
        "# These are used to embed gene perturbations so they align with scGPT vocabulary space\n",
        "# If you have a local copy, adjust path accordingly\n",
        "gene_embed_file = Path('../../data/L1000_phase1_cotrain/pert_embed.csv')\n",
        "\n",
        "# scGPT pretrained model\n",
        "load_model_dir = Path('../../data/scGPT_human')\n",
        "model_config_file = load_model_dir / 'args.json'\n",
        "model_file = load_model_dir / 'best_model.pt'\n",
        "vocab_file = load_model_dir / 'vocab.json'\n",
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "## 4. Hyperparameters and Modes\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 4,
      "metadata": {},
      "outputs": [],
      "source": [
        "# Training hyperparameters\n",
        "batch_size = 64\n",
        "eval_batch_size = 64\n",
        "log_interval = 100\n",
        "\n",
        "# Cell encoder hyperparameters\n",
        "embsize = 512\n",
        "d_hid = 512\n",
        "nlayers = 12\n",
        "nhead = 8\n",
        "n_layers_cls = 3\n",
        "dropout = 0.2\n",
        "use_fast_transformer = True\n",
        "amp = True\n",
        "# Adjust based on available GPUs\n",
        "device_ids = [0]\n",
        "\n",
        "# Perturbation settings\n",
        "pert_mode = 'drug_gene'  # joint training for drug + gene\n",
        "pert_flag_mode = False   # use binary flags when True; set False for co-train\n",
        "attn_gate_mode = True\n",
        "load_cxg_weight = True\n",
        "mask_mode = True\n",
        "use_scgpt_layer = True\n",
        "use_scgpt_input = True\n",
        "add_token = True\n",
        "init_mode = False\n",
        "cross_mode = True\n",
        "pert_type_mode = True\n",
        "delta_mode = False\n",
        "\n",
        "# Co-train alignment settings (match scFlamingo_drug_v27)\n",
        "co_train_mode = 1\n",
        "align_loss = 'mmd'\n",
        "alpha_mmd = 0.2\n",
        "\n",
        "# Drug embedding settings\n",
        "drug_embed_mode = 'ecfp'  # 'grover' | 'rdkit' | 'morgan' | 'ecfp' | 'chemberta_st' | 'grover+rdkit'\n",
        "\n",
        "# Dosage settings (applied to drugs only)\n",
        "dosage_mode = False       # v27 disables dosage\n",
        "dosage_mode_type = -1     # -1 disables; 0/1/2 refer to strategies if enabled\n",
        "\n",
        "# Scheduler\n",
        "epochs = 20\n",
        "lr = 1e-5                 # v27 uses lower LR\n",
        "scheduler_type = 'cosine_warm'  # v27\n",
        "schedule_interval = 5\n",
        "\n",
        "# Gradient clipping\n",
        "max_norm = 1.0\n",
        "\n",
        "# Control whether to copy encoder weights into encoder_plus when adding tokens\n",
        "load_encoder_plus = True\n",
        "include_zero_gene = \"all\"\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "## 5. Scheduler and Training/Eval Routines\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 5,
      "metadata": {},
      "outputs": [],
      "source": [
        "from torch.optim.lr_scheduler import LambdaLR\n",
        "import math\n",
        "\n",
        "\n",
        "def get_cosine_schedule_with_warmup(optimizer, num_warmup_steps, num_training_steps):\n",
        "    def lr_lambda(current_step):\n",
        "        if current_step < num_warmup_steps:\n",
        "            return float(current_step) / float(max(1, num_warmup_steps))\n",
        "        progress = (current_step - num_warmup_steps) / float(max(1, num_training_steps - num_warmup_steps))\n",
        "        return max(0.0, 0.5 * (1.0 + math.cos(math.pi * progress)))\n",
        "    return LambdaLR(optimizer, lr_lambda)\n",
        "\n",
        "\n",
        "def _split_perturbations_and_build_embeddings(\n",
        "    batch_perts,\n",
        "    batch_pert_types,\n",
        "    pert_embed_dict: Dict[str, np.ndarray],\n",
        "    gpt_emb_dim_gene: int,\n",
        "    gpt_emb_dim_drug: int,\n",
        "    dosage_list=None,\n",
        "    device=None,\n",
        "):\n",
        "    \"\"\"\n",
        "    Build per-batch embedding tensors for gene and drug perturbations using a boolean drug_mask.\n",
        "    - batch_perts: list[str], format \"drug1; gene2 | dose1; dose2\" or \"geneA; geneB | ctrl\" for genes\n",
        "    - batch_pert_types: list[str], entries like 'trt_cp' (drug) or 'trt_sh.cgs' (gene)\n",
        "    - dosage_list: Optional[List[List[float]]] for drugs only, already aligned by sample\n",
        "    Returns: (batch_pert_embed_gene, batch_pert_embed_drug, pert_mask, batch_dosages_pad, drug_mask)\n",
        "    \"\"\"\n",
        "    B = len(batch_perts)\n",
        "    # tokenize perts and dosages\n",
        "    perts_split = [p.split(' | ')[0].split('; ') for p in batch_perts]\n",
        "    max_len = max(len(p) for p in perts_split) if B > 0 else 1\n",
        "\n",
        "    drug_mask = np.array(batch_pert_types) == 'trt_cp'\n",
        "\n",
        "    emb_gene = torch.zeros(sum(~drug_mask), max_len, gpt_emb_dim_gene).float()\n",
        "    emb_drug = torch.zeros(sum(drug_mask), max_len, gpt_emb_dim_drug).float()\n",
        "    pert_mask = torch.ones(B, max_len)\n",
        "    dos_pad = torch.zeros(sum(drug_mask), max_len).float()\n",
        "\n",
        "    # fill mask\n",
        "    for i, perts in enumerate(perts_split):\n",
        "        for j, pert in enumerate(perts):\n",
        "            pert_mask[i, j] = 0\n",
        "\n",
        "    # fill gene/drug embeddings separately (keep relative order)\n",
        "    gi = 0\n",
        "    di = 0\n",
        "    for idx, perts in enumerate(perts_split):\n",
        "        if drug_mask[idx]:\n",
        "            for j, pert in enumerate(perts):\n",
        "                emb_drug[di, j, :] = torch.tensor(np.array(pert_embed_dict[pert]))\n",
        "                if dosage_list is not None:\n",
        "                    dos_pad[di, j] = dosage_list[idx][j]\n",
        "            di += 1\n",
        "        else:\n",
        "            for j, pert in enumerate(perts):\n",
        "                emb_gene[gi, j, :] = torch.tensor(np.array(pert_embed_dict[pert]))\n",
        "            gi += 1\n",
        "\n",
        "    if device is not None:\n",
        "        emb_gene = emb_gene.to(device)\n",
        "        emb_drug = emb_drug.to(device)\n",
        "        dos_pad = dos_pad.to(device)\n",
        "        pert_mask = pert_mask.to(device)\n",
        "\n",
        "    return emb_gene, emb_drug, pert_mask, dos_pad, drug_mask\n",
        "\n",
        "\n",
        "def train(model: nn.Module, train_loader: torch.utils.data.DataLoader) -> None:\n",
        "    model.train()\n",
        "    total_loss, total_mse = 0.0, 0.0\n",
        "    start_time = time.time()\n",
        "    num_batches = len(train_loader)\n",
        "\n",
        "    for batch, batch_data in enumerate(tqdm(train_loader)):\n",
        "        batch_size = len(batch_data.y)\n",
        "        batch_data.to(device)\n",
        "        x: torch.Tensor = batch_data.x\n",
        "        ori_gene_values = x\n",
        "        if pert_mode == 'drug' and pert_flag_mode:\n",
        "            pert_flags = batch_data.pert_flags\n",
        "        else:\n",
        "            pert_flags = batch_data.pert_flags.long()\n",
        "        target_gene_values = batch_data.y\n",
        "\n",
        "        # Prepare perts/types and optional dosages (only for drugs)\n",
        "        batch_perts = batch_data.pert\n",
        "        batch_pert_types = batch_data.pert_type\n",
        "        # Extract dosages for drug items only (log10(1 + dose)) if dosage_mode\n",
        "        dosage_list = None\n",
        "        if dosage_mode:\n",
        "            drug_indices = np.where(np.array(batch_pert_types) == 'trt_cp')[0]\n",
        "            dosage_list_full = [[np.log10(float(v) + 1) for v in s.split(' | ')[1].split('; ')] for s in batch_data.pert]\n",
        "            dosage_list = [dosage_list_full[i] for i in drug_indices]\n",
        "\n",
        "        batch_pert_embed_gene, batch_pert_embed_drug, pert_mask, batch_dosages_pad, drug_mask = _split_perturbations_and_build_embeddings(\n",
        "            batch_perts=batch_perts,\n",
        "            batch_pert_types=batch_pert_types,\n",
        "            pert_embed_dict=pert_embed_dict,\n",
        "            gpt_emb_dim_gene=gpt_emb_dim_gene,\n",
        "            gpt_emb_dim_drug=gpt_emb_dim_drug,\n",
        "            dosage_list=dosage_list if dosage_mode else None,\n",
        "            device=device,\n",
        "        )\n",
        "\n",
        "        # Select input genes\n",
        "        if include_zero_gene in [\"all\", \"batch-wise\"]:\n",
        "            if include_zero_gene == \"all\":\n",
        "                input_gene_ids = torch.arange(n_genes, device=device, dtype=torch.long)\n",
        "            else:\n",
        "                input_gene_ids = (\n",
        "                    ori_gene_values.nonzero()[:, 1].flatten().unique().sort()[0]\n",
        "                )\n",
        "            if len(input_gene_ids) > max_seq_len:\n",
        "                input_gene_ids = torch.randperm(len(input_gene_ids), device=device)[:max_seq_len]\n",
        "            input_values = ori_gene_values[:, input_gene_ids]\n",
        "            input_pert_flags = pert_flags[:, input_gene_ids]\n",
        "            target_values = target_gene_values[:, input_gene_ids]\n",
        "            if delta_mode:\n",
        "                target_values = target_values - input_values\n",
        "            mapped_input_gene_ids = map_raw_id_to_vocab_id(input_gene_ids, gene_ids)\n",
        "            mapped_input_gene_ids = mapped_input_gene_ids.repeat(batch_size, 1)\n",
        "            src_key_padding_mask = torch.zeros_like(input_values, dtype=torch.bool, device=device)\n",
        "\n",
        "        with torch.cuda.amp.autocast(enabled=amp):\n",
        "            output_dict = model(\n",
        "                mapped_input_gene_ids,\n",
        "                input_values,\n",
        "                input_pert_flags,\n",
        "                src_key_padding_mask=src_key_padding_mask,\n",
        "                batch_pert_embed=None,  # not used in co-train; we pass per-modality below\n",
        "                pert_mask=pert_mask,\n",
        "                batch_dosages_pad=batch_dosages_pad if dosage_mode else None,\n",
        "                batch_pert_embed_gene=batch_pert_embed_gene,\n",
        "                batch_pert_embed_drug=batch_pert_embed_drug,\n",
        "                drug_mask=drug_mask,\n",
        "                CLS=False, CCE=False, MVC=False, ECS=False,\n",
        "            )\n",
        "            output_values = output_dict[\"mlm_output\"]\n",
        "            masked_positions = torch.ones_like(input_values, dtype=torch.bool)\n",
        "            loss = loss_mse = criterion(output_values, target_values, masked_positions)\n",
        "\n",
        "        model.zero_grad()\n",
        "        scaler.scale(loss).backward()\n",
        "        scaler.unscale_(optimizer)\n",
        "        torch.nn.utils.clip_grad_norm_(\n",
        "            model.parameters(), max_norm, error_if_nonfinite=False if scaler.is_enabled() else True\n",
        "        )\n",
        "        scaler.step(optimizer)\n",
        "        scaler.update()\n",
        "        if scheduler_type in ['cosine', 'cosine_warm']:\n",
        "            scheduler.step()\n",
        "\n",
        "        total_loss += loss.item()\n",
        "        total_mse += loss_mse.item()\n",
        "        if batch % log_interval == 0 and batch > 0:\n",
        "            lr_ = scheduler.get_last_lr()[0] if scheduler_type != 'steplr' else scheduler.get_last_lr()[0]\n",
        "            ms_per_batch = (time.time() - start_time) * 1000 / log_interval\n",
        "            cur_loss = total_loss / log_interval\n",
        "            cur_mse = total_mse / log_interval\n",
        "            logger.info(f\"| epoch {epoch:3d} | {batch:3d}/{num_batches:3d} batches | lr {lr_:07.6f} | ms/batch {ms_per_batch:5.2f} | loss {cur_loss:5.5f} | mse {cur_mse:5.5f} |\")\n",
        "            total_loss = 0\n",
        "            total_mse = 0\n",
        "            start_time = time.time()\n",
        "\n",
        "\n",
        "@torch.no_grad()\n",
        "def evaluate(model: nn.Module, val_loader: torch.utils.data.DataLoader):\n",
        "    model.eval()\n",
        "    total_loss = 0.0\n",
        "    total_error = 0.0\n",
        "    for batch, batch_data in enumerate(val_loader):\n",
        "        batch_size = len(batch_data.y)\n",
        "        batch_data.to(device)\n",
        "        x: torch.Tensor = batch_data.x\n",
        "        ori_gene_values = x\n",
        "        if pert_mode == 'drug' and pert_flag_mode:\n",
        "            pert_flags = batch_data.pert_flags\n",
        "        else:\n",
        "            pert_flags = batch_data.pert_flags.long()\n",
        "        target_gene_values = batch_data.y\n",
        "\n",
        "        batch_perts = batch_data.pert\n",
        "        batch_pert_types = batch_data.pert_type\n",
        "        dosage_list = None\n",
        "        if dosage_mode:\n",
        "            drug_indices = np.where(np.array(batch_pert_types) == 'trt_cp')[0]\n",
        "            dosage_list_full = [[np.log10(float(v) + 1) for v in s.split(' | ')[1].split('; ')] for s in batch_data.pert]\n",
        "            dosage_list = [dosage_list_full[i] for i in drug_indices]\n",
        "\n",
        "        batch_pert_embed_gene, batch_pert_embed_drug, pert_mask, batch_dosages_pad, drug_mask = _split_perturbations_and_build_embeddings(\n",
        "            batch_perts=batch_perts,\n",
        "            batch_pert_types=batch_pert_types,\n",
        "            pert_embed_dict=pert_embed_dict,\n",
        "            gpt_emb_dim_gene=gpt_emb_dim_gene,\n",
        "            gpt_emb_dim_drug=gpt_emb_dim_drug,\n",
        "            dosage_list=dosage_list if dosage_mode else None,\n",
        "            device=device,\n",
        "        )\n",
        "\n",
        "        if include_zero_gene in [\"all\", \"batch-wise\"]:\n",
        "            if include_zero_gene == \"all\":\n",
        "                input_gene_ids = torch.arange(n_genes, device=device)\n",
        "            else:\n",
        "                input_gene_ids = (\n",
        "                    ori_gene_values.nonzero()[:, 1].flatten().unique().sort()[0]\n",
        "                )\n",
        "            if len(input_gene_ids) > max_seq_len:\n",
        "                input_gene_ids = torch.randperm(len(input_gene_ids), device=device)[:max_seq_len]\n",
        "            input_values = ori_gene_values[:, input_gene_ids]\n",
        "            input_pert_flags = pert_flags[:, input_gene_ids]\n",
        "            target_values = target_gene_values[:, input_gene_ids]\n",
        "            if delta_mode:\n",
        "                target_values = target_values - input_values\n",
        "            mapped_input_gene_ids = map_raw_id_to_vocab_id(input_gene_ids, gene_ids)\n",
        "            mapped_input_gene_ids = mapped_input_gene_ids.repeat(batch_size, 1)\n",
        "            src_key_padding_mask = torch.zeros_like(input_values, dtype=torch.bool, device=input_values.device)\n",
        "\n",
        "        with torch.cuda.amp.autocast(enabled=amp):\n",
        "            output_dict = model(\n",
        "                mapped_input_gene_ids,\n",
        "                input_values,\n",
        "                input_pert_flags,\n",
        "                src_key_padding_mask=src_key_padding_mask,\n",
        "                batch_pert_embed=None,\n",
        "                pert_mask=pert_mask,\n",
        "                batch_dosages_pad=batch_dosages_pad if dosage_mode else None,\n",
        "                batch_pert_embed_gene=batch_pert_embed_gene,\n",
        "                batch_pert_embed_drug=batch_pert_embed_drug,\n",
        "                drug_mask=drug_mask,\n",
        "                do_sample=True,\n",
        "            )\n",
        "            output_values = output_dict[\"mlm_output\"]\n",
        "            masked_positions = torch.ones_like(input_values, dtype=torch.bool, device=input_values.device)\n",
        "            loss = criterion(output_values, target_values, masked_positions)\n",
        "        total_loss += loss.item()\n",
        "        total_error += masked_relative_error(output_values, target_values, masked_positions).item()\n",
        "    return total_loss / len(val_loader), total_error / len(val_loader)\n",
        "\n",
        "\n",
        "@torch.no_grad()\n",
        "def pred_perturb_new(model, batch_data, include_zero_gene=\"batch-wise\", gene_ids=None, amp=True):\n",
        "    model.eval()\n",
        "    device = next(model.parameters()).device\n",
        "    batch_data.to(device)\n",
        "    batch_size = len(batch_data.pert)\n",
        "    x: torch.Tensor = batch_data.x\n",
        "    ori_gene_values = x\n",
        "    if pert_mode == 'drug' and pert_flag_mode:\n",
        "        pert_flags = batch_data.pert_flags\n",
        "    else:\n",
        "        pert_flags = batch_data.pert_flags.long()\n",
        "\n",
        "    batch_perts = batch_data.pert\n",
        "    batch_pert_types = batch_data.pert_type\n",
        "\n",
        "    dosage_list = None\n",
        "    if dosage_mode:\n",
        "        drug_indices = np.where(np.array(batch_pert_types) == 'trt_cp')[0]\n",
        "        dosage_list_full = [[np.log10(float(v) + 1) for v in s.split(' | ')[1].split('; ')] for s in batch_data.pert]\n",
        "        dosage_list = [dosage_list_full[i] for i in drug_indices]\n",
        "\n",
        "    batch_pert_embed_gene, batch_pert_embed_drug, pert_mask, batch_dosages_pad, drug_mask = _split_perturbations_and_build_embeddings(\n",
        "        batch_perts=batch_perts,\n",
        "        batch_pert_types=batch_pert_types,\n",
        "        pert_embed_dict=pert_embed_dict,\n",
        "        gpt_emb_dim_gene=gpt_emb_dim_gene,\n",
        "        gpt_emb_dim_drug=gpt_emb_dim_drug,\n",
        "        dosage_list=dosage_list if dosage_mode else None,\n",
        "        device=device,\n",
        "    )\n",
        "\n",
        "    if include_zero_gene in [\"all\", \"batch-wise\"]:\n",
        "        assert gene_ids is not None\n",
        "        if include_zero_gene == \"all\":\n",
        "            input_gene_ids = torch.arange(ori_gene_values.size(1), device=device)\n",
        "        else:\n",
        "            input_gene_ids = (\n",
        "                ori_gene_values.nonzero()[:, 1].flatten().unique().sort()[0]\n",
        "            )\n",
        "        input_values = ori_gene_values[:, input_gene_ids]\n",
        "        input_pert_flags = pert_flags[:, input_gene_ids]\n",
        "        mapped_input_gene_ids = map_raw_id_to_vocab_id(input_gene_ids, gene_ids)\n",
        "        mapped_input_gene_ids = mapped_input_gene_ids.repeat(batch_size, 1)\n",
        "        src_key_padding_mask = torch.zeros_like(input_values, dtype=torch.bool, device=device)\n",
        "        with torch.cuda.amp.autocast(enabled=amp):\n",
        "            output_dict = model(\n",
        "                mapped_input_gene_ids,\n",
        "                input_values,\n",
        "                input_pert_flags,\n",
        "                src_key_padding_mask=src_key_padding_mask,\n",
        "                batch_pert_embed=None,\n",
        "                pert_mask=pert_mask,\n",
        "                batch_dosages_pad=batch_dosages_pad if dosage_mode else None,\n",
        "                batch_pert_embed_gene=batch_pert_embed_gene,\n",
        "                batch_pert_embed_drug=batch_pert_embed_drug,\n",
        "                drug_mask=drug_mask,\n",
        "                CLS=False, CCE=False, MVC=False, ECS=False,\n",
        "                do_sample=True,\n",
        "            )\n",
        "        output_values = output_dict[\"mlm_output\"].float()\n",
        "        pred_gene_values = torch.zeros_like(ori_gene_values)\n",
        "        pred_gene_values[:, input_gene_ids] = output_values\n",
        "        if delta_mode:\n",
        "            pred_gene_values = input_values + pred_gene_values\n",
        "    return pred_gene_values\n",
        "\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 15,
      "metadata": {},
      "outputs": [],
      "source": [
        "@torch.no_grad()\n",
        "def pred_perturb_new(\n",
        "    model,\n",
        "    batch_data,\n",
        "    include_zero_gene=\"batch-wise\",\n",
        "    gene_ids=None,\n",
        "    amp=True,\n",
        "):\n",
        "    \"\"\"\n",
        "    Args:\n",
        "        batch_data: a dictionary of input data with keys.\n",
        "\n",
        "    Returns:\n",
        "        output Tensor of shape [N, seq_len]\n",
        "    \"\"\"\n",
        "    model.eval()\n",
        "    device = next(model.parameters()).device\n",
        "    batch_data.to(device)\n",
        "    batch_size = len(batch_data.pert)\n",
        "    x: torch.Tensor = batch_data.x\n",
        "    # ori_gene_values = x[:, 0].view(batch_size, n_genes)\n",
        "    ori_gene_values = x\n",
        "    # pert_flags = x[:, 1].long().view(batch_size, n_genes)\n",
        "    if pert_mode == 'drug' and pert_flag_mode:\n",
        "        pert_flags = batch_data.pert_flags\n",
        "    else:\n",
        "        pert_flags = batch_data.pert_flags.long()\n",
        "\n",
        "\n",
        "    batch_perts = batch_data.pert\n",
        "    batch_pert_types = batch_data.pert_type\n",
        "    drug_mask = np.array(batch_pert_types)=='trt_cp'\n",
        "\n",
        "    # batch_perts = [[i.split('+')[0]] if 'ctrl' in i else i.split('+') for i in batch_perts] # we first focus on the single perts\n",
        "    batch_perts = [i.split(' | ')[0].split(\"; \") for i in batch_perts]\n",
        "    # batch_dosages = [[np.log10(int(j)) for j in i.split(' | ')[1].split(\"; \")] for i in batch_data.pert]\n",
        "    batch_dosages = [[np.log10(float(j)+1) for j in i.split(' | ')[1].split(\"; \")] for i in np.array(batch_data.pert)[drug_mask]]\n",
        "    max_pert_len = max([len(perts) for perts in batch_perts])\n",
        "\n",
        "    batch_pert_embed_gene = torch.zeros(sum(~drug_mask), max_pert_len, gpt_emb_dim_gene).float()\n",
        "    batch_pert_embed_drug = torch.zeros(sum(drug_mask), max_pert_len, gpt_emb_dim_drug).float()\n",
        "\n",
        "    pert_mask = torch.ones(batch_size, max_pert_len)\n",
        "    batch_dosages_pad = torch.zeros(sum(drug_mask), max_pert_len).float()\n",
        "    # batch_pert_embed = torch.tensor(np.array([(pert_embed_dict[pert[0]]) for pert in batch_perts])).float()\n",
        "    for i, perts in enumerate(batch_perts):\n",
        "        for j, pert in enumerate(perts):\n",
        "            # batch_pert_embed[i, j, :] = torch.tensor(np.array(pert_embed_dict[pert]))\n",
        "            pert_mask[i, j] = 0\n",
        "            # batch_dosages_pad[i, j] = batch_dosages[i][j]\n",
        "\n",
        "    for i, perts in enumerate(np.array(batch_perts)[~drug_mask]):\n",
        "        for j, pert in enumerate(perts):\n",
        "            batch_pert_embed_gene[i, j, :] = torch.tensor(np.array(pert_embed_dict[pert]))\n",
        "\n",
        "    for i, perts in enumerate(np.array(batch_perts)[drug_mask]):\n",
        "        for j, pert in enumerate(perts):\n",
        "            batch_pert_embed_drug[i, j, :] = torch.tensor(np.array(pert_embed_dict[pert]))\n",
        "            batch_dosages_pad[i, j] = batch_dosages[i][j]\n",
        "\n",
        "    batch_pert_embed_gene = batch_pert_embed_gene.to(device)\n",
        "    batch_pert_embed_drug = batch_pert_embed_drug.to(device)\n",
        "    batch_dosages_pad = batch_dosages_pad.to(device)\n",
        "\n",
        "    if not dosage_mode:\n",
        "        batch_dosages_pad = None\n",
        "\n",
        "    if include_zero_gene in [\"all\", \"batch-wise\"]:\n",
        "        assert gene_ids is not None\n",
        "        if include_zero_gene == \"all\":\n",
        "            input_gene_ids = torch.arange(ori_gene_values.size(1), device=device)\n",
        "        else:  # batch-wise\n",
        "            input_gene_ids = (\n",
        "                ori_gene_values.nonzero()[:, 1].flatten().unique().sort()[0]\n",
        "            )\n",
        "        input_values = ori_gene_values[:, input_gene_ids]\n",
        "        input_pert_flags = pert_flags[:, input_gene_ids]\n",
        "\n",
        "        mapped_input_gene_ids = map_raw_id_to_vocab_id(input_gene_ids, gene_ids)\n",
        "        mapped_input_gene_ids = mapped_input_gene_ids.repeat(batch_size, 1)\n",
        "\n",
        "        src_key_padding_mask = torch.zeros_like(\n",
        "            input_values, dtype=torch.bool, device=device\n",
        "        )\n",
        "        with torch.cuda.amp.autocast(enabled=amp):\n",
        "            output_dict = model(\n",
        "                mapped_input_gene_ids,\n",
        "                input_values,\n",
        "                input_pert_flags,\n",
        "                src_key_padding_mask=src_key_padding_mask,\n",
        "                batch_pert_embed = None,\n",
        "                pert_mask = pert_mask,\n",
        "                batch_dosages_pad = batch_dosages_pad,\n",
        "\n",
        "                batch_pert_embed_gene = batch_pert_embed_gene,\n",
        "                batch_pert_embed_drug = batch_pert_embed_drug,\n",
        "                drug_mask = drug_mask,\n",
        "\n",
        "                CLS=False,\n",
        "                CCE=False,\n",
        "                MVC=False,\n",
        "                ECS=False,\n",
        "                do_sample=True,\n",
        "            )\n",
        "        output_values = output_dict[\"mlm_output\"].float()\n",
        "        pred_gene_values = torch.zeros_like(ori_gene_values)\n",
        "        pred_gene_values[:, input_gene_ids] = output_values\n",
        "\n",
        "        if delta_mode:\n",
        "            pred_gene_values = input_values + pred_gene_values\n",
        "    return pred_gene_values\n",
        "\n",
        "\n",
        "\n",
        "@torch.no_grad()\n",
        "def eval_perturb_new(\n",
        "    loader: torch.utils.data.DataLoader, model: TransformerGenerator, device: torch.device\n",
        ") -> Dict:\n",
        "    \"\"\"\n",
        "    Run model in inference mode using a given data loader\n",
        "    \"\"\"\n",
        "\n",
        "    model.eval()\n",
        "    model.to(device)\n",
        "    pert_cat = []\n",
        "    pred = []\n",
        "    truth = []\n",
        "    pred_de = []\n",
        "    truth_de = []\n",
        "    results = {}\n",
        "    logvar = []\n",
        "\n",
        "    for itr, batch in enumerate(loader):\n",
        "        batch.to(device)\n",
        "        pert_cat.extend(batch.pert)\n",
        "\n",
        "        with torch.no_grad():\n",
        "            p = pred_perturb_new(model, batch, include_zero_gene, gene_ids=gene_ids)\n",
        "            t = batch.y\n",
        "            pred.extend(p.cpu())\n",
        "            truth.extend(t.cpu())\n",
        "\n",
        "            # Differentially expressed genes\n",
        "            for itr, de_idx in enumerate(batch.de_idx):\n",
        "                pred_de.append(p[itr, de_idx])\n",
        "                truth_de.append(t[itr, de_idx])\n",
        "\n",
        "        # if debug_mode_1:\n",
        "        #     break\n",
        "\n",
        "    # all genes\n",
        "    results[\"pert_cat\"] = np.array(pert_cat)\n",
        "    pred = torch.stack(pred)\n",
        "    truth = torch.stack(truth)\n",
        "    results[\"pred\"] = pred.detach().cpu().numpy().astype(np.float64)\n",
        "    results[\"truth\"] = truth.detach().cpu().numpy().astype(np.float64)\n",
        "\n",
        "    pred_de = torch.stack(pred_de)\n",
        "    truth_de = torch.stack(truth_de)\n",
        "    results[\"pred_de\"] = pred_de.detach().cpu().numpy().astype(np.float64)\n",
        "    results[\"truth_de\"] = truth_de.detach().cpu().numpy().astype(np.float64)\n",
        "\n",
        "    return results\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 7,
      "metadata": {},
      "outputs": [],
      "source": [
        "def pairwise_dist2(a, b):\n",
        "    a2 = (a ** 2).sum(-1, keepdim=True)\n",
        "    b2 = (b ** 2).sum(-1, keepdim=True)\n",
        "    return a2 + b2.T - 2 * a @ b.T\n",
        "\n",
        "\n",
        "def mmd_rbf(z_d, z_g, eps=1e-6):\n",
        "    z_d = z_d.squeeze().float()\n",
        "    z_g = z_g.squeeze().float()\n",
        "    dist2_dg = pairwise_dist2(z_d, z_g)\n",
        "    sigma2 = torch.quantile(dist2_dg, 0.5).clamp_min(eps)\n",
        "    k_dd = torch.exp(-pairwise_dist2(z_d, z_d) / (2 * sigma2)).mean()\n",
        "    k_gg = torch.exp(-pairwise_dist2(z_g, z_g) / (2 * sigma2)).mean()\n",
        "    k_dg = torch.exp(-dist2_dg / (2 * sigma2)).mean()\n",
        "    return k_dd + k_gg - 2 * k_dg\n",
        "\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 8,
      "metadata": {},
      "outputs": [],
      "source": [
        "# Override train/evaluate to include MMD alignment when enabled\n",
        "\n",
        "def train(model: nn.Module, train_loader: torch.utils.data.DataLoader) -> None:\n",
        "    model.train()\n",
        "    total_loss, total_mse = 0.0, 0.0\n",
        "    start_time = time.time()\n",
        "    num_batches = len(train_loader)\n",
        "\n",
        "    for batch, batch_data in enumerate(tqdm(train_loader)):\n",
        "        batch_size = len(batch_data.y)\n",
        "        batch_data.to(device)\n",
        "        x: torch.Tensor = batch_data.x\n",
        "        ori_gene_values = x\n",
        "        if pert_mode == 'drug' and pert_flag_mode:\n",
        "            pert_flags = batch_data.pert_flags\n",
        "        else:\n",
        "            pert_flags = batch_data.pert_flags.long()\n",
        "        target_gene_values = batch_data.y\n",
        "\n",
        "        batch_perts = batch_data.pert\n",
        "        batch_pert_types = batch_data.pert_type\n",
        "\n",
        "        dosage_list = None  # v27 disables dosage\n",
        "\n",
        "        batch_pert_embed_gene, batch_pert_embed_drug, pert_mask, batch_dosages_pad, drug_mask = _split_perturbations_and_build_embeddings(\n",
        "            batch_perts=batch_perts,\n",
        "            batch_pert_types=batch_pert_types,\n",
        "            pert_embed_dict=pert_embed_dict,\n",
        "            gpt_emb_dim_gene=gpt_emb_dim_gene,\n",
        "            gpt_emb_dim_drug=gpt_emb_dim_drug,\n",
        "            dosage_list=dosage_list,\n",
        "            device=device,\n",
        "        )\n",
        "\n",
        "        if include_zero_gene in [\"all\", \"batch-wise\"]:\n",
        "            if include_zero_gene == \"all\":\n",
        "                input_gene_ids = torch.arange(n_genes, device=device, dtype=torch.long)\n",
        "            else:\n",
        "                input_gene_ids = (\n",
        "                    ori_gene_values.nonzero()[:, 1].flatten().unique().sort()[0]\n",
        "                )\n",
        "            if len(input_gene_ids) > max_seq_len:\n",
        "                input_gene_ids = torch.randperm(len(input_gene_ids), device=device)[:max_seq_len]\n",
        "            input_values = ori_gene_values[:, input_gene_ids]\n",
        "            input_pert_flags = pert_flags[:, input_gene_ids]\n",
        "            target_values = target_gene_values[:, input_gene_ids]\n",
        "            if delta_mode:\n",
        "                target_values = target_values - input_values\n",
        "            mapped_input_gene_ids = map_raw_id_to_vocab_id(input_gene_ids, gene_ids)\n",
        "            mapped_input_gene_ids = mapped_input_gene_ids.repeat(batch_size, 1)\n",
        "            src_key_padding_mask = torch.zeros_like(input_values, dtype=torch.bool, device=device)\n",
        "\n",
        "        with torch.cuda.amp.autocast(enabled=amp):\n",
        "            output_dict = model(\n",
        "                mapped_input_gene_ids,\n",
        "                input_values,\n",
        "                input_pert_flags,\n",
        "                src_key_padding_mask=src_key_padding_mask,\n",
        "                batch_pert_embed=None,\n",
        "                pert_mask=pert_mask,\n",
        "                batch_dosages_pad=None,\n",
        "                batch_pert_embed_gene=batch_pert_embed_gene,\n",
        "                batch_pert_embed_drug=batch_pert_embed_drug,\n",
        "                drug_mask=drug_mask,\n",
        "                CLS=False, CCE=False, MVC=False, ECS=False,\n",
        "            )\n",
        "            output_values = output_dict[\"mlm_output\"]\n",
        "            masked_positions = torch.ones_like(input_values, dtype=torch.bool)\n",
        "            loss = loss_mse = criterion(output_values, target_values, masked_positions)\n",
        "\n",
        "            if align_loss == 'mmd' and 'latent_pert' in output_dict:\n",
        "                latent_pert = output_dict['latent_pert']\n",
        "                gene_mask = ~torch.tensor(drug_mask, dtype=torch.bool, device=latent_pert.device)\n",
        "                drug_mask_t = torch.tensor(drug_mask, dtype=torch.bool, device=latent_pert.device)\n",
        "                if drug_mask_t.any() and gene_mask.any():\n",
        "                    loss_mmd = mmd_rbf(latent_pert[drug_mask_t], latent_pert[gene_mask])\n",
        "                else:\n",
        "                    loss_mmd = latent_pert.new_tensor(0.)\n",
        "                loss = loss + alpha_mmd * loss_mmd\n",
        "\n",
        "        model.zero_grad()\n",
        "        scaler.scale(loss).backward()\n",
        "        scaler.unscale_(optimizer)\n",
        "        torch.nn.utils.clip_grad_norm_(\n",
        "            model.parameters(), max_norm, error_if_nonfinite=False if scaler.is_enabled() else True\n",
        "        )\n",
        "        scaler.step(optimizer)\n",
        "        scaler.update()\n",
        "        if scheduler_type in ['cosine', 'cosine_warm']:\n",
        "            scheduler.step()\n",
        "\n",
        "        total_loss += loss.item()\n",
        "        total_mse += loss_mse.item()\n",
        "        if batch % log_interval == 0 and batch > 0:\n",
        "            lr_ = scheduler.get_last_lr()[0] if scheduler_type != 'steplr' else scheduler.get_last_lr()[0]\n",
        "            ms_per_batch = (time.time() - start_time) * 1000 / log_interval\n",
        "            cur_loss = total_loss / log_interval\n",
        "            cur_mse = total_mse / log_interval\n",
        "            logger.info(f\"| epoch {epoch:3d} | {batch:3d}/{num_batches:3d} batches | lr {lr_:07.6f} | ms/batch {ms_per_batch:5.2f} | loss {cur_loss:5.5f} | mse {cur_mse:5.5f} |\")\n",
        "            total_loss = 0\n",
        "            total_mse = 0\n",
        "            start_time = time.time()\n",
        "\n",
        "\n",
        "@torch.no_grad()\n",
        "def evaluate(model: nn.Module, val_loader: torch.utils.data.DataLoader):\n",
        "    model.eval()\n",
        "    total_loss = 0.0\n",
        "    total_error = 0.0\n",
        "    for batch, batch_data in enumerate(val_loader):\n",
        "        batch_size = len(batch_data.y)\n",
        "        batch_data.to(device)\n",
        "        x: torch.Tensor = batch_data.x\n",
        "        ori_gene_values = x\n",
        "        if pert_mode == 'drug' and pert_flag_mode:\n",
        "            pert_flags = batch_data.pert_flags\n",
        "        else:\n",
        "            pert_flags = batch_data.pert_flags.long()\n",
        "        target_gene_values = batch_data.y\n",
        "\n",
        "        batch_perts = batch_data.pert\n",
        "        batch_pert_types = batch_data.pert_type\n",
        "\n",
        "        batch_pert_embed_gene, batch_pert_embed_drug, pert_mask, batch_dosages_pad, drug_mask = _split_perturbations_and_build_embeddings(\n",
        "            batch_perts=batch_perts,\n",
        "            batch_pert_types=batch_pert_types,\n",
        "            pert_embed_dict=pert_embed_dict,\n",
        "            gpt_emb_dim_gene=gpt_emb_dim_gene,\n",
        "            gpt_emb_dim_drug=gpt_emb_dim_drug,\n",
        "            dosage_list=None,\n",
        "            device=device,\n",
        "        )\n",
        "\n",
        "        if include_zero_gene in [\"all\", \"batch-wise\"]:\n",
        "            if include_zero_gene == \"all\":\n",
        "                input_gene_ids = torch.arange(n_genes, device=device)\n",
        "            else:\n",
        "                input_gene_ids = (\n",
        "                    ori_gene_values.nonzero()[:, 1].flatten().unique().sort()[0]\n",
        "                )\n",
        "            if len(input_gene_ids) > max_seq_len:\n",
        "                input_gene_ids = torch.randperm(len(input_gene_ids), device=device)[:max_seq_len]\n",
        "            input_values = ori_gene_values[:, input_gene_ids]\n",
        "            input_pert_flags = pert_flags[:, input_gene_ids]\n",
        "            target_values = target_gene_values[:, input_gene_ids]\n",
        "            if delta_mode:\n",
        "                target_values = target_values - input_values\n",
        "            mapped_input_gene_ids = map_raw_id_to_vocab_id(input_gene_ids, gene_ids)\n",
        "            mapped_input_gene_ids = mapped_input_gene_ids.repeat(batch_size, 1)\n",
        "            src_key_padding_mask = torch.zeros_like(input_values, dtype=torch.bool, device=input_values.device)\n",
        "\n",
        "        with torch.cuda.amp.autocast(enabled=amp):\n",
        "            output_dict = model(\n",
        "                mapped_input_gene_ids,\n",
        "                input_values,\n",
        "                input_pert_flags,\n",
        "                src_key_padding_mask=src_key_padding_mask,\n",
        "                batch_pert_embed=None,\n",
        "                pert_mask=pert_mask,\n",
        "                batch_dosages_pad=None,\n",
        "                batch_pert_embed_gene=batch_pert_embed_gene,\n",
        "                batch_pert_embed_drug=batch_pert_embed_drug,\n",
        "                drug_mask=drug_mask,\n",
        "                do_sample=True,\n",
        "            )\n",
        "            output_values = output_dict[\"mlm_output\"]\n",
        "            masked_positions = torch.ones_like(input_values, dtype=torch.bool, device=input_values.device)\n",
        "            loss = criterion(output_values, target_values, masked_positions)\n",
        "            if align_loss == 'mmd' and 'latent_pert' in output_dict:\n",
        "                latent_pert = output_dict['latent_pert']\n",
        "                gene_mask = ~torch.tensor(drug_mask, dtype=torch.bool, device=latent_pert.device)\n",
        "                drug_mask_t = torch.tensor(drug_mask, dtype=torch.bool, device=latent_pert.device)\n",
        "                if drug_mask_t.any() and gene_mask.any():\n",
        "                    loss_mmd = mmd_rbf(latent_pert[drug_mask_t], latent_pert[gene_mask])\n",
        "                else:\n",
        "                    loss_mmd = latent_pert.new_tensor(0.)\n",
        "                loss = loss + alpha_mmd * loss_mmd\n",
        "        total_loss += loss.item()\n",
        "        total_error += masked_relative_error(output_values, target_values, masked_positions).item()\n",
        "    return total_loss / len(val_loader), total_error / len(val_loader)\n",
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "## 6. Data Loading and Tokenization\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 9,
      "metadata": {},
      "outputs": [
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 8461/8461 [00:19<00:00, 441.48it/s]\n",
            "100%|██████████| 944/944 [00:02<00:00, 471.76it/s]\n",
            "0it [00:00, ?it/s]"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "========== get Data_scGPT finished!\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "\n"
          ]
        }
      ],
      "source": [
        "seed = 2024\n",
        "bs_train = 32\n",
        "bs_test = bs_train * 2\n",
        "\n",
        "# Load preprocessed pert_data object\n",
        "import pickle\n",
        "pert_data = pickle.load(open(data_dir / pert_data_version, 'rb'))\n",
        "\n",
        "# Set var_genes\n",
        "pert_data.var_genes = pert_data.adata_split.var_names\n",
        "\n",
        "# Build X-Pert training datasets\n",
        "# For co-train, keep split_col consistent with L1000 drug-split for comparability\n",
        "pert_data.get_Data_scgpt_2(\n",
        "    num_de_genes=pert_data.num_de_genes,\n",
        "    dataset_name=['train', 'test', 'val'],\n",
        "    add_control=add_control,\n",
        "    split_col=split_col,\n",
        "    pert_flag_mode=pert_flag_mode,\n",
        "    drug_embed_dir=drug_embed_dir,\n",
        "    pert_type_mode=pert_type_mode,\n",
        ")\n",
        "\n",
        "# Attach full AnnData and gene names\n",
        "pert_data.adata = pert_data.adata_split\n",
        "pert_data.adata.var[\"gene_name\"] = pert_data.adata.var_names\n",
        "\n",
        "# DataLoaders\n",
        "trainloader, testloader, valloader = pert_data.get_dataloader(\n",
        "    mode='all', \n",
        "    bs_train=int(bs_train)*max(1, len(device_ids)), \n",
        "    bs_test=int(bs_test)*max(1, len(device_ids))\n",
        ")\n",
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "## 7. Load Gene and Drug Embeddings\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 10,
      "metadata": {},
      "outputs": [
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "INFO:xpert:Loaded 3302 gene perts (dim=1536) and 6082 drug perts (dim=2048)\n"
          ]
        }
      ],
      "source": [
        "# Load gene embeddings (scGPT token space) and drug embeddings\n",
        "assert gene_embed_file.exists(), f\"Gene embedding file not found: {gene_embed_file}\"\n",
        "gene_embed = pd.read_csv(gene_embed_file, sep=',', index_col=0)\n",
        "\n",
        "# Load drug embeddings based on mode\n",
        "if drug_embed_mode == 'grover':\n",
        "    drug_embed = pd.read_csv(drug_embed_dir / 'embed_grover.csv', sep=',', index_col=0)\n",
        "elif drug_embed_mode == 'rdkit':\n",
        "    drug_embed = pd.read_csv(drug_embed_dir / 'embed_rdkit.csv', sep=',', index_col=0)\n",
        "elif drug_embed_mode == 'morgan':\n",
        "    drug_embed = pd.read_csv(drug_embed_dir / 'embed_morgan.csv', sep=',', index_col=0)\n",
        "elif drug_embed_mode == 'ecfp':\n",
        "    drug_embed = pd.read_csv(drug_embed_dir / 'embed_ecfp.csv', sep=',', index_col=0)\n",
        "elif drug_embed_mode == 'chemberta_st':\n",
        "    drug_embed = pd.read_csv(drug_embed_dir / 'embed_chemberta_st.csv', sep=',', index_col=0)\n",
        "elif drug_embed_mode == 'grover+rdkit':\n",
        "    emb1 = pd.read_csv(drug_embed_dir / 'embed_grover.csv', sep=',', index_col=0)\n",
        "    emb2 = pd.read_csv(drug_embed_dir / 'embed_rdkit.csv', sep=',', index_col=0)\n",
        "    from sklearn.preprocessing import StandardScaler\n",
        "    scaler = StandardScaler()\n",
        "    drug_embed = pd.concat([emb1, emb2], axis=0)\n",
        "    drug_embed = pd.DataFrame(scaler.fit_transform(drug_embed.T).T, index=drug_embed.index, columns=drug_embed.columns)\n",
        "else:\n",
        "    raise ValueError(f\"Unknown drug_embed_mode: {drug_embed_mode}\")\n",
        "\n",
        "# Build joint pert_embed_dict for both gene and drug perturbations\n",
        "# Identify unique gene and drug perts from the filtered perturbation list\n",
        "all_perts = np.array(pert_data.filter_perturbation_list)\n",
        "all_perts_names = [i.split(' | ')[0] for i in all_perts]\n",
        "\n",
        "# Split by observed pert types from obs if available\n",
        "obs = pert_data.adata_split.obs\n",
        "gene_perts = obs.loc[obs['pert_type']=='trt_sh.cgs', 'perturbation_new'].unique()\n",
        "drug_perts = obs.loc[obs['pert_type']=='trt_cp', 'perturbation_new'].unique()\n",
        "\n",
        "pert_embed_dict: Dict[str, np.ndarray] = {}\n",
        "np.random.seed(2024)\n",
        "\n",
        "# Gene embeddings\n",
        "for pert in gene_perts:\n",
        "    if pert in gene_embed.columns:\n",
        "        pert_embed_dict[pert] = gene_embed.loc[:, pert].values\n",
        "    else:\n",
        "        logger.warning(f'{pert} not in gene_embed, using random gene embedding proxy')\n",
        "        pert_embed_dict[pert] = gene_embed.loc[:, np.random.choice(gene_embed.columns, 1)[0]].values\n",
        "\n",
        "# Drug embeddings\n",
        "for pert in drug_perts:\n",
        "    if pert in drug_embed.columns:\n",
        "        pert_embed_dict[pert] = drug_embed.loc[:, pert].values\n",
        "    else:\n",
        "        logger.warning(f'{pert} not in drug_embed, using random drug embedding proxy')\n",
        "        pert_embed_dict[pert] = drug_embed.loc[:, np.random.choice(drug_embed.columns, 1)[0]].values\n",
        "\n",
        "# Embedding dims per modality\n",
        "gpt_emb_dim_gene = gene_embed.shape[0]\n",
        "gpt_emb_dim_drug = drug_embed.shape[0]\n",
        "\n",
        "logger.info(f\"Loaded {len(gene_perts)} gene perts (dim={gpt_emb_dim_gene}) and {len(drug_perts)} drug perts (dim={gpt_emb_dim_drug})\")\n",
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "## 8. Vocabulary, Tokens and Sequence Setup\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 11,
      "metadata": {},
      "outputs": [
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "INFO:xpert:match 978/978 genes in vocabulary of size 60726.\n"
          ]
        }
      ],
      "source": [
        "# Load vocab and extend if needed\n",
        "vocab = GeneVocab.from_file(vocab_file)\n",
        "special_tokens = [\"<pad>\", \"<cls>\", \"<eoc>\"]\n",
        "for s in special_tokens:\n",
        "    if s not in vocab:\n",
        "        vocab.append_token(s)\n",
        "\n",
        "vocab_ori = copy.deepcopy(vocab)\n",
        "if add_token:\n",
        "    add_genes = np.setdiff1d(pert_data.adata.var_names, list(vocab.get_stoi().keys()))\n",
        "    for gene in add_genes:\n",
        "        if gene not in vocab:\n",
        "            vocab.append_token(gene)\n",
        "\n",
        "pert_data.adata.var[\"id_in_vocab\"] = [1 if gene in vocab else -1 for gene in pert_data.adata.var_names]\n",
        "gene_ids_in_vocab = np.array(pert_data.adata.var[\"id_in_vocab\"])\n",
        "logger.info(f\"match {np.sum(gene_ids_in_vocab >= 0)}/{len(gene_ids_in_vocab)} genes in vocabulary of size {len(vocab)}.\")\n",
        "\n",
        "# Build token ids for input genes\n",
        "genes = pert_data.adata.var[\"gene_name\"].tolist()\n",
        "vocab.set_default_index(vocab[\"<pad>\"])\n",
        "gene_ids = np.array([vocab[g] if g in vocab else vocab[\"<pad>\"] for g in genes], dtype=int)\n",
        "n_genes = len(genes)\n",
        "\n",
        "# Sequence length\n",
        "max_seq_len = 6000\n",
        "\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {},
      "outputs": [],
      "source": [
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "## 9. Build Model, Optimizer and Scheduler\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 12,
      "metadata": {},
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Using simple batchnorm instead of domain specific batchnorm\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "INFO:xpert:Missing keys: 0 | Unexpected keys: 0\n"
          ]
        }
      ],
      "source": [
        "# Rebuild model and optim to enforce v27 settings\n",
        "with open(model_config_file, 'r') as f:\n",
        "    model_configs = json.load(f)\n",
        "embsize = model_configs.get('embsize', embsize)\n",
        "d_hid = model_configs.get('d_hid', d_hid)\n",
        "n_layers_cls = model_configs.get('n_layers_cls', n_layers_cls)\n",
        "\n",
        "ntokens = len(vocab_ori)\n",
        "model = TransformerGenerator(\n",
        "    ntokens,\n",
        "    embsize,\n",
        "    nhead,\n",
        "    d_hid,\n",
        "    nlayers,\n",
        "    nlayers_cls=n_layers_cls,\n",
        "    n_cls=1,\n",
        "    vocab=vocab,\n",
        "    dropout=dropout,\n",
        "    pad_token='<pad>',\n",
        "    pad_value=0,\n",
        "    pert_pad_id=2,\n",
        "    use_fast_transformer=use_fast_transformer,\n",
        "    pert_embed_dict=pert_embed_dict,\n",
        "    gpt_emb_dim=gpt_emb_dim_gene,\n",
        "    model_mode='scFlamingo_drug_v27',\n",
        "    attn_gate_mode=attn_gate_mode,\n",
        "    pert_mode=pert_mode,\n",
        "    drug_embed_mode=drug_embed_mode,\n",
        "    pert_flag_mode=pert_flag_mode,\n",
        "    use_scgpt_layer=use_scgpt_layer,\n",
        "    use_scgpt_input=use_scgpt_input,\n",
        "    mask_mode=mask_mode,\n",
        "    add_token=add_token,\n",
        "    init_mode=init_mode,\n",
        "    dosage_mode_type=dosage_mode_type,\n",
        "    cross_mode=cross_mode,\n",
        "    gpt_emb_dim_gene=gpt_emb_dim_gene,\n",
        "    gpt_emb_dim_drug=gpt_emb_dim_drug,\n",
        "    co_train_mode=co_train_mode,\n",
        "    align_loss=align_loss,\n",
        ")\n",
        "\n",
        "if load_cxg_weight:\n",
        "    load_param_prefixs = [\"encoder\", \"value_encoder\", \"transformer_encoder\"]\n",
        "    state = torch.load(model_file)\n",
        "    model_dict = model.state_dict()\n",
        "    pretrained_dict = {k: v for k, v in state.items() if any([k.startswith(p) for p in load_param_prefixs])}\n",
        "    model_dict.update(pretrained_dict)\n",
        "    _missing = model.load_state_dict(model_dict, strict=False)\n",
        "    logger.info(f\"Missing keys: {len(_missing.missing_keys)} | Unexpected keys: {len(_missing.unexpected_keys)}\")\n",
        "\n",
        "if add_token and load_encoder_plus:\n",
        "    with torch.no_grad():\n",
        "        n = model.encoder.embedding.num_embeddings\n",
        "        model.encoder_plus.embedding.weight[:n] = model.encoder.embedding.weight\n",
        "        model.encoder_plus.enc_norm.weight[:] = model.encoder.enc_norm.weight\n",
        "        model.encoder_plus.enc_norm.bias[:] = model.encoder.enc_norm.bias\n",
        "        pretrained_embed = model.encoder.embedding.weight\n",
        "        mean = pretrained_embed.mean(dim=0)\n",
        "        std = pretrained_embed.std(dim=0)\n",
        "        m = model.encoder_plus.embedding.num_embeddings - n\n",
        "        if m > 0:\n",
        "            model.encoder_plus.embedding.weight[n:] = torch.normal(mean=mean.expand(m, -1), std=std.expand(m, -1))\n",
        "\n",
        "device = torch.device(f\"cuda:{device_ids[0]}\" if torch.cuda.is_available() else \"cpu\")\n",
        "if torch.cuda.device_count() > 1:\n",
        "    try:\n",
        "        device_ids\n",
        "    except NameError:\n",
        "        device_ids = list(range(torch.cuda.device_count()))\n",
        "    model = torch.nn.DataParallel(model, device_ids=device_ids).to(device)\n",
        "else:\n",
        "    model = model.to(device)\n",
        "\n",
        "loss_type = 'mse'\n",
        "criterion = masked_mse_loss if loss_type == 'mse' else masked_huber_loss\n",
        "optimizer = torch.optim.Adam(model.parameters(), lr=lr)\n",
        "if scheduler_type == 'steplr':\n",
        "    scheduler = torch.optim.lr_scheduler.StepLR(optimizer, schedule_interval, gamma=0.95)\n",
        "elif scheduler_type == 'cosine':\n",
        "    scheduler = torch.optim.lr_scheduler.CosineAnnealingLR(optimizer, T_max=len(trainloader)*epochs)\n",
        "elif scheduler_type == 'cosine_warm':\n",
        "    scheduler = get_cosine_schedule_with_warmup(\n",
        "        optimizer,\n",
        "        num_warmup_steps=int(len(trainloader)*epochs*0.05),\n",
        "        num_training_steps=len(trainloader)*epochs,\n",
        "    )\n",
        "else:\n",
        "    raise ValueError('Unknown scheduler')\n",
        "\n",
        "scaler = torch.cuda.amp.GradScaler(enabled=amp)\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "## 10. Training Loop and Validation\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 13,
      "metadata": {},
      "outputs": [
        {
          "data": {
            "text/plain": [
              "20"
            ]
          },
          "execution_count": 13,
          "metadata": {},
          "output_type": "execute_result"
        }
      ],
      "source": [
        "epochs"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 16,
      "metadata": {},
      "outputs": [
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            " 37%|███▋      | 100/272 [00:21<00:36,  4.72it/s]INFO:xpert:| epoch   1 | 100/272 batches | lr 0.000010 | ms/batch 215.50 | loss 1.00488 | mse 1.00488 |\n",
            " 74%|███████▎  | 200/272 [00:42<00:15,  4.65it/s]INFO:xpert:| epoch   1 | 200/272 batches | lr 0.000010 | ms/batch 213.78 | loss 0.99828 | mse 0.99828 |\n",
            "100%|██████████| 272/272 [00:58<00:00,  4.68it/s]\n",
            "INFO:xpert:-----------------------------------------------------------------------------------------\n",
            "INFO:xpert:| end of epoch   1 | time: 59.90s | valid loss/mse 0.9853 |\n",
            "INFO:xpert:-----------------------------------------------------------------------------------------\n",
            "INFO:xpert:Best model with score 0.9853\n",
            " 37%|███▋      | 100/272 [00:21<00:36,  4.65it/s]INFO:xpert:| epoch   2 | 100/272 batches | lr 0.000010 | ms/batch 217.89 | loss 0.97153 | mse 0.97153 |\n",
            " 74%|███████▎  | 200/272 [00:43<00:15,  4.60it/s]INFO:xpert:| epoch   2 | 200/272 batches | lr 0.000010 | ms/batch 216.41 | loss 1.00774 | mse 1.00774 |\n",
            "100%|██████████| 272/272 [00:58<00:00,  4.64it/s]\n",
            "INFO:xpert:-----------------------------------------------------------------------------------------\n",
            "INFO:xpert:| end of epoch   2 | time: 60.42s | valid loss/mse 0.9806 |\n",
            "INFO:xpert:-----------------------------------------------------------------------------------------\n",
            "INFO:xpert:Best model with score 0.9806\n",
            " 37%|███▋      | 100/272 [00:21<00:37,  4.63it/s]INFO:xpert:| epoch   3 | 100/272 batches | lr 0.000010 | ms/batch 217.97 | loss 1.00138 | mse 1.00138 |\n",
            " 74%|███████▎  | 200/272 [00:43<00:15,  4.63it/s]INFO:xpert:| epoch   3 | 200/272 batches | lr 0.000009 | ms/batch 216.29 | loss 0.95521 | mse 0.95521 |\n",
            "100%|██████████| 272/272 [00:58<00:00,  4.63it/s]\n",
            "INFO:xpert:-----------------------------------------------------------------------------------------\n",
            "INFO:xpert:| end of epoch   3 | time: 60.46s | valid loss/mse 0.9823 |\n",
            "INFO:xpert:-----------------------------------------------------------------------------------------\n",
            " 37%|███▋      | 100/272 [00:21<00:37,  4.61it/s]INFO:xpert:| epoch   4 | 100/272 batches | lr 0.000009 | ms/batch 218.43 | loss 1.00173 | mse 1.00173 |\n",
            " 74%|███████▎  | 200/272 [00:43<00:15,  4.66it/s]INFO:xpert:| epoch   4 | 200/272 batches | lr 0.000009 | ms/batch 216.12 | loss 0.94610 | mse 0.94610 |\n",
            "100%|██████████| 272/272 [00:58<00:00,  4.63it/s]\n",
            "INFO:xpert:-----------------------------------------------------------------------------------------\n",
            "INFO:xpert:| end of epoch   4 | time: 60.45s | valid loss/mse 0.9963 |\n",
            "INFO:xpert:-----------------------------------------------------------------------------------------\n",
            " 37%|███▋      | 100/272 [00:21<00:37,  4.65it/s]INFO:xpert:| epoch   5 | 100/272 batches | lr 0.000009 | ms/batch 217.11 | loss 0.96026 | mse 0.96026 |\n",
            " 74%|███████▎  | 200/272 [00:43<00:15,  4.64it/s]INFO:xpert:| epoch   5 | 200/272 batches | lr 0.000009 | ms/batch 215.90 | loss 0.95277 | mse 0.95277 |\n",
            "100%|██████████| 272/272 [00:58<00:00,  4.64it/s]\n",
            "INFO:xpert:-----------------------------------------------------------------------------------------\n",
            "INFO:xpert:| end of epoch   5 | time: 60.30s | valid loss/mse 0.9906 |\n",
            "INFO:xpert:-----------------------------------------------------------------------------------------\n",
            " 37%|███▋      | 100/272 [00:21<00:37,  4.59it/s]INFO:xpert:| epoch   6 | 100/272 batches | lr 0.000008 | ms/batch 217.74 | loss 0.95488 | mse 0.95488 |\n",
            " 74%|███████▎  | 200/272 [00:43<00:15,  4.61it/s]INFO:xpert:| epoch   6 | 200/272 batches | lr 0.000008 | ms/batch 216.65 | loss 0.94644 | mse 0.94644 |\n",
            "100%|██████████| 272/272 [00:58<00:00,  4.63it/s]\n",
            "INFO:xpert:-----------------------------------------------------------------------------------------\n",
            "INFO:xpert:| end of epoch   6 | time: 60.49s | valid loss/mse 0.9987 |\n",
            "INFO:xpert:-----------------------------------------------------------------------------------------\n",
            " 37%|███▋      | 100/272 [00:21<00:37,  4.62it/s]INFO:xpert:| epoch   7 | 100/272 batches | lr 0.000007 | ms/batch 218.08 | loss 0.91249 | mse 0.91249 |\n",
            " 74%|███████▎  | 200/272 [00:43<00:15,  4.62it/s]INFO:xpert:| epoch   7 | 200/272 batches | lr 0.000007 | ms/batch 216.93 | loss 0.94172 | mse 0.94172 |\n",
            "100%|██████████| 272/272 [00:58<00:00,  4.63it/s]\n",
            "INFO:xpert:-----------------------------------------------------------------------------------------\n",
            "INFO:xpert:| end of epoch   7 | time: 60.54s | valid loss/mse 0.9965 |\n",
            "INFO:xpert:-----------------------------------------------------------------------------------------\n",
            " 37%|███▋      | 100/272 [00:21<00:36,  4.65it/s]INFO:xpert:| epoch   8 | 100/272 batches | lr 0.000007 | ms/batch 217.24 | loss 0.94788 | mse 0.94788 |\n",
            " 74%|███████▎  | 200/272 [00:43<00:15,  4.64it/s]INFO:xpert:| epoch   8 | 200/272 batches | lr 0.000006 | ms/batch 216.07 | loss 0.91103 | mse 0.91103 |\n",
            "100%|██████████| 272/272 [00:58<00:00,  4.65it/s]\n",
            "INFO:xpert:-----------------------------------------------------------------------------------------\n",
            "INFO:xpert:| end of epoch   8 | time: 60.28s | valid loss/mse 1.0237 |\n",
            "INFO:xpert:-----------------------------------------------------------------------------------------\n",
            " 37%|███▋      | 100/272 [00:21<00:37,  4.64it/s]INFO:xpert:| epoch   9 | 100/272 batches | lr 0.000006 | ms/batch 216.84 | loss 0.90037 | mse 0.90037 |\n",
            " 74%|███████▎  | 200/272 [00:42<00:15,  4.63it/s]INFO:xpert:| epoch   9 | 200/272 batches | lr 0.000006 | ms/batch 215.30 | loss 0.93209 | mse 0.93209 |\n",
            "100%|██████████| 272/272 [00:58<00:00,  4.65it/s]\n",
            "INFO:xpert:-----------------------------------------------------------------------------------------\n",
            "INFO:xpert:| end of epoch   9 | time: 60.19s | valid loss/mse 0.9994 |\n",
            "INFO:xpert:-----------------------------------------------------------------------------------------\n",
            " 37%|███▋      | 100/272 [00:21<00:37,  4.64it/s]INFO:xpert:| epoch  10 | 100/272 batches | lr 0.000005 | ms/batch 216.54 | loss 0.92763 | mse 0.92763 |\n",
            " 74%|███████▎  | 200/272 [00:42<00:15,  4.65it/s]INFO:xpert:| epoch  10 | 200/272 batches | lr 0.000005 | ms/batch 215.16 | loss 0.90561 | mse 0.90561 |\n",
            "100%|██████████| 272/272 [00:58<00:00,  4.65it/s]\n",
            "INFO:xpert:-----------------------------------------------------------------------------------------\n",
            "INFO:xpert:| end of epoch  10 | time: 60.21s | valid loss/mse 1.0008 |\n",
            "INFO:xpert:-----------------------------------------------------------------------------------------\n",
            " 37%|███▋      | 100/272 [00:21<00:37,  4.64it/s]INFO:xpert:| epoch  11 | 100/272 batches | lr 0.000004 | ms/batch 217.25 | loss 0.91454 | mse 0.91454 |\n",
            " 74%|███████▎  | 200/272 [00:43<00:15,  4.69it/s]INFO:xpert:| epoch  11 | 200/272 batches | lr 0.000004 | ms/batch 215.04 | loss 0.93146 | mse 0.93146 |\n",
            "100%|██████████| 272/272 [00:58<00:00,  4.65it/s]\n",
            "INFO:xpert:-----------------------------------------------------------------------------------------\n",
            "INFO:xpert:| end of epoch  11 | time: 60.23s | valid loss/mse 1.0044 |\n",
            "INFO:xpert:-----------------------------------------------------------------------------------------\n",
            " 37%|███▋      | 100/272 [00:21<00:37,  4.64it/s]INFO:xpert:| epoch  12 | 100/272 batches | lr 0.000003 | ms/batch 217.02 | loss 0.93312 | mse 0.93312 |\n",
            " 74%|███████▎  | 200/272 [00:43<00:15,  4.64it/s]INFO:xpert:| epoch  12 | 200/272 batches | lr 0.000003 | ms/batch 215.48 | loss 0.88561 | mse 0.88561 |\n",
            "100%|██████████| 272/272 [00:58<00:00,  4.65it/s]\n",
            "INFO:xpert:-----------------------------------------------------------------------------------------\n",
            "INFO:xpert:| end of epoch  12 | time: 60.21s | valid loss/mse 1.0112 |\n",
            "INFO:xpert:-----------------------------------------------------------------------------------------\n",
            " 37%|███▋      | 100/272 [00:21<00:37,  4.64it/s]INFO:xpert:| epoch  13 | 100/272 batches | lr 0.000003 | ms/batch 217.25 | loss 0.92406 | mse 0.92406 |\n",
            " 74%|███████▎  | 200/272 [00:43<00:15,  4.63it/s]INFO:xpert:| epoch  13 | 200/272 batches | lr 0.000002 | ms/batch 215.73 | loss 0.89997 | mse 0.89997 |\n",
            "100%|██████████| 272/272 [00:58<00:00,  4.65it/s]\n",
            "INFO:xpert:-----------------------------------------------------------------------------------------\n",
            "INFO:xpert:| end of epoch  13 | time: 60.28s | valid loss/mse 1.0017 |\n",
            "INFO:xpert:-----------------------------------------------------------------------------------------\n",
            " 37%|███▋      | 100/272 [00:21<00:36,  4.68it/s]INFO:xpert:| epoch  14 | 100/272 batches | lr 0.000002 | ms/batch 217.04 | loss 0.90202 | mse 0.90202 |\n",
            " 74%|███████▎  | 200/272 [00:43<00:15,  4.63it/s]INFO:xpert:| epoch  14 | 200/272 batches | lr 0.000002 | ms/batch 215.50 | loss 0.93869 | mse 0.93869 |\n",
            "100%|██████████| 272/272 [00:58<00:00,  4.66it/s]\n",
            "INFO:xpert:-----------------------------------------------------------------------------------------\n",
            "INFO:xpert:| end of epoch  14 | time: 60.15s | valid loss/mse 1.0029 |\n",
            "INFO:xpert:-----------------------------------------------------------------------------------------\n",
            " 37%|███▋      | 100/272 [00:21<00:37,  4.64it/s]INFO:xpert:| epoch  15 | 100/272 batches | lr 0.000001 | ms/batch 217.00 | loss 0.91758 | mse 0.91758 |\n",
            " 74%|███████▎  | 200/272 [00:43<00:15,  4.63it/s]INFO:xpert:| epoch  15 | 200/272 batches | lr 0.000001 | ms/batch 215.93 | loss 0.90154 | mse 0.90154 |\n",
            "100%|██████████| 272/272 [00:58<00:00,  4.65it/s]\n",
            "INFO:xpert:-----------------------------------------------------------------------------------------\n",
            "INFO:xpert:| end of epoch  15 | time: 60.30s | valid loss/mse 1.0071 |\n",
            "INFO:xpert:-----------------------------------------------------------------------------------------\n",
            " 37%|███▋      | 100/272 [00:21<00:37,  4.63it/s]INFO:xpert:| epoch  16 | 100/272 batches | lr 0.000001 | ms/batch 217.23 | loss 0.91806 | mse 0.91806 |\n",
            " 74%|███████▎  | 200/272 [00:43<00:15,  4.64it/s]INFO:xpert:| epoch  16 | 200/272 batches | lr 0.000001 | ms/batch 215.36 | loss 0.90915 | mse 0.90915 |\n",
            "100%|██████████| 272/272 [00:58<00:00,  4.65it/s]\n",
            "INFO:xpert:-----------------------------------------------------------------------------------------\n",
            "INFO:xpert:| end of epoch  16 | time: 60.23s | valid loss/mse 1.0055 |\n",
            "INFO:xpert:-----------------------------------------------------------------------------------------\n",
            " 37%|███▋      | 100/272 [00:21<00:37,  4.62it/s]INFO:xpert:| epoch  17 | 100/272 batches | lr 0.000000 | ms/batch 217.45 | loss 0.92557 | mse 0.92557 |\n",
            " 74%|███████▎  | 200/272 [00:43<00:15,  4.63it/s]INFO:xpert:| epoch  17 | 200/272 batches | lr 0.000000 | ms/batch 217.65 | loss 0.90771 | mse 0.90771 |\n",
            "100%|██████████| 272/272 [00:58<00:00,  4.63it/s]\n",
            "INFO:xpert:-----------------------------------------------------------------------------------------\n",
            "INFO:xpert:| end of epoch  17 | time: 60.52s | valid loss/mse 1.0058 |\n",
            "INFO:xpert:-----------------------------------------------------------------------------------------\n",
            " 37%|███▋      | 100/272 [00:21<00:37,  4.64it/s]INFO:xpert:| epoch  18 | 100/272 batches | lr 0.000000 | ms/batch 217.61 | loss 0.90280 | mse 0.90280 |\n",
            " 74%|███████▎  | 200/272 [00:43<00:15,  4.64it/s]INFO:xpert:| epoch  18 | 200/272 batches | lr 0.000000 | ms/batch 215.64 | loss 0.91027 | mse 0.91027 |\n",
            "100%|██████████| 272/272 [00:58<00:00,  4.64it/s]\n",
            "INFO:xpert:-----------------------------------------------------------------------------------------\n",
            "INFO:xpert:| end of epoch  18 | time: 60.39s | valid loss/mse 1.0067 |\n",
            "INFO:xpert:-----------------------------------------------------------------------------------------\n",
            " 37%|███▋      | 100/272 [00:21<00:37,  4.63it/s]INFO:xpert:| epoch  19 | 100/272 batches | lr 0.000000 | ms/batch 218.07 | loss 0.92397 | mse 0.92397 |\n",
            " 74%|███████▎  | 200/272 [00:43<00:15,  4.62it/s]INFO:xpert:| epoch  19 | 200/272 batches | lr 0.000000 | ms/batch 216.18 | loss 0.89055 | mse 0.89055 |\n",
            "100%|██████████| 272/272 [00:58<00:00,  4.64it/s]\n",
            "INFO:xpert:-----------------------------------------------------------------------------------------\n",
            "INFO:xpert:| end of epoch  19 | time: 60.42s | valid loss/mse 1.0067 |\n",
            "INFO:xpert:-----------------------------------------------------------------------------------------\n",
            " 37%|███▋      | 100/272 [00:21<00:37,  4.63it/s]INFO:xpert:| epoch  20 | 100/272 batches | lr 0.000000 | ms/batch 217.85 | loss 0.92940 | mse 0.92940 |\n",
            " 74%|███████▎  | 200/272 [00:43<00:15,  4.64it/s]INFO:xpert:| epoch  20 | 200/272 batches | lr 0.000000 | ms/batch 214.97 | loss 0.87172 | mse 0.87172 |\n",
            "100%|██████████| 272/272 [00:58<00:00,  4.65it/s]\n",
            "INFO:xpert:-----------------------------------------------------------------------------------------\n",
            "INFO:xpert:| end of epoch  20 | time: 60.27s | valid loss/mse 1.0068 |\n",
            "INFO:xpert:-----------------------------------------------------------------------------------------\n"
          ]
        }
      ],
      "source": [
        "best_val_loss = float('inf')\n",
        "best_model = None\n",
        "patience = 0\n",
        "early_stop = epochs\n",
        "\n",
        "train_metrics_list, train_metrics_pert_list = [], []\n",
        "val_metrics_list, val_metrics_pert_list = [], []\n",
        "\n",
        "for epoch in range(1, epochs + 1):\n",
        "    epoch_start_time = time.time()\n",
        "    train_loader = pert_data.dataloader[\"train_loader\"]\n",
        "    valid_loader = pert_data.dataloader[\"val_loader\"]\n",
        "\n",
        "    train(model, train_loader)\n",
        "    val_loss, val_mre = evaluate(model, valid_loader)\n",
        "    elapsed = time.time() - epoch_start_time\n",
        "    logger.info(\"-\" * 89)\n",
        "    logger.info(f\"| end of epoch {epoch:3d} | time: {elapsed:5.2f}s | valid loss/mse {val_loss:5.4f} |\")\n",
        "    logger.info(\"-\" * 89)\n",
        "\n",
        "    if val_loss < best_val_loss:\n",
        "        best_val_loss = val_loss\n",
        "        best_model = copy.deepcopy(model)\n",
        "        logger.info(f\"Best model with score {best_val_loss:5.4f}\")\n",
        "        patience = 0\n",
        "        torch.save(best_model.state_dict(), save_root / \"model_best.pt\")\n",
        "    else:\n",
        "        patience += 1\n",
        "        if patience >= early_stop:\n",
        "            logger.info(f\"Early stop at epoch {epoch}\")\n",
        "            break\n",
        "\n",
        "    if np.isnan(val_loss):\n",
        "        logger.warning(f\"NaN loss detected at epoch {epoch}, stopping training\")\n",
        "        break\n",
        "\n",
        "    # Optional metric collection (costly)\n",
        "    train_res = eval_perturb_new(train_loader, model, device)\n",
        "    val_res = eval_perturb_new(valid_loader, model, device)\n",
        "    if train_res is not None:\n",
        "        train_metrics, train_metrics_pert = compute_metrics(train_res)\n",
        "        train_metrics_list.append(train_metrics)\n",
        "        train_metrics_pert_list.append(train_metrics_pert)\n",
        "    if val_res is not None:\n",
        "        val_metrics, val_metrics_pert = compute_metrics(val_res)\n",
        "        val_metrics_list.append(val_metrics)\n",
        "        val_metrics_pert_list.append(val_metrics_pert)\n",
        "\n",
        "    if scheduler_type == 'steplr':\n",
        "        scheduler.step()\n",
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "## 11. Final Evaluation and Save Artifacts\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 17,
      "metadata": {},
      "outputs": [
        {
          "data": {
            "image/png": "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",
            "text/plain": [
              "<Figure size 1000x800 with 4 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        },
        {
          "data": {
            "image/png": "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",
            "text/plain": [
              "<Figure size 1000x800 with 4 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        }
      ],
      "source": [
        "# Save best model and evaluate on test set\n",
        "torch.save(best_model.state_dict(), save_root / \"model_best.pt\")\n",
        "\n",
        "test_loader = pert_data.dataloader[\"test_loader\"]\n",
        "\n",
        "test_res = eval_perturb_new(test_loader, best_model, device)\n",
        "\n",
        "# Save results\n",
        "model_prefix = f'lr_{lr}_cotrain_mode'\n",
        "result_dir = save_root / f'result_{model_prefix}'\n",
        "result_dir.mkdir(parents=True, exist_ok=True)\n",
        "\n",
        "import pickle\n",
        "pickle.dump(test_res, open(result_dir / 'test_res.pkl', 'wb'))\n",
        "if len(train_metrics_list) > 0:\n",
        "    pickle.dump(train_metrics_list, open(result_dir / 'train_metrics_list.pkl', 'wb'))\n",
        "    pickle.dump(train_metrics_pert_list, open(result_dir / 'train_metrics_pert_list.pkl', 'wb'))\n",
        "    pickle.dump(val_metrics_list, open(result_dir / 'val_metrics_list.pkl', 'wb'))\n",
        "    pickle.dump(val_metrics_pert_list, open(result_dir / 'val_metrics_pert_list.pkl', 'wb'))\n",
        "\n",
        "# Optional plotting if metrics were collected\n",
        "if len(train_metrics_list) > 0:\n",
        "    try:\n",
        "        merge_plot(train_metrics_list, 'train', str(result_dir / 'train.png'))\n",
        "        merge_plot(val_metrics_list, 'val', str(result_dir / 'val.png'))\n",
        "    except Exception as e:\n",
        "        logger.warning(f\"Plotting skipped: {e}\")\n",
        "\n",
        "# Free CUDA cache\n",
        "torch.cuda.empty_cache()\n",
        "\n"
      ]
    }
  ],
  "metadata": {
    "kernelspec": {
      "display_name": "scGPT_2",
      "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": "3.10.14"
    }
  },
  "nbformat": 4,
  "nbformat_minor": 2
}