tutorials/FinetuneData_GSE135194BW_random_context.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "e48c5f86-f0aa-4963-8913-604443fa0a3c",
   "metadata": {},
   "source": [
    "# Tutorial: Finetune pretrained model using single cell data with addtional genes and prediction catagories"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5aac073f-ef61-49f9-bf3e-13b2a3137121",
   "metadata": {},
   "source": [
    "## Introduction\n",
    "\n",
    "This tutorial demonstrates how to use **scGenAI** to fine-tune a model for predicting cell types using a new single-cell RNA sequencing (scRNA-seq) dataset that contains additional genes and cell types. In Fine-tune mode, the model template and context method will be automatically adopted from the pretrained model specified in the configuration file.\n",
    "\n",
    "\n",
    "**Note:**\n",
    "1. The dataset used for this tutorial is reported in [**Hu et. al., 2021**](https://pubmed.ncbi.nlm.nih.gov/32556286/). The input data train and validation files can be downloaded from link of [**GSE135194**](https://zenodo.org/records/14036452).\n",
    "\n",
    "\n",
    "**Reference**\n",
    "1. Wu, Z., Gao, S., Diamond, C., Kajigaya, S., Chen, J., Shi, R., Palmer, C., Hsu, A. P., Calvo, K. R., Hickstein, D. D., Holland, S. M., & Young, N. S. (2020). Sequencing of RNA in single cells reveals a distinct transcriptome signature of hematopoiesis in GATA2 deficiency. Blood advances, 4(12), 2656–2670. https://doi.org/10.1182/bloodadvances.2019001352\n",
    "\n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8294154a-4036-4788-b358-0e91342ac224",
   "metadata": {},
   "source": [
    "## Fine-tune the Model (Multiple GPU)\n",
    "\n",
    "The training process can be easily initiated by calling the `run_finetune_from_config` function in **scGenAI**, using the input files specified in a configuration YAML file. Example template YAML files are available in the `scgenai/examples/config_templates` folder of the package. These templates provide the necessary structure for defining training parameters, model configurations, and input data locations.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "2989a022-6bea-4f9c-861d-452b0d85e11f",
   "metadata": {},
   "outputs": [],
   "source": [
    "from scGenAI.finetuning.finetune import run_finetune_from_config"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "5a140f4c-6d3c-4a18-9f15-87d148179ac8",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "model_dir: /home/ubuntu/fulldataset/model/GSE181989BW_random_context/best_model/\n",
      "finetune_dir: /home/ubuntu/fulldataset/model/Finetune_GSE135194BW/\n",
      "train_file: /home/ubuntu/fulldataset/GSE135194_train.h5ad\n",
      "val_file: /home/ubuntu/fulldataset/BoneMarrowRefine2_val.h5ad\n",
      "log_dir: /home/ubuntu/scgenai/examples/logs/\n",
      "output_dir: /home/ubuntu/scgenai/examples/output\n",
      "seed: 1314521\n",
      "cache_dir: /home/ubuntu/scgenai/examples/tmp/cache\n",
      "mode: Finetune\n",
      "min_cells: 50\n",
      "target_feature: ct\n",
      "multiomics: No\n",
      "keepIntermediateFiles: Yes\n",
      "savelog: Yes\n",
      "savesetting: Yes\n",
      "model_backbone_name: llama\n",
      "model_backbone_size: small\n",
      "context_method: random\n",
      "num_bins: 10\n",
      "optimizer: AdamW\n",
      "weight_decay: 0.01\n",
      "depth: 2\n",
      "max_length: 3072\n",
      "batch_size: 1\n",
      "learning_rate: 1e-05\n",
      "num_epochs: 30\n",
      "world_size: 1\n",
      "evaluate_during_training: True\n",
      "hidden_size: 768\n",
      "num_layers: 12\n",
      "num_heads: 12\n",
      "intermediate_size: 1024\n",
      "stride: 1536\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\u001b[32m2024-11-01 03:28:36.647\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.finetuning.finetune\u001b[0m:\u001b[36mFinetunePreprocessing\u001b[0m:\u001b[36m68\u001b[0m - \u001b[1mNew genes found: 3473. Updating custom_tokenizer.\u001b[0m\n",
      "\u001b[32m2024-11-01 03:28:36.657\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.finetuning.finetune\u001b[0m:\u001b[36mFinetunePreprocessing\u001b[0m:\u001b[36m79\u001b[0m - \u001b[1mNew target labels found: 2. Updating label encoder.\u001b[0m\n",
      "2024-11-01 03:31:14.929 | INFO     | scGenAI.utils.distributed:setup_distributed:26 - Distributed setup complete for GPU 0, Total GPUs used: 4\n",
      "2024-11-01 03:31:14.938 | INFO     | scGenAI.utils.distributed:setup_distributed:26 - Distributed setup complete for GPU 3, Total GPUs used: 4\n",
      "2024-11-01 03:31:14.943 | INFO     | scGenAI.utils.distributed:setup_distributed:26 - Distributed setup complete for GPU 1, Total GPUs used: 4\n",
      "2024-11-01 03:31:14.945 | INFO     | scGenAI.utils.distributed:setup_distributed:26 - Distributed setup complete for GPU 2, Total GPUs used: 4\n",
      "2024-11-01 03:31:17.651 | INFO     | scGenAI.finetuning.finetune:model_finetune_and_eval:132 - Total number of trainable parameters: 82198100; Max Sequence Legth: 1536 genes\n",
      "Epoch 1/30: 100%|██████████| 9246/9246 [1:01:25<00:00,  2.51it/s, GPU_ID=3, train_accuracy=0.672, train_loss=0.964]\n",
      "Epoch 1/30: 100%|██████████| 9246/9246 [1:01:25<00:00,  2.51it/s, GPU_ID=2, train_accuracy=0.674, train_loss=0.974]\n",
      "\n",
      "Epoch 1/30: 100%|██████████| 9246/9246 [1:01:25<00:00,  2.51it/s, GPU_ID=1, train_accuracy=0.666, train_loss=0.988]\n",
      "Evaluating Epoch 1/30: 100%|██████████| 3993/3993 [07:19<00:00,  9.08it/s]\n",
      "Evaluating Epoch 1/30: 100%|██████████| 3993/3993 [07:23<00:00,  8.99it/s]\n",
      "Evaluating Epoch 1/30: 100%|██████████| 3993/3993 [07:28<00:00,  8.90it/s]\n",
      "Evaluating Epoch 1/30: 100%|██████████| 3993/3993 [07:29<00:00,  8.87it/s]\n",
      "2024-11-01 04:40:13.939 | INFO     | scGenAI.training.train:evaluate_and_save_predictions:328 - Epoch 1/30, Train Loss: 0.9657, Train Accuracy: 0.6723, Context-level Val Loss: 1.3063, Context-level Val Accuracy: 0.7022\n",
      "Epoch 2/30:   0%|          | 0/9246 [00:00<?, ?it/s]2024-11-01 04:40:18.523 | INFO     | scGenAI.utils.distributed:savemodel:54 - Model state saved to /home/ubuntu/fulldataset/model/Finetune_GSE135194BW/best_model/scGenAI_model.pt\n",
      "2024-11-01 04:40:18.659 | INFO     | scGenAI.utils.distributed:savemodel:59 - Tokenizer saved to /home/ubuntu/fulldataset/model/Finetune_GSE135194BW/best_model\n",
      "2024-11-01 04:40:18.664 | INFO     | scGenAI.utils.distributed:savemodel:75 - Custom vocab and label encoder saved\n",
      "2024-11-01 04:40:18.665 | INFO     | scGenAI.utils.distributed:savemodel:81 - Model configuration saved to /home/ubuntu/fulldataset/model/Finetune_GSE135194BW/best_model\n",
      "Epoch 2/30: 100%|██████████| 9246/9246 [1:01:33<00:00,  2.50it/s, GPU_ID=3, train_accuracy=0.831, train_loss=0.493] \n",
      "\n",
      "Epoch 2/30: 100%|██████████| 9246/9246 [1:01:27<00:00,  2.51it/s, GPU_ID=0, train_accuracy=0.834, train_loss=0.477]\n",
      "Epoch 2/30: 100%|██████████| 9246/9246 [1:01:28<00:00,  2.51it/s, GPU_ID=1, train_accuracy=0.828, train_loss=0.498]\n",
      "Evaluating Epoch 2/30: 100%|██████████| 3993/3993 [07:18<00:00,  9.10it/s]\n",
      "Evaluating Epoch 2/30: 100%|██████████| 3993/3993 [07:23<00:00,  9.01it/s]\n",
      "Evaluating Epoch 2/30: 100%|██████████| 3993/3993 [07:28<00:00,  8.91it/s]\n",
      "Evaluating Epoch 2/30: 100%|██████████| 3993/3993 [07:29<00:00,  8.89it/s]\n",
      "2024-11-01 05:49:15.130 | INFO     | scGenAI.training.train:evaluate_and_save_predictions:328 - Epoch 2/30, Train Loss: 0.4771, Train Accuracy: 0.8335, Context-level Val Loss: 1.2320, Context-level Val Accuracy: 0.7688\n",
      "Epoch 3/30:   0%|          | 0/9246 [00:00<?, ?it/s]2024-11-01 05:49:21.340 | INFO     | scGenAI.utils.distributed:savemodel:54 - Model state saved to /home/ubuntu/fulldataset/model/Finetune_GSE135194BW/best_model/scGenAI_model.pt\n",
      "2024-11-01 05:49:21.490 | INFO     | scGenAI.utils.distributed:savemodel:59 - Tokenizer saved to /home/ubuntu/fulldataset/model/Finetune_GSE135194BW/best_model\n",
      "2024-11-01 05:49:21.494 | INFO     | scGenAI.utils.distributed:savemodel:75 - Custom vocab and label encoder saved\n",
      "2024-11-01 05:49:21.495 | INFO     | scGenAI.utils.distributed:savemodel:81 - Model configuration saved to /home/ubuntu/fulldataset/model/Finetune_GSE135194BW/best_model\n",
      "Epoch 3/30: 100%|██████████| 9246/9246 [1:01:30<00:00,  2.51it/s, GPU_ID=3, train_accuracy=0.895, train_loss=0.315] \n",
      "Epoch 3/30: 100%|██████████| 9246/9246 [1:01:34<00:00,  2.50it/s, GPU_ID=2, train_accuracy=0.9, train_loss=0.288]\n",
      "Epoch 3/30: 100%|██████████| 9246/9246 [1:01:22<00:00,  2.51it/s, GPU_ID=0, train_accuracy=0.898, train_loss=0.29]\n",
      "Epoch 3/30: 100%|██████████| 9246/9246 [1:01:25<00:00,  2.51it/s, GPU_ID=1, train_accuracy=0.895, train_loss=0.306]\n",
      "Evaluating Epoch 3/30: 100%|██████████| 3993/3993 [07:18<00:00,  9.10it/s]\n",
      "Evaluating Epoch 3/30: 100%|██████████| 3993/3993 [07:22<00:00,  9.02it/s]\n",
      "Evaluating Epoch 3/30: 100%|██████████| 3993/3993 [07:27<00:00,  8.92it/s]\n",
      "Evaluating Epoch 3/30: 100%|██████████| 3993/3993 [07:27<00:00,  8.92it/s]\n",
      "2024-11-01 06:58:11.161 | INFO     | scGenAI.training.train:evaluate_and_save_predictions:328 - Epoch 3/30, Train Loss: 0.2903, Train Accuracy: 0.8984, Context-level Val Loss: 1.2491, Context-level Val Accuracy: 0.7919\n",
      "Epoch 4/30:   0%|          | 0/9246 [00:00<?, ?it/s]2024-11-01 06:58:17.380 | INFO     | scGenAI.utils.distributed:savemodel:54 - Model state saved to /home/ubuntu/fulldataset/model/Finetune_GSE135194BW/best_model/scGenAI_model.pt\n",
      "2024-11-01 06:58:17.531 | INFO     | scGenAI.utils.distributed:savemodel:59 - Tokenizer saved to /home/ubuntu/fulldataset/model/Finetune_GSE135194BW/best_model\n",
      "2024-11-01 06:58:17.536 | INFO     | scGenAI.utils.distributed:savemodel:75 - Custom vocab and label encoder saved\n",
      "2024-11-01 06:58:17.537 | INFO     | scGenAI.utils.distributed:savemodel:81 - Model configuration saved to /home/ubuntu/fulldataset/model/Finetune_GSE135194BW/best_model\n",
      "Epoch 4/30: 100%|██████████| 9246/9246 [1:01:29<00:00,  2.51it/s, GPU_ID=3, train_accuracy=0.938, train_loss=0.189] \n",
      "Epoch 4/30: 100%|██████████| 9246/9246 [1:01:22<00:00,  2.51it/s, GPU_ID=0, train_accuracy=0.939, train_loss=0.178]\n",
      "Epoch 4/30: 100%|██████████| 9246/9246 [1:01:33<00:00,  2.50it/s, GPU_ID=2, train_accuracy=0.942, train_loss=0.174]\n",
      "Epoch 4/30: 100%|██████████| 9246/9246 [1:01:24<00:00,  2.51it/s, GPU_ID=1, train_accuracy=0.937, train_loss=0.184]\n",
      "Evaluating Epoch 4/30: 100%|██████████| 3993/3993 [07:18<00:00,  9.10it/s]\n",
      "Evaluating Epoch 4/30: 100%|██████████| 3993/3993 [07:22<00:00,  9.02it/s]\n",
      "Evaluating Epoch 4/30: 100%|██████████| 3993/3993 [07:27<00:00,  8.92it/s]\n",
      "Evaluating Epoch 4/30: 100%|██████████| 3993/3993 [07:28<00:00,  8.91it/s]\n",
      "2024-11-01 08:07:08.034 | INFO     | scGenAI.training.train:evaluate_and_save_predictions:328 - Epoch 4/30, Train Loss: 0.1785, Train Accuracy: 0.9389, Context-level Val Loss: 1.2771, Context-level Val Accuracy: 0.7986\n",
      "Epoch 5/30:   0%|          | 0/9246 [00:00<?, ?it/s]2024-11-01 08:07:14.299 | INFO     | scGenAI.utils.distributed:savemodel:54 - Model state saved to /home/ubuntu/fulldataset/model/Finetune_GSE135194BW/best_model/scGenAI_model.pt\n",
      "2024-11-01 08:07:14.452 | INFO     | scGenAI.utils.distributed:savemodel:59 - Tokenizer saved to /home/ubuntu/fulldataset/model/Finetune_GSE135194BW/best_model\n",
      "2024-11-01 08:07:14.457 | INFO     | scGenAI.utils.distributed:savemodel:75 - Custom vocab and label encoder saved\n",
      "2024-11-01 08:07:14.458 | INFO     | scGenAI.utils.distributed:savemodel:81 - Model configuration saved to /home/ubuntu/fulldataset/model/Finetune_GSE135194BW/best_model\n",
      "Epoch 5/30: 100%|██████████| 9246/9246 [1:01:28<00:00,  2.51it/s, GPU_ID=3, train_accuracy=0.962, train_loss=0.122]  \n",
      "Epoch 5/30: 100%|██████████| 9246/9246 [1:01:32<00:00,  2.50it/s, GPU_ID=2, train_accuracy=0.962, train_loss=0.119]\n",
      "Epoch 5/30: 100%|██████████| 9246/9246 [1:01:23<00:00,  2.51it/s, GPU_ID=1, train_accuracy=0.961, train_loss=0.119]\n",
      "Epoch 5/30: 100%|██████████| 9246/9246 [1:01:20<00:00,  2.51it/s, GPU_ID=0, train_accuracy=0.961, train_loss=0.12]\n",
      "Evaluating Epoch 5/30: 100%|██████████| 3993/3993 [07:18<00:00,  9.11it/s]\n",
      "Evaluating Epoch 5/30: 100%|██████████| 3993/3993 [07:23<00:00,  9.01it/s]\n",
      "Evaluating Epoch 5/30: 100%|██████████| 3993/3993 [07:27<00:00,  8.92it/s]\n",
      "Evaluating Epoch 5/30: 100%|██████████| 3993/3993 [07:28<00:00,  8.91it/s]\n",
      "2024-11-01 09:16:03.024 | INFO     | scGenAI.training.train:evaluate_and_save_predictions:328 - Epoch 5/30, Train Loss: 0.1200, Train Accuracy: 0.9606, Context-level Val Loss: 1.4024, Context-level Val Accuracy: 0.8042\n",
      "Epoch 6/30:   0%|          | 0/9246 [00:00<?, ?it/s]2024-11-01 09:16:09.251 | INFO     | scGenAI.utils.distributed:savemodel:54 - Model state saved to /home/ubuntu/fulldataset/model/Finetune_GSE135194BW/best_model/scGenAI_model.pt\n",
      "2024-11-01 09:16:09.408 | INFO     | scGenAI.utils.distributed:savemodel:59 - Tokenizer saved to /home/ubuntu/fulldataset/model/Finetune_GSE135194BW/best_model\n",
      "2024-11-01 09:16:09.413 | INFO     | scGenAI.utils.distributed:savemodel:75 - Custom vocab and label encoder saved\n",
      "2024-11-01 09:16:09.414 | INFO     | scGenAI.utils.distributed:savemodel:81 - Model configuration saved to /home/ubuntu/fulldataset/model/Finetune_GSE135194BW/best_model\n",
      "Epoch 6/30: 100%|██████████| 9246/9246 [1:01:27<00:00,  2.51it/s, GPU_ID=3, train_accuracy=0.972, train_loss=0.0861] \n",
      "Epoch 6/30: 100%|██████████| 9246/9246 [1:01:31<00:00,  2.50it/s, GPU_ID=2, train_accuracy=0.974, train_loss=0.0824]\n",
      "Epoch 6/30: 100%|██████████| 9246/9246 [1:01:20<00:00,  2.51it/s, GPU_ID=0, train_accuracy=0.972, train_loss=0.0894]\n",
      "Epoch 6/30: 100%|██████████| 9246/9246 [1:01:23<00:00,  2.51it/s, GPU_ID=1, train_accuracy=0.974, train_loss=0.0809]\n",
      "Evaluating Epoch 6/30: 100%|██████████| 3993/3993 [07:18<00:00,  9.11it/s]\n",
      "Evaluating Epoch 6/30: 100%|██████████| 3993/3993 [07:22<00:00,  9.03it/s]\n",
      "Evaluating Epoch 6/30: 100%|██████████| 3993/3993 [07:26<00:00,  8.94it/s]\n",
      "Evaluating Epoch 6/30: 100%|██████████| 3993/3993 [07:27<00:00,  8.93it/s]\n",
      "2024-11-01 10:24:56.843 | INFO     | scGenAI.training.train:evaluate_and_save_predictions:328 - Epoch 6/30, Train Loss: 0.0894, Train Accuracy: 0.9719, Context-level Val Loss: 1.4609, Context-level Val Accuracy: 0.8029\n",
      "Epoch 7/30: 100%|██████████| 9246/9246 [1:01:25<00:00,  2.51it/s, GPU_ID=3, train_accuracy=0.978, train_loss=0.0692] \n",
      "Epoch 7/30: 100%|██████████| 9246/9246 [1:01:28<00:00,  2.51it/s, GPU_ID=2, train_accuracy=0.981, train_loss=0.0575]\n",
      "\n",
      "Epoch 7/30: 100%|██████████| 9246/9246 [1:01:20<00:00,  2.51it/s, GPU_ID=1, train_accuracy=0.978, train_loss=0.0674]\n",
      "Evaluating Epoch 7/30: 100%|██████████| 3993/3993 [07:18<00:00,  9.11it/s]\n",
      "Evaluating Epoch 7/30: 100%|██████████| 3993/3993 [07:21<00:00,  9.04it/s]\n",
      "Evaluating Epoch 7/30: 100%|██████████| 3993/3993 [07:27<00:00,  8.93it/s]\n",
      "2024-11-01 11:33:48.207 | INFO     | scGenAI.training.train:evaluate_and_save_predictions:328 - Epoch 7/30, Train Loss: 0.0594, Train Accuracy: 0.9804, Context-level Val Loss: 1.6868, Context-level Val Accuracy: 0.7921\n",
      "Evaluating Epoch 7/30: 100%|██████████| 3993/3993 [07:27<00:00,  8.92it/s]\n",
      "Epoch 8/30: 100%|██████████| 9246/9246 [1:01:24<00:00,  2.51it/s, GPU_ID=3, train_accuracy=0.984, train_loss=0.0514] \n",
      "Epoch 8/30: 100%|██████████| 9246/9246 [1:01:19<00:00,  2.51it/s, GPU_ID=0, train_accuracy=0.984, train_loss=0.0484]\n",
      "Epoch 8/30: 100%|██████████| 9246/9246 [1:01:28<00:00,  2.51it/s, GPU_ID=2, train_accuracy=0.984, train_loss=0.051]\n",
      "Epoch 8/30: 100%|██████████| 9246/9246 [1:01:19<00:00,  2.51it/s, GPU_ID=1, train_accuracy=0.983, train_loss=0.0582]\n",
      "Evaluating Epoch 8/30: 100%|██████████| 3993/3993 [07:17<00:00,  9.12it/s]\n",
      "Evaluating Epoch 8/30: 100%|██████████| 3993/3993 [07:22<00:00,  9.02it/s]\n",
      "Evaluating Epoch 8/30: 100%|██████████| 3993/3993 [07:26<00:00,  8.94it/s]\n",
      "Evaluating Epoch 8/30: 100%|██████████| 3993/3993 [07:26<00:00,  8.94it/s]\n",
      "2024-11-01 12:42:38.503 | INFO     | scGenAI.training.train:evaluate_and_save_predictions:328 - Epoch 8/30, Train Loss: 0.0484, Train Accuracy: 0.9839, Context-level Val Loss: 1.5749, Context-level Val Accuracy: 0.8089\n",
      "Epoch 9/30:   0%|          | 0/9246 [00:00<?, ?it/s]2024-11-01 12:42:44.750 | INFO     | scGenAI.utils.distributed:savemodel:54 - Model state saved to /home/ubuntu/fulldataset/model/Finetune_GSE135194BW/best_model/scGenAI_model.pt\n",
      "2024-11-01 12:42:44.905 | INFO     | scGenAI.utils.distributed:savemodel:59 - Tokenizer saved to /home/ubuntu/fulldataset/model/Finetune_GSE135194BW/best_model\n",
      "2024-11-01 12:42:44.909 | INFO     | scGenAI.utils.distributed:savemodel:75 - Custom vocab and label encoder saved\n",
      "2024-11-01 12:42:44.910 | INFO     | scGenAI.utils.distributed:savemodel:81 - Model configuration saved to /home/ubuntu/fulldataset/model/Finetune_GSE135194BW/best_model\n",
      "Epoch 9/30: 100%|██████████| 9246/9246 [1:01:27<00:00,  2.51it/s, GPU_ID=3, train_accuracy=0.988, train_loss=0.0426] \n",
      "Epoch 9/30: 100%|██████████| 9246/9246 [1:01:21<00:00,  2.51it/s, GPU_ID=0, train_accuracy=0.989, train_loss=0.0384]\n",
      "\n",
      "Epoch 9/30: 100%|██████████| 9246/9246 [1:01:23<00:00,  2.51it/s, GPU_ID=1, train_accuracy=0.987, train_loss=0.0429]\n",
      "Evaluating Epoch 9/30: 100%|██████████| 3993/3993 [07:18<00:00,  9.10it/s]\n",
      "Evaluating Epoch 9/30: 100%|██████████| 3993/3993 [07:22<00:00,  9.02it/s]\n",
      "Evaluating Epoch 9/30: 100%|██████████| 3993/3993 [07:27<00:00,  8.93it/s]\n",
      "Evaluating Epoch 9/30: 100%|██████████| 3993/3993 [07:27<00:00,  8.92it/s]\n",
      "2024-11-01 13:51:33.948 | INFO     | scGenAI.training.train:evaluate_and_save_predictions:328 - Epoch 9/30, Train Loss: 0.0384, Train Accuracy: 0.9889, Context-level Val Loss: 1.7360, Context-level Val Accuracy: 0.8054\n",
      "Epoch 10/30: 100%|██████████| 9246/9246 [1:01:27<00:00,  2.51it/s, GPU_ID=3, train_accuracy=0.988, train_loss=0.0457] \n",
      "Epoch 10/30: 100%|██████████| 9246/9246 [1:01:22<00:00,  2.51it/s, GPU_ID=0, train_accuracy=0.992, train_loss=0.0259]\n",
      "\n",
      "Epoch 10/30: 100%|██████████| 9246/9246 [1:01:30<00:00,  2.51it/s, GPU_ID=2, train_accuracy=0.99, train_loss=0.0349]\n",
      "Evaluating Epoch 10/30: 100%|██████████| 3993/3993 [07:19<00:00,  9.08it/s]\n",
      "Evaluating Epoch 10/30: 100%|██████████| 3993/3993 [07:24<00:00,  8.98it/s]\n",
      "Evaluating Epoch 10/30: 100%|██████████| 3993/3993 [07:28<00:00,  8.89it/s]\n",
      "Evaluating Epoch 10/30: 100%|██████████| 3993/3993 [07:29<00:00,  8.87it/s]\n",
      "2024-11-01 15:00:30.131 | INFO     | scGenAI.training.train:evaluate_and_save_predictions:328 - Epoch 10/30, Train Loss: 0.0259, Train Accuracy: 0.9920, Context-level Val Loss: 1.7598, Context-level Val Accuracy: 0.8074\n",
      "Epoch 11/30:  34%|███▍      | 3165/9246 [21:02<40:25,  2.51it/s, GPU_ID=0, train_accuracy=0.991, train_loss=0.013]]]  IOPub data rate exceeded.\n",
      "The Jupyter server will temporarily stop sending output\n",
      "to the client in order to avoid crashing it.\n",
      "To change this limit, set the config variable\n",
      "`--ServerApp.iopub_data_rate_limit`.\n",
      "\n",
      "Current values:\n",
      "ServerApp.iopub_data_rate_limit=1000000.0 (bytes/sec)\n",
      "ServerApp.rate_limit_window=3.0 (secs)\n",
      "\n",
      "Epoch 11/30: 100%|██████████| 9246/9246 [1:01:29<00:00,  2.51it/s, GPU_ID=1, train_accuracy=0.989, train_loss=0.0327]\n",
      "Epoch 11/30: 100%|██████████| 9246/9246 [1:01:28<00:00,  2.51it/s, GPU_ID=0, train_accuracy=0.992, train_loss=0.029]]\n",
      "Epoch 11/30: 100%|██████████| 9246/9246 [1:01:38<00:00,  2.50it/s, GPU_ID=2, train_accuracy=0.992, train_loss=0.0311]\n",
      "\n",
      "Epoch 12/30: 100%|██████████| 9246/9246 [1:01:46<00:00,  2.49it/s, GPU_ID=2, train_accuracy=0.992, train_loss=0.0244]\n",
      "Epoch 12/30: 100%|██████████| 9246/9246 [1:01:42<00:00,  2.50it/s, GPU_ID=3, train_accuracy=0.993, train_loss=0.0232]\n",
      "Epoch 12/30: 100%|██████████| 9246/9246 [1:01:37<00:00,  2.50it/s, GPU_ID=1, train_accuracy=0.991, train_loss=0.0253]\n",
      "Epoch 12/30: 100%|██████████| 9246/9246 [1:01:35<00:00,  2.50it/s, GPU_ID=0, train_accuracy=0.991, train_loss=0.031]\n",
      "Evaluating Epoch 12/30: 100%|██████████| 3993/3993 [07:22<00:00,  9.03it/s]\n",
      "Evaluating Epoch 12/30: 100%|██████████| 3993/3993 [07:25<00:00,  8.97it/s]\n",
      "Evaluating Epoch 12/30: 100%|██████████| 3993/3993 [07:31<00:00,  8.85it/s]\n",
      "Evaluating Epoch 12/30: 100%|██████████| 3993/3993 [07:33<00:00,  8.81it/s]\n",
      "2024-11-01 17:18:47.739 | INFO     | scGenAI.training.train:evaluate_and_save_predictions:328 - Epoch 12/30, Train Loss: 0.0310, Train Accuracy: 0.9907, Context-level Val Loss: 1.8777, Context-level Val Accuracy: 0.8087\n",
      "Epoch 13/30: 100%|██████████| 9246/9246 [1:02:00<00:00,  2.49it/s, GPU_ID=3, train_accuracy=0.994, train_loss=0.0225] \n",
      "Epoch 13/30: 100%|██████████| 9246/9246 [1:01:51<00:00,  2.49it/s, GPU_ID=0, train_accuracy=0.993, train_loss=0.0268]\n",
      "Epoch 13/30: 100%|██████████| 9246/9246 [1:01:54<00:00,  2.49it/s, GPU_ID=1, train_accuracy=0.993, train_loss=0.0245]\n",
      "\n",
      "Evaluating Epoch 13/30: 100%|██████████| 3993/3993 [07:24<00:00,  8.98it/s]\n",
      "Evaluating Epoch 13/30: 100%|██████████| 3993/3993 [07:27<00:00,  8.92it/s]\n",
      "Evaluating Epoch 13/30: 100%|██████████| 3993/3993 [07:32<00:00,  8.82it/s]\n",
      "Evaluating Epoch 13/30: 100%|██████████| 3993/3993 [07:38<00:00,  8.72it/s]\n",
      "2024-11-01 18:28:22.043 | INFO     | scGenAI.training.train:evaluate_and_save_predictions:328 - Epoch 13/30, Train Loss: 0.0268, Train Accuracy: 0.9931, Context-level Val Loss: 1.7089, Context-level Val Accuracy: 0.8114\n",
      "2024-11-01 18:28:28.331 | INFO     | scGenAI.utils.distributed:savemodel:54 - Model state saved to /home/ubuntu/fulldataset/model/Finetune_GSE135194BW/best_model/scGenAI_model.pt\n",
      "2024-11-01 18:28:28.499 | INFO     | scGenAI.utils.distributed:savemodel:59 - Tokenizer saved to /home/ubuntu/fulldataset/model/Finetune_GSE135194BW/best_model\n",
      "2024-11-01 18:28:28.504 | INFO     | scGenAI.utils.distributed:savemodel:75 - Custom vocab and label encoder saved\n",
      "2024-11-01 18:28:28.505 | INFO     | scGenAI.utils.distributed:savemodel:81 - Model configuration saved to /home/ubuntu/fulldataset/model/Finetune_GSE135194BW/best_model\n",
      "Epoch 14/30: 100%|██████████| 9246/9246 [1:02:22<00:00,  2.47it/s, GPU_ID=2, train_accuracy=0.995, train_loss=0.0196] \n",
      "Epoch 14/30: 100%|██████████| 9246/9246 [1:02:19<00:00,  2.47it/s, GPU_ID=3, train_accuracy=0.994, train_loss=0.0218]\n",
      "Epoch 14/30: 100%|██████████| 9246/9246 [1:02:14<00:00,  2.48it/s, GPU_ID=1, train_accuracy=0.994, train_loss=0.0195]\n",
      "Epoch 14/30: 100%|██████████| 9246/9246 [1:02:06<00:00,  2.48it/s, GPU_ID=0, train_accuracy=0.994, train_loss=0.0212]\n",
      "Evaluating Epoch 14/30: 100%|██████████| 3993/3993 [07:24<00:00,  8.97it/s]\n",
      "Evaluating Epoch 14/30: 100%|██████████| 3993/3993 [07:28<00:00,  8.90it/s]\n",
      "Evaluating Epoch 14/30: 100%|██████████| 3993/3993 [07:33<00:00,  8.81it/s]\n",
      "Evaluating Epoch 14/30: 100%|██████████| 3993/3993 [07:39<00:00,  8.69it/s]\n",
      "2024-11-01 19:38:14.381 | INFO     | scGenAI.training.train:evaluate_and_save_predictions:328 - Epoch 14/30, Train Loss: 0.0212, Train Accuracy: 0.9942, Context-level Val Loss: 1.8707, Context-level Val Accuracy: 0.8107\n",
      "Epoch 15/30: 100%|██████████| 9246/9246 [1:02:06<00:00,  2.48it/s, GPU_ID=3, train_accuracy=0.996, train_loss=0.0142] \n",
      "Epoch 15/30: 100%|██████████| 9246/9246 [1:01:56<00:00,  2.49it/s, GPU_ID=0, train_accuracy=0.995, train_loss=0.0214]\n",
      "Epoch 15/30: 100%|██████████| 9246/9246 [1:02:10<00:00,  2.48it/s, GPU_ID=2, train_accuracy=0.995, train_loss=0.0173]\n",
      "Epoch 15/30: 100%|██████████| 9246/9246 [1:02:02<00:00,  2.48it/s, GPU_ID=1, train_accuracy=0.995, train_loss=0.0215]\n",
      "Evaluating Epoch 15/30: 100%|██████████| 3993/3993 [07:24<00:00,  8.99it/s]\n",
      "Evaluating Epoch 15/30: 100%|██████████| 3993/3993 [07:26<00:00,  8.94it/s]\n",
      "Evaluating Epoch 15/30: 100%|██████████| 3993/3993 [07:32<00:00,  8.83it/s]\n",
      "Evaluating Epoch 15/30: 100%|██████████| 3993/3993 [07:36<00:00,  8.75it/s]\n",
      "2024-11-01 20:47:51.030 | INFO     | scGenAI.training.train:evaluate_and_save_predictions:328 - Epoch 15/30, Train Loss: 0.0214, Train Accuracy: 0.9946, Context-level Val Loss: 1.7256, Context-level Val Accuracy: 0.8107\n",
      "Epoch 16/30: 100%|██████████| 9246/9246 [1:01:57<00:00,  2.49it/s, GPU_ID=3, train_accuracy=0.997, train_loss=0.0163] \n",
      "Epoch 16/30: 100%|██████████| 9246/9246 [1:01:48<00:00,  2.49it/s, GPU_ID=0, train_accuracy=0.995, train_loss=0.0142]\n",
      "\n",
      "Epoch 16/30: 100%|██████████| 9246/9246 [1:01:52<00:00,  2.49it/s, GPU_ID=1, train_accuracy=0.996, train_loss=0.0144]\n",
      "Evaluating Epoch 16/30: 100%|██████████| 3993/3993 [07:21<00:00,  9.04it/s]\n",
      "Evaluating Epoch 16/30: 100%|██████████| 3993/3993 [07:25<00:00,  8.97it/s]\n",
      "Evaluating Epoch 16/30: 100%|██████████| 3993/3993 [07:30<00:00,  8.86it/s]\n",
      "Evaluating Epoch 16/30: 100%|██████████| 3993/3993 [07:33<00:00,  8.80it/s]\n",
      "2024-11-01 21:57:16.876 | INFO     | scGenAI.training.train:evaluate_and_save_predictions:328 - Epoch 16/30, Train Loss: 0.0142, Train Accuracy: 0.9952, Context-level Val Loss: 1.8533, Context-level Val Accuracy: 0.8147\n",
      "Epoch 17/30:   0%|          | 0/9246 [00:00<?, ?it/s]2024-11-01 21:57:23.146 | INFO     | scGenAI.utils.distributed:savemodel:54 - Model state saved to /home/ubuntu/fulldataset/model/Finetune_GSE135194BW/best_model/scGenAI_model.pt\n",
      "2024-11-01 21:57:23.297 | INFO     | scGenAI.utils.distributed:savemodel:59 - Tokenizer saved to /home/ubuntu/fulldataset/model/Finetune_GSE135194BW/best_model\n",
      "2024-11-01 21:57:23.302 | INFO     | scGenAI.utils.distributed:savemodel:75 - Custom vocab and label encoder saved\n",
      "2024-11-01 21:57:23.303 | INFO     | scGenAI.utils.distributed:savemodel:81 - Model configuration saved to /home/ubuntu/fulldataset/model/Finetune_GSE135194BW/best_model\n",
      "Epoch 17/30: 100%|██████████| 9246/9246 [1:01:49<00:00,  2.49it/s, GPU_ID=2, train_accuracy=0.994, train_loss=0.0209] \n",
      "Epoch 17/30: 100%|██████████| 9246/9246 [1:01:46<00:00,  2.49it/s, GPU_ID=3, train_accuracy=0.997, train_loss=0.0126]\n",
      "\n",
      "Epoch 17/30: 100%|██████████| 9246/9246 [1:01:35<00:00,  2.50it/s, GPU_ID=0, train_accuracy=0.996, train_loss=0.015]\n",
      "Evaluating Epoch 17/30: 100%|██████████| 3993/3993 [07:19<00:00,  9.08it/s]\n",
      "Evaluating Epoch 17/30: 100%|██████████| 3993/3993 [07:23<00:00,  9.00it/s]\n",
      "Evaluating Epoch 17/30: 100%|██████████| 3993/3993 [07:27<00:00,  8.93it/s]\n",
      "Evaluating Epoch 17/30: 100%|██████████| 3993/3993 [07:30<00:00,  8.87it/s]\n",
      "2024-11-01 23:06:28.728 | INFO     | scGenAI.training.train:evaluate_and_save_predictions:328 - Epoch 17/30, Train Loss: 0.0150, Train Accuracy: 0.9959, Context-level Val Loss: 1.9526, Context-level Val Accuracy: 0.8119\n",
      "Epoch 18/30: 100%|██████████| 9246/9246 [1:01:32<00:00,  2.50it/s, GPU_ID=2, train_accuracy=0.996, train_loss=0.0123]]\n",
      "Epoch 18/30: 100%|██████████| 9246/9246 [1:01:28<00:00,  2.51it/s, GPU_ID=3, train_accuracy=0.997, train_loss=0.00962]\n",
      "\n",
      "Epoch 18/30: 100%|██████████| 9246/9246 [1:01:22<00:00,  2.51it/s, GPU_ID=0, train_accuracy=0.996, train_loss=0.0173]\n",
      "Evaluating Epoch 18/30: 100%|██████████| 3993/3993 [07:18<00:00,  9.11it/s]\n",
      "Evaluating Epoch 18/30: 100%|██████████| 3993/3993 [07:22<00:00,  9.02it/s]\n",
      "Evaluating Epoch 18/30: 100%|██████████| 3993/3993 [07:25<00:00,  8.96it/s]\n",
      "Evaluating Epoch 18/30: 100%|██████████| 3993/3993 [07:28<00:00,  8.91it/s]\n",
      "2024-11-02 00:15:23.438 | INFO     | scGenAI.training.train:evaluate_and_save_predictions:328 - Epoch 18/30, Train Loss: 0.0173, Train Accuracy: 0.9958, Context-level Val Loss: 2.2824, Context-level Val Accuracy: 0.7961\n",
      "Epoch 19/30: 100%|██████████| 9246/9246 [1:01:20<00:00,  2.51it/s, GPU_ID=3, train_accuracy=0.997, train_loss=0.0134] \n",
      "Epoch 19/30: 100%|██████████| 9246/9246 [1:01:24<00:00,  2.51it/s, GPU_ID=2, train_accuracy=0.996, train_loss=0.0148]\n",
      "Epoch 19/30: 100%|██████████| 9246/9246 [1:01:14<00:00,  2.52it/s, GPU_ID=0, train_accuracy=0.996, train_loss=0.0127]\n",
      "Epoch 19/30: 100%|██████████| 9246/9246 [1:01:17<00:00,  2.51it/s, GPU_ID=1, train_accuracy=0.996, train_loss=0.0152]\n",
      "Evaluating Epoch 19/30: 100%|██████████| 3993/3993 [07:17<00:00,  9.13it/s]\n",
      "Evaluating Epoch 19/30: 100%|██████████| 3993/3993 [07:22<00:00,  9.03it/s]\n",
      "Evaluating Epoch 19/30: 100%|██████████| 3993/3993 [07:25<00:00,  8.97it/s]\n",
      "Evaluating Epoch 19/30: 100%|██████████| 3993/3993 [07:25<00:00,  8.97it/s]\n",
      "2024-11-02 01:24:07.685 | INFO     | scGenAI.training.train:evaluate_and_save_predictions:328 - Epoch 19/30, Train Loss: 0.0127, Train Accuracy: 0.9964, Context-level Val Loss: 2.0420, Context-level Val Accuracy: 0.8054\n",
      "Epoch 20/30: 100%|██████████| 9246/9246 [1:01:08<00:00,  2.52it/s, GPU_ID=3, train_accuracy=0.997, train_loss=0.0115] \n",
      "Epoch 20/30: 100%|██████████| 9246/9246 [1:01:05<00:00,  2.52it/s, GPU_ID=0, train_accuracy=0.996, train_loss=0.0126]\n",
      "Epoch 20/30: 100%|██████████| 9246/9246 [1:01:13<00:00,  2.52it/s, GPU_ID=2, train_accuracy=0.996, train_loss=0.0193]\n",
      "Epoch 20/30: 100%|██████████| 9246/9246 [1:01:06<00:00,  2.52it/s, GPU_ID=1, train_accuracy=0.996, train_loss=0.0152]\n",
      "Evaluating Epoch 20/30: 100%|██████████| 3993/3993 [07:15<00:00,  9.16it/s]\n",
      "Evaluating Epoch 20/30: 100%|██████████| 3993/3993 [07:21<00:00,  9.05it/s]\n",
      "Evaluating Epoch 20/30: 100%|██████████| 3993/3993 [07:24<00:00,  8.99it/s]\n",
      "Evaluating Epoch 20/30: 100%|██████████| 3993/3993 [07:24<00:00,  8.98it/s]\n",
      "2024-11-02 02:32:42.202 | INFO     | scGenAI.training.train:evaluate_and_save_predictions:328 - Epoch 20/30, Train Loss: 0.0126, Train Accuracy: 0.9962, Context-level Val Loss: 1.8108, Context-level Val Accuracy: 0.8159\n",
      "Epoch 21/30:   0%|          | 0/9246 [00:00<?, ?it/s]2024-11-02 02:32:48.490 | INFO     | scGenAI.utils.distributed:savemodel:54 - Model state saved to /home/ubuntu/fulldataset/model/Finetune_GSE135194BW/best_model/scGenAI_model.pt\n",
      "2024-11-02 02:32:48.644 | INFO     | scGenAI.utils.distributed:savemodel:59 - Tokenizer saved to /home/ubuntu/fulldataset/model/Finetune_GSE135194BW/best_model\n",
      "2024-11-02 02:32:48.648 | INFO     | scGenAI.utils.distributed:savemodel:75 - Custom vocab and label encoder saved\n",
      "2024-11-02 02:32:48.649 | INFO     | scGenAI.utils.distributed:savemodel:81 - Model configuration saved to /home/ubuntu/fulldataset/model/Finetune_GSE135194BW/best_model\n",
      "Epoch 21/30: 100%|██████████| 9246/9246 [1:01:18<00:00,  2.51it/s, GPU_ID=2, train_accuracy=0.997, train_loss=0.00856]\n",
      "Epoch 21/30: 100%|██████████| 9246/9246 [1:01:13<00:00,  2.52it/s, GPU_ID=3, train_accuracy=0.998, train_loss=0.00837]\n",
      "Epoch 21/30: 100%|██████████| 9246/9246 [1:01:10<00:00,  2.52it/s, GPU_ID=1, train_accuracy=0.996, train_loss=0.0154]\n",
      "Epoch 21/30: 100%|██████████| 9246/9246 [1:01:07<00:00,  2.52it/s, GPU_ID=0, train_accuracy=0.997, train_loss=0.015]\n",
      "Evaluating Epoch 21/30: 100%|██████████| 3993/3993 [07:15<00:00,  9.16it/s]\n",
      "Evaluating Epoch 21/30: 100%|██████████| 3993/3993 [07:20<00:00,  9.06it/s]\n",
      "Evaluating Epoch 21/30: 100%|██████████| 3993/3993 [07:24<00:00,  8.99it/s]\n",
      "Evaluating Epoch 21/30: 100%|██████████| 3993/3993 [07:24<00:00,  8.98it/s]\n",
      "2024-11-02 03:41:21.270 | INFO     | scGenAI.training.train:evaluate_and_save_predictions:328 - Epoch 21/30, Train Loss: 0.0150, Train Accuracy: 0.9970, Context-level Val Loss: 1.9330, Context-level Val Accuracy: 0.8137\n",
      "Epoch 22/30: 100%|██████████| 9246/9246 [1:01:12<00:00,  2.52it/s, GPU_ID=2, train_accuracy=0.997, train_loss=0.0112]]\n",
      "Epoch 22/30: 100%|██████████| 9246/9246 [1:01:07<00:00,  2.52it/s, GPU_ID=3, train_accuracy=0.998, train_loss=0.00977]\n",
      "Epoch 22/30: 100%|██████████| 9246/9246 [1:01:04<00:00,  2.52it/s, GPU_ID=1, train_accuracy=0.997, train_loss=0.0122]\n",
      "Epoch 22/30: 100%|██████████| 9246/9246 [1:01:03<00:00,  2.52it/s, GPU_ID=0, train_accuracy=0.997, train_loss=0.0135]\n",
      "Evaluating Epoch 22/30: 100%|██████████| 3993/3993 [07:14<00:00,  9.18it/s]\n",
      "Evaluating Epoch 22/30: 100%|██████████| 3993/3993 [07:19<00:00,  9.08it/s]\n",
      "Evaluating Epoch 22/30: 100%|██████████| 3993/3993 [07:24<00:00,  8.99it/s]\n",
      "Evaluating Epoch 22/30: 100%|██████████| 3993/3993 [07:24<00:00,  8.99it/s]\n",
      "2024-11-02 04:49:53.275 | INFO     | scGenAI.training.train:evaluate_and_save_predictions:328 - Epoch 22/30, Train Loss: 0.0135, Train Accuracy: 0.9965, Context-level Val Loss: 1.9706, Context-level Val Accuracy: 0.8082\n",
      "Epoch 23/30: 100%|██████████| 9246/9246 [1:01:15<00:00,  2.52it/s, GPU_ID=2, train_accuracy=0.997, train_loss=0.0134]]\n",
      "Epoch 23/30: 100%|██████████| 9246/9246 [1:01:10<00:00,  2.52it/s, GPU_ID=3, train_accuracy=0.997, train_loss=0.0111]\n",
      "Epoch 23/30: 100%|██████████| 9246/9246 [1:01:06<00:00,  2.52it/s, GPU_ID=1, train_accuracy=0.997, train_loss=0.0116]\n",
      "Epoch 23/30: 100%|██████████| 9246/9246 [1:01:05<00:00,  2.52it/s, GPU_ID=0, train_accuracy=0.998, train_loss=0.00944]\n",
      "Evaluating Epoch 23/30: 100%|██████████| 3993/3993 [07:15<00:00,  9.17it/s]\n",
      "Evaluating Epoch 23/30: 100%|██████████| 3993/3993 [07:19<00:00,  9.08it/s]\n",
      "Evaluating Epoch 23/30: 100%|██████████| 3993/3993 [07:24<00:00,  8.99it/s]\n",
      "Evaluating Epoch 23/30: 100%|██████████| 3993/3993 [07:25<00:00,  8.97it/s]\n",
      "2024-11-02 05:58:28.574 | INFO     | scGenAI.training.train:evaluate_and_save_predictions:328 - Epoch 23/30, Train Loss: 0.0094, Train Accuracy: 0.9982, Context-level Val Loss: 1.9017, Context-level Val Accuracy: 0.8117\n",
      "Epoch 24/30: 100%|██████████| 9246/9246 [1:01:08<00:00,  2.52it/s, GPU_ID=3, train_accuracy=0.998, train_loss=0.0123]]\n",
      "Epoch 24/30: 100%|██████████| 9246/9246 [1:01:12<00:00,  2.52it/s, GPU_ID=2, train_accuracy=0.996, train_loss=0.014]\n",
      "Epoch 24/30: 100%|██████████| 9246/9246 [1:01:03<00:00,  2.52it/s, GPU_ID=0, train_accuracy=0.997, train_loss=0.00977]\n",
      "Epoch 24/30: 100%|██████████| 9246/9246 [1:01:04<00:00,  2.52it/s, GPU_ID=1, train_accuracy=0.998, train_loss=0.0063]\n",
      "Evaluating Epoch 24/30: 100%|██████████| 3993/3993 [07:14<00:00,  9.18it/s]\n",
      "Evaluating Epoch 24/30: 100%|██████████| 3993/3993 [07:19<00:00,  9.08it/s]\n",
      "Evaluating Epoch 24/30: 100%|██████████| 3993/3993 [07:24<00:00,  8.98it/s]\n",
      "Evaluating Epoch 24/30: 100%|██████████| 3993/3993 [07:25<00:00,  8.97it/s]\n",
      "2024-11-02 07:07:01.383 | INFO     | scGenAI.training.train:evaluate_and_save_predictions:328 - Epoch 24/30, Train Loss: 0.0098, Train Accuracy: 0.9972, Context-level Val Loss: 2.2318, Context-level Val Accuracy: 0.8057\n",
      "Epoch 25/30: 100%|██████████| 9246/9246 [1:01:07<00:00,  2.52it/s, GPU_ID=3, train_accuracy=0.998, train_loss=0.00933]\n",
      "Epoch 25/30: 100%|██████████| 9246/9246 [1:01:02<00:00,  2.52it/s, GPU_ID=0, train_accuracy=0.997, train_loss=0.0118]\n",
      "Epoch 25/30: 100%|██████████| 9246/9246 [1:01:12<00:00,  2.52it/s, GPU_ID=2, train_accuracy=0.998, train_loss=0.0104]\n",
      "Epoch 25/30: 100%|██████████| 9246/9246 [1:01:03<00:00,  2.52it/s, GPU_ID=1, train_accuracy=0.997, train_loss=0.0135]\n",
      "Evaluating Epoch 25/30: 100%|██████████| 3993/3993 [07:14<00:00,  9.18it/s]\n",
      "Evaluating Epoch 25/30: 100%|██████████| 3993/3993 [07:19<00:00,  9.08it/s]\n",
      "Evaluating Epoch 25/30: 100%|██████████| 3993/3993 [07:24<00:00,  8.98it/s]\n",
      "Evaluating Epoch 25/30: 100%|██████████| 3993/3993 [07:24<00:00,  8.97it/s]\n",
      "2024-11-02 08:15:32.984 | INFO     | scGenAI.training.train:evaluate_and_save_predictions:328 - Epoch 25/30, Train Loss: 0.0118, Train Accuracy: 0.9971, Context-level Val Loss: 2.0592, Context-level Val Accuracy: 0.8009\n",
      "Epoch 26/30: 100%|██████████| 9246/9246 [1:01:01<00:00,  2.53it/s, GPU_ID=0, train_accuracy=0.998, train_loss=0.0104]]\n",
      "Epoch 26/30: 100%|██████████| 9246/9246 [1:01:01<00:00,  2.53it/s, GPU_ID=1, train_accuracy=0.997, train_loss=0.0149]\n",
      "Epoch 26/30: 100%|██████████| 9246/9246 [1:01:06<00:00,  2.52it/s, GPU_ID=3, train_accuracy=0.998, train_loss=0.00809]\n",
      "Epoch 26/30: 100%|██████████| 9246/9246 [1:01:11<00:00,  2.52it/s, GPU_ID=2, train_accuracy=0.998, train_loss=0.00818]\n",
      "Evaluating Epoch 26/30: 100%|██████████| 3993/3993 [07:14<00:00,  9.18it/s]\n",
      "Evaluating Epoch 26/30: 100%|██████████| 3993/3993 [07:18<00:00,  9.10it/s]\n",
      "Evaluating Epoch 26/30: 100%|██████████| 3993/3993 [07:23<00:00,  9.00it/s]\n",
      "Evaluating Epoch 26/30: 100%|██████████| 3993/3993 [07:24<00:00,  8.97it/s]\n",
      "2024-11-02 09:24:03.316 | INFO     | scGenAI.training.train:evaluate_and_save_predictions:328 - Epoch 26/30, Train Loss: 0.0104, Train Accuracy: 0.9977, Context-level Val Loss: 1.9710, Context-level Val Accuracy: 0.8074\n",
      "Epoch 27/30: 100%|██████████| 9246/9246 [1:01:11<00:00,  2.52it/s, GPU_ID=2, train_accuracy=0.997, train_loss=0.0122]]\n",
      "Epoch 27/30: 100%|██████████| 9246/9246 [1:01:08<00:00,  2.52it/s, GPU_ID=3, train_accuracy=0.998, train_loss=0.00663]\n",
      "Epoch 27/30: 100%|██████████| 9246/9246 [1:01:03<00:00,  2.52it/s, GPU_ID=1, train_accuracy=0.998, train_loss=0.00672]\n",
      "Epoch 27/30: 100%|██████████| 9246/9246 [1:01:01<00:00,  2.52it/s, GPU_ID=0, train_accuracy=0.999, train_loss=0.00535]\n",
      "Evaluating Epoch 27/30: 100%|██████████| 3993/3993 [07:14<00:00,  9.19it/s]\n",
      "Evaluating Epoch 27/30: 100%|██████████| 3993/3993 [07:18<00:00,  9.10it/s]\n",
      "Evaluating Epoch 27/30: 100%|██████████| 3993/3993 [07:22<00:00,  9.02it/s]\n",
      "Evaluating Epoch 27/30: 100%|██████████| 3993/3993 [07:24<00:00,  8.97it/s]\n",
      "2024-11-02 10:32:34.468 | INFO     | scGenAI.training.train:evaluate_and_save_predictions:328 - Epoch 27/30, Train Loss: 0.0053, Train Accuracy: 0.9990, Context-level Val Loss: 2.0197, Context-level Val Accuracy: 0.8092\n",
      "Epoch 28/30: 100%|██████████| 9246/9246 [1:01:09<00:00,  2.52it/s, GPU_ID=2, train_accuracy=0.998, train_loss=0.0104]]\n",
      "Epoch 28/30: 100%|██████████| 9246/9246 [1:01:06<00:00,  2.52it/s, GPU_ID=3, train_accuracy=0.998, train_loss=0.00663]\n",
      "Epoch 28/30: 100%|██████████| 9246/9246 [1:01:02<00:00,  2.52it/s, GPU_ID=1, train_accuracy=0.998, train_loss=0.00983]\n",
      "Epoch 28/30: 100%|██████████| 9246/9246 [1:00:59<00:00,  2.53it/s, GPU_ID=0, train_accuracy=0.999, train_loss=0.00539]\n",
      "Evaluating Epoch 28/30: 100%|██████████| 3993/3993 [07:14<00:00,  9.19it/s]\n",
      "Evaluating Epoch 28/30: 100%|██████████| 3993/3993 [07:19<00:00,  9.09it/s]\n",
      "Evaluating Epoch 28/30: 100%|██████████| 3993/3993 [07:22<00:00,  9.03it/s]\n",
      "Evaluating Epoch 28/30: 100%|██████████| 3993/3993 [07:24<00:00,  8.97it/s]\n",
      "2024-11-02 11:41:03.516 | INFO     | scGenAI.training.train:evaluate_and_save_predictions:328 - Epoch 28/30, Train Loss: 0.0054, Train Accuracy: 0.9987, Context-level Val Loss: 2.0104, Context-level Val Accuracy: 0.8079\n",
      "Epoch 29/30: 100%|██████████| 9246/9246 [1:01:10<00:00,  2.52it/s, GPU_ID=2, train_accuracy=0.999, train_loss=0.00819]\n",
      "Epoch 29/30: 100%|██████████| 9246/9246 [1:01:03<00:00,  2.52it/s, GPU_ID=1, train_accuracy=0.998, train_loss=0.00686]\n",
      "Epoch 29/30: 100%|██████████| 9246/9246 [1:01:06<00:00,  2.52it/s, GPU_ID=3, train_accuracy=0.999, train_loss=0.0074]\n",
      "Epoch 29/30: 100%|██████████| 9246/9246 [1:01:00<00:00,  2.53it/s, GPU_ID=0, train_accuracy=0.998, train_loss=0.00858]\n",
      "Evaluating Epoch 29/30: 100%|██████████| 3993/3993 [07:14<00:00,  9.19it/s]\n",
      "Evaluating Epoch 29/30: 100%|██████████| 3993/3993 [07:19<00:00,  9.09it/s]\n",
      "Evaluating Epoch 29/30: 100%|██████████| 3993/3993 [07:22<00:00,  9.03it/s]\n",
      "Evaluating Epoch 29/30: 100%|██████████| 3993/3993 [07:24<00:00,  8.97it/s]\n",
      "2024-11-02 12:49:33.443 | INFO     | scGenAI.training.train:evaluate_and_save_predictions:328 - Epoch 29/30, Train Loss: 0.0086, Train Accuracy: 0.9984, Context-level Val Loss: 2.0497, Context-level Val Accuracy: 0.8102\n",
      "Epoch 30/30: 100%|██████████| 9246/9246 [1:01:00<00:00,  2.53it/s, GPU_ID=0, train_accuracy=0.998, train_loss=0.00926]\n",
      "Epoch 30/30: 100%|██████████| 9246/9246 [1:01:06<00:00,  2.52it/s, GPU_ID=3, train_accuracy=0.998, train_loss=0.00914]\n",
      "Epoch 30/30: 100%|██████████| 9246/9246 [1:01:03<00:00,  2.52it/s, GPU_ID=1, train_accuracy=0.998, train_loss=0.0108]]\n",
      "\n",
      "Evaluating Epoch 30/30: 100%|██████████| 3993/3993 [07:14<00:00,  9.18it/s]\n",
      "Evaluating Epoch 30/30: 100%|██████████| 3993/3993 [07:18<00:00,  9.10it/s]\n",
      "Evaluating Epoch 30/30: 100%|██████████| 3993/3993 [07:23<00:00,  9.01it/s]\n",
      "Evaluating Epoch 30/30: 100%|██████████| 3993/3993 [07:25<00:00,  8.97it/s]\n",
      "2024-11-02 13:58:03.769 | INFO     | scGenAI.training.train:evaluate_and_save_predictions:328 - Epoch 30/30, Train Loss: 0.0093, Train Accuracy: 0.9984, Context-level Val Loss: 1.9343, Context-level Val Accuracy: 0.8097\n",
      "2024-11-02 13:58:08.559 | INFO     | scGenAI.utils.distributed:savemodel:54 - Model state saved to /home/ubuntu/fulldataset/model/Finetune_GSE135194BW/last_model/scGenAI_model.pt\n",
      "2024-11-02 13:58:08.699 | INFO     | scGenAI.utils.distributed:savemodel:59 - Tokenizer saved to /home/ubuntu/fulldataset/model/Finetune_GSE135194BW/last_model\n",
      "2024-11-02 13:58:08.703 | INFO     | scGenAI.utils.distributed:savemodel:75 - Custom vocab and label encoder saved\n",
      "2024-11-02 13:58:08.704 | INFO     | scGenAI.utils.distributed:savemodel:81 - Model configuration saved to /home/ubuntu/fulldataset/model/Finetune_GSE135194BW/last_model\n",
      "2024-11-02 13:58:08.705 | INFO     | scGenAI.utils.distributed:cleanup:33 - Cleaned up DDP environment\n",
      "2024-11-02 13:58:08.705 | INFO     | scGenAI.utils.distributed:cleanup:33 - Cleaned up DDP environment\n",
      "2024-11-02 13:58:08.705 | INFO     | scGenAI.utils.distributed:cleanup:33 - Cleaned up DDP environment\n",
      "2024-11-02 13:58:08.705 | INFO     | scGenAI.utils.distributed:cleanup:33 - Cleaned up DDP environment\n",
      "\u001b[32m2024-11-02 13:58:10.245\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mmerge_predictions_across_gpus\u001b[0m:\u001b[36m147\u001b[0m - \u001b[1mMerging predictions for epoch 1\u001b[0m\n",
      "\u001b[32m2024-11-02 13:58:17.092\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mpostprocess_epoch\u001b[0m:\u001b[36m211\u001b[0m - \u001b[1mEpoch 1 - Validation Accuracy: 0.6420\u001b[0m\n",
      "\u001b[32m2024-11-02 13:58:17.577\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mmerge_predictions_across_gpus\u001b[0m:\u001b[36m147\u001b[0m - \u001b[1mMerging predictions for epoch 2\u001b[0m\n",
      "\u001b[32m2024-11-02 13:58:24.515\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mpostprocess_epoch\u001b[0m:\u001b[36m211\u001b[0m - \u001b[1mEpoch 2 - Validation Accuracy: 0.7420\u001b[0m\n",
      "\u001b[32m2024-11-02 13:58:25.002\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mmerge_predictions_across_gpus\u001b[0m:\u001b[36m147\u001b[0m - \u001b[1mMerging predictions for epoch 3\u001b[0m\n",
      "\u001b[32m2024-11-02 13:58:31.925\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mpostprocess_epoch\u001b[0m:\u001b[36m211\u001b[0m - \u001b[1mEpoch 3 - Validation Accuracy: 0.7753\u001b[0m\n",
      "\u001b[32m2024-11-02 13:58:32.413\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mmerge_predictions_across_gpus\u001b[0m:\u001b[36m147\u001b[0m - \u001b[1mMerging predictions for epoch 4\u001b[0m\n",
      "\u001b[32m2024-11-02 13:58:39.339\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mpostprocess_epoch\u001b[0m:\u001b[36m211\u001b[0m - \u001b[1mEpoch 4 - Validation Accuracy: 0.7873\u001b[0m\n",
      "\u001b[32m2024-11-02 13:58:39.827\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mmerge_predictions_across_gpus\u001b[0m:\u001b[36m147\u001b[0m - \u001b[1mMerging predictions for epoch 5\u001b[0m\n",
      "\u001b[32m2024-11-02 13:58:46.749\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mpostprocess_epoch\u001b[0m:\u001b[36m211\u001b[0m - \u001b[1mEpoch 5 - Validation Accuracy: 0.7907\u001b[0m\n",
      "\u001b[32m2024-11-02 13:58:47.238\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mmerge_predictions_across_gpus\u001b[0m:\u001b[36m147\u001b[0m - \u001b[1mMerging predictions for epoch 6\u001b[0m\n",
      "\u001b[32m2024-11-02 13:58:54.148\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mpostprocess_epoch\u001b[0m:\u001b[36m211\u001b[0m - \u001b[1mEpoch 6 - Validation Accuracy: 0.7980\u001b[0m\n",
      "\u001b[32m2024-11-02 13:58:54.636\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mmerge_predictions_across_gpus\u001b[0m:\u001b[36m147\u001b[0m - \u001b[1mMerging predictions for epoch 7\u001b[0m\n",
      "\u001b[32m2024-11-02 13:59:01.571\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mpostprocess_epoch\u001b[0m:\u001b[36m211\u001b[0m - \u001b[1mEpoch 7 - Validation Accuracy: 0.7853\u001b[0m\n",
      "\u001b[32m2024-11-02 13:59:02.060\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mmerge_predictions_across_gpus\u001b[0m:\u001b[36m147\u001b[0m - \u001b[1mMerging predictions for epoch 8\u001b[0m\n",
      "\u001b[32m2024-11-02 13:59:09.299\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mpostprocess_epoch\u001b[0m:\u001b[36m211\u001b[0m - \u001b[1mEpoch 8 - Validation Accuracy: 0.7993\u001b[0m\n",
      "\u001b[32m2024-11-02 13:59:09.788\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mmerge_predictions_across_gpus\u001b[0m:\u001b[36m147\u001b[0m - \u001b[1mMerging predictions for epoch 9\u001b[0m\n",
      "\u001b[32m2024-11-02 13:59:16.713\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mpostprocess_epoch\u001b[0m:\u001b[36m211\u001b[0m - \u001b[1mEpoch 9 - Validation Accuracy: 0.7913\u001b[0m\n",
      "\u001b[32m2024-11-02 13:59:17.201\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mmerge_predictions_across_gpus\u001b[0m:\u001b[36m147\u001b[0m - \u001b[1mMerging predictions for epoch 10\u001b[0m\n",
      "\u001b[32m2024-11-02 13:59:24.136\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mpostprocess_epoch\u001b[0m:\u001b[36m211\u001b[0m - \u001b[1mEpoch 10 - Validation Accuracy: 0.7987\u001b[0m\n",
      "\u001b[32m2024-11-02 13:59:24.626\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mmerge_predictions_across_gpus\u001b[0m:\u001b[36m147\u001b[0m - \u001b[1mMerging predictions for epoch 11\u001b[0m\n",
      "\u001b[32m2024-11-02 13:59:31.555\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mpostprocess_epoch\u001b[0m:\u001b[36m211\u001b[0m - \u001b[1mEpoch 11 - Validation Accuracy: 0.7813\u001b[0m\n",
      "\u001b[32m2024-11-02 13:59:32.044\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mmerge_predictions_across_gpus\u001b[0m:\u001b[36m147\u001b[0m - \u001b[1mMerging predictions for epoch 12\u001b[0m\n",
      "\u001b[32m2024-11-02 13:59:38.952\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mpostprocess_epoch\u001b[0m:\u001b[36m211\u001b[0m - \u001b[1mEpoch 12 - Validation Accuracy: 0.7907\u001b[0m\n",
      "\u001b[32m2024-11-02 13:59:39.442\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mmerge_predictions_across_gpus\u001b[0m:\u001b[36m147\u001b[0m - \u001b[1mMerging predictions for epoch 13\u001b[0m\n",
      "\u001b[32m2024-11-02 13:59:46.361\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mpostprocess_epoch\u001b[0m:\u001b[36m211\u001b[0m - \u001b[1mEpoch 13 - Validation Accuracy: 0.8047\u001b[0m\n",
      "\u001b[32m2024-11-02 13:59:46.848\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mmerge_predictions_across_gpus\u001b[0m:\u001b[36m147\u001b[0m - \u001b[1mMerging predictions for epoch 14\u001b[0m\n",
      "\u001b[32m2024-11-02 13:59:54.290\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mpostprocess_epoch\u001b[0m:\u001b[36m211\u001b[0m - \u001b[1mEpoch 14 - Validation Accuracy: 0.7947\u001b[0m\n",
      "\u001b[32m2024-11-02 13:59:54.779\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mmerge_predictions_across_gpus\u001b[0m:\u001b[36m147\u001b[0m - \u001b[1mMerging predictions for epoch 15\u001b[0m\n",
      "\u001b[32m2024-11-02 14:00:01.708\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mpostprocess_epoch\u001b[0m:\u001b[36m211\u001b[0m - \u001b[1mEpoch 15 - Validation Accuracy: 0.8087\u001b[0m\n",
      "\u001b[32m2024-11-02 14:00:02.196\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mmerge_predictions_across_gpus\u001b[0m:\u001b[36m147\u001b[0m - \u001b[1mMerging predictions for epoch 16\u001b[0m\n",
      "\u001b[32m2024-11-02 14:00:09.102\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mpostprocess_epoch\u001b[0m:\u001b[36m211\u001b[0m - \u001b[1mEpoch 16 - Validation Accuracy: 0.7993\u001b[0m\n",
      "\u001b[32m2024-11-02 14:00:09.590\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mmerge_predictions_across_gpus\u001b[0m:\u001b[36m147\u001b[0m - \u001b[1mMerging predictions for epoch 17\u001b[0m\n",
      "\u001b[32m2024-11-02 14:00:16.489\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mpostprocess_epoch\u001b[0m:\u001b[36m211\u001b[0m - \u001b[1mEpoch 17 - Validation Accuracy: 0.8000\u001b[0m\n",
      "\u001b[32m2024-11-02 14:00:16.977\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mmerge_predictions_across_gpus\u001b[0m:\u001b[36m147\u001b[0m - \u001b[1mMerging predictions for epoch 18\u001b[0m\n",
      "\u001b[32m2024-11-02 14:00:23.922\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mpostprocess_epoch\u001b[0m:\u001b[36m211\u001b[0m - \u001b[1mEpoch 18 - Validation Accuracy: 0.7853\u001b[0m\n",
      "\u001b[32m2024-11-02 14:00:24.410\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mmerge_predictions_across_gpus\u001b[0m:\u001b[36m147\u001b[0m - \u001b[1mMerging predictions for epoch 19\u001b[0m\n",
      "\u001b[32m2024-11-02 14:00:31.323\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mpostprocess_epoch\u001b[0m:\u001b[36m211\u001b[0m - \u001b[1mEpoch 19 - Validation Accuracy: 0.7933\u001b[0m\n",
      "\u001b[32m2024-11-02 14:00:31.812\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mmerge_predictions_across_gpus\u001b[0m:\u001b[36m147\u001b[0m - \u001b[1mMerging predictions for epoch 20\u001b[0m\n",
      "\u001b[32m2024-11-02 14:00:38.735\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mpostprocess_epoch\u001b[0m:\u001b[36m211\u001b[0m - \u001b[1mEpoch 20 - Validation Accuracy: 0.8000\u001b[0m\n",
      "\u001b[32m2024-11-02 14:00:39.222\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mmerge_predictions_across_gpus\u001b[0m:\u001b[36m147\u001b[0m - \u001b[1mMerging predictions for epoch 21\u001b[0m\n",
      "\u001b[32m2024-11-02 14:00:46.470\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mpostprocess_epoch\u001b[0m:\u001b[36m211\u001b[0m - \u001b[1mEpoch 21 - Validation Accuracy: 0.8000\u001b[0m\n",
      "\u001b[32m2024-11-02 14:00:46.958\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mmerge_predictions_across_gpus\u001b[0m:\u001b[36m147\u001b[0m - \u001b[1mMerging predictions for epoch 22\u001b[0m\n",
      "\u001b[32m2024-11-02 14:00:53.879\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mpostprocess_epoch\u001b[0m:\u001b[36m211\u001b[0m - \u001b[1mEpoch 22 - Validation Accuracy: 0.7920\u001b[0m\n",
      "\u001b[32m2024-11-02 14:00:54.366\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mmerge_predictions_across_gpus\u001b[0m:\u001b[36m147\u001b[0m - \u001b[1mMerging predictions for epoch 23\u001b[0m\n",
      "\u001b[32m2024-11-02 14:01:01.276\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mpostprocess_epoch\u001b[0m:\u001b[36m211\u001b[0m - \u001b[1mEpoch 23 - Validation Accuracy: 0.7967\u001b[0m\n",
      "\u001b[32m2024-11-02 14:01:01.763\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mmerge_predictions_across_gpus\u001b[0m:\u001b[36m147\u001b[0m - \u001b[1mMerging predictions for epoch 24\u001b[0m\n",
      "\u001b[32m2024-11-02 14:01:08.657\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mpostprocess_epoch\u001b[0m:\u001b[36m211\u001b[0m - \u001b[1mEpoch 24 - Validation Accuracy: 0.7860\u001b[0m\n",
      "\u001b[32m2024-11-02 14:01:09.144\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mmerge_predictions_across_gpus\u001b[0m:\u001b[36m147\u001b[0m - \u001b[1mMerging predictions for epoch 25\u001b[0m\n",
      "\u001b[32m2024-11-02 14:01:16.037\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mpostprocess_epoch\u001b[0m:\u001b[36m211\u001b[0m - \u001b[1mEpoch 25 - Validation Accuracy: 0.7813\u001b[0m\n",
      "\u001b[32m2024-11-02 14:01:16.523\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mmerge_predictions_across_gpus\u001b[0m:\u001b[36m147\u001b[0m - \u001b[1mMerging predictions for epoch 26\u001b[0m\n",
      "\u001b[32m2024-11-02 14:01:23.422\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mpostprocess_epoch\u001b[0m:\u001b[36m211\u001b[0m - \u001b[1mEpoch 26 - Validation Accuracy: 0.7887\u001b[0m\n",
      "\u001b[32m2024-11-02 14:01:23.909\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mmerge_predictions_across_gpus\u001b[0m:\u001b[36m147\u001b[0m - \u001b[1mMerging predictions for epoch 27\u001b[0m\n",
      "\u001b[32m2024-11-02 14:01:31.252\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mpostprocess_epoch\u001b[0m:\u001b[36m211\u001b[0m - \u001b[1mEpoch 27 - Validation Accuracy: 0.7960\u001b[0m\n",
      "\u001b[32m2024-11-02 14:01:31.739\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mmerge_predictions_across_gpus\u001b[0m:\u001b[36m147\u001b[0m - \u001b[1mMerging predictions for epoch 28\u001b[0m\n",
      "\u001b[32m2024-11-02 14:01:38.649\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mpostprocess_epoch\u001b[0m:\u001b[36m211\u001b[0m - \u001b[1mEpoch 28 - Validation Accuracy: 0.7980\u001b[0m\n",
      "\u001b[32m2024-11-02 14:01:39.136\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mmerge_predictions_across_gpus\u001b[0m:\u001b[36m147\u001b[0m - \u001b[1mMerging predictions for epoch 29\u001b[0m\n",
      "\u001b[32m2024-11-02 14:01:46.032\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mpostprocess_epoch\u001b[0m:\u001b[36m211\u001b[0m - \u001b[1mEpoch 29 - Validation Accuracy: 0.7973\u001b[0m\n",
      "\u001b[32m2024-11-02 14:01:46.517\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mmerge_predictions_across_gpus\u001b[0m:\u001b[36m147\u001b[0m - \u001b[1mMerging predictions for epoch 30\u001b[0m\n",
      "\u001b[32m2024-11-02 14:01:53.404\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mpostprocess_epoch\u001b[0m:\u001b[36m211\u001b[0m - \u001b[1mEpoch 30 - Validation Accuracy: 0.7980\u001b[0m\n",
      "\u001b[32m2024-11-02 14:01:53.891\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36msummarize_epochs\u001b[0m:\u001b[36m234\u001b[0m - \u001b[1mCompleted post-processing and summary of all epochs.\u001b[0m\n",
      "\u001b[32m2024-11-02 14:01:53.895\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mcombine_epoch_results\u001b[0m:\u001b[36m61\u001b[0m - \u001b[1mCombined results saved to /home/ubuntu/fulldataset/model/Finetune_GSE135194BW/combined_epoch_results.csv.\u001b[0m\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Plot saved to /home/ubuntu/fulldataset/model/Finetune_GSE135194BW/train_summary.pdf\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1000x600 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "config_file = '/home/ubuntu/scgenai/tutorials/yaml/config_Finetune_GSE135194BW_random_context.yaml'\n",
    "run_finetune_from_config(config_file)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2fe550d5-16c0-425a-bfd7-d257c1ecfe1d",
   "metadata": {},
   "source": [
    "## Post-Training Model and Results\n",
    "\n",
    "Upon completion of the training, the trained model will be saved in the `model_dir` as defined in the configuration file. Within this directory, two subfolders will be created: `best_model` and `last_model`. The `best_model` folder contains the model that achieved the highest training performance, while the `last_model` contains the model from the final epoch. Additionally, a `combined_epoch_results.csv` file will be generated in the `model_dir`, summarizing the training performance metrics across all epochs."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "cc88e128-eaba-4557-8fca-eae0966e482d",
   "metadata": {},
   "source": [
    "### Training Summary\n",
    "\n",
    "In addition to the output summary files generated after training, the training summary can be directly accessed in Python by loading the saved configuration file.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "9cd73cce-7ddc-47ab-b59f-fc4ac955b57f",
   "metadata": {},
   "outputs": [],
   "source": [
    "from scGenAI.config import Config\n",
    "from scGenAI.data.postprocess import TrainSummary"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "9bc20420-8143-4ff7-b474-d4a3f6936991",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Epoch</th>\n",
       "      <th>Validation Accuracy</th>\n",
       "      <th>Train Loss</th>\n",
       "      <th>Train Accuracy</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>0.642000</td>\n",
       "      <td>0.965655</td>\n",
       "      <td>0.672291</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2</td>\n",
       "      <td>0.742000</td>\n",
       "      <td>0.477099</td>\n",
       "      <td>0.833550</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3</td>\n",
       "      <td>0.775333</td>\n",
       "      <td>0.290277</td>\n",
       "      <td>0.898443</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>4</td>\n",
       "      <td>0.787333</td>\n",
       "      <td>0.178465</td>\n",
       "      <td>0.938892</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>5</td>\n",
       "      <td>0.790667</td>\n",
       "      <td>0.119970</td>\n",
       "      <td>0.960632</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>6</td>\n",
       "      <td>0.798000</td>\n",
       "      <td>0.089447</td>\n",
       "      <td>0.971880</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>7</td>\n",
       "      <td>0.785333</td>\n",
       "      <td>0.059383</td>\n",
       "      <td>0.980424</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>8</td>\n",
       "      <td>0.799333</td>\n",
       "      <td>0.048372</td>\n",
       "      <td>0.983885</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>9</td>\n",
       "      <td>0.791333</td>\n",
       "      <td>0.038447</td>\n",
       "      <td>0.988860</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>10</td>\n",
       "      <td>0.798667</td>\n",
       "      <td>0.025881</td>\n",
       "      <td>0.991997</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>11</td>\n",
       "      <td>0.781333</td>\n",
       "      <td>0.029033</td>\n",
       "      <td>0.991888</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>12</td>\n",
       "      <td>0.790667</td>\n",
       "      <td>0.030986</td>\n",
       "      <td>0.990699</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>13</td>\n",
       "      <td>0.804667</td>\n",
       "      <td>0.026771</td>\n",
       "      <td>0.993078</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>14</td>\n",
       "      <td>0.794667</td>\n",
       "      <td>0.021172</td>\n",
       "      <td>0.994160</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>15</td>\n",
       "      <td>0.808667</td>\n",
       "      <td>0.021423</td>\n",
       "      <td>0.994592</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>16</td>\n",
       "      <td>0.799333</td>\n",
       "      <td>0.014152</td>\n",
       "      <td>0.995241</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>17</td>\n",
       "      <td>0.800000</td>\n",
       "      <td>0.015010</td>\n",
       "      <td>0.995890</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>18</td>\n",
       "      <td>0.785333</td>\n",
       "      <td>0.017258</td>\n",
       "      <td>0.995782</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>19</td>\n",
       "      <td>0.793333</td>\n",
       "      <td>0.012694</td>\n",
       "      <td>0.996431</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>20</td>\n",
       "      <td>0.800000</td>\n",
       "      <td>0.012555</td>\n",
       "      <td>0.996215</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20</th>\n",
       "      <td>21</td>\n",
       "      <td>0.800000</td>\n",
       "      <td>0.015002</td>\n",
       "      <td>0.996972</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>21</th>\n",
       "      <td>22</td>\n",
       "      <td>0.792000</td>\n",
       "      <td>0.013493</td>\n",
       "      <td>0.996539</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22</th>\n",
       "      <td>23</td>\n",
       "      <td>0.796667</td>\n",
       "      <td>0.009442</td>\n",
       "      <td>0.998161</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>23</th>\n",
       "      <td>24</td>\n",
       "      <td>0.786000</td>\n",
       "      <td>0.009768</td>\n",
       "      <td>0.997188</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24</th>\n",
       "      <td>25</td>\n",
       "      <td>0.781333</td>\n",
       "      <td>0.011801</td>\n",
       "      <td>0.997080</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25</th>\n",
       "      <td>26</td>\n",
       "      <td>0.788667</td>\n",
       "      <td>0.010418</td>\n",
       "      <td>0.997729</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>26</th>\n",
       "      <td>27</td>\n",
       "      <td>0.796000</td>\n",
       "      <td>0.005350</td>\n",
       "      <td>0.999027</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27</th>\n",
       "      <td>28</td>\n",
       "      <td>0.798000</td>\n",
       "      <td>0.005390</td>\n",
       "      <td>0.998702</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>28</th>\n",
       "      <td>29</td>\n",
       "      <td>0.797333</td>\n",
       "      <td>0.008577</td>\n",
       "      <td>0.998378</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>29</th>\n",
       "      <td>30</td>\n",
       "      <td>0.798000</td>\n",
       "      <td>0.009257</td>\n",
       "      <td>0.998378</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    Epoch  Validation Accuracy  Train Loss  Train Accuracy\n",
       "0       1             0.642000    0.965655        0.672291\n",
       "1       2             0.742000    0.477099        0.833550\n",
       "2       3             0.775333    0.290277        0.898443\n",
       "3       4             0.787333    0.178465        0.938892\n",
       "4       5             0.790667    0.119970        0.960632\n",
       "5       6             0.798000    0.089447        0.971880\n",
       "6       7             0.785333    0.059383        0.980424\n",
       "7       8             0.799333    0.048372        0.983885\n",
       "8       9             0.791333    0.038447        0.988860\n",
       "9      10             0.798667    0.025881        0.991997\n",
       "10     11             0.781333    0.029033        0.991888\n",
       "11     12             0.790667    0.030986        0.990699\n",
       "12     13             0.804667    0.026771        0.993078\n",
       "13     14             0.794667    0.021172        0.994160\n",
       "14     15             0.808667    0.021423        0.994592\n",
       "15     16             0.799333    0.014152        0.995241\n",
       "16     17             0.800000    0.015010        0.995890\n",
       "17     18             0.785333    0.017258        0.995782\n",
       "18     19             0.793333    0.012694        0.996431\n",
       "19     20             0.800000    0.012555        0.996215\n",
       "20     21             0.800000    0.015002        0.996972\n",
       "21     22             0.792000    0.013493        0.996539\n",
       "22     23             0.796667    0.009442        0.998161\n",
       "23     24             0.786000    0.009768        0.997188\n",
       "24     25             0.781333    0.011801        0.997080\n",
       "25     26             0.788667    0.010418        0.997729\n",
       "26     27             0.796000    0.005350        0.999027\n",
       "27     28             0.798000    0.005390        0.998702\n",
       "28     29             0.797333    0.008577        0.998378\n",
       "29     30             0.798000    0.009257        0.998378"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "config = Config(config_file)\n",
    "summary = TrainSummary(config.finetune_dir)\n",
    "dfsummary = summary.table()\n",
    "dfsummary"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a0ba9299-9ce7-4625-8f21-c49a6e3fd75a",
   "metadata": {},
   "source": [
    "It can also be plotted using the `summary.plot`:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "24962a8b-d8b7-4b38-b982-10633416b86e",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Plot saved to /home/ubuntu/fulldataset/model/Finetune_GSE135194BW/train_summary.pdf\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1000x600 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "summary.plot(saveplot=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "346b4f66-b17c-462e-947f-0a3d08972cb4",
   "metadata": {},
   "source": [
    "## Prediction Using the Trained Model\n",
    "\n",
    "Similar to the training step, predictions can be run using the `run_prediction_from_config` function in scGenAI with a simple configuration file. Unlike the training configuration, the prediction configuration file only requires the following parameters to be defined: `mode`, `predict_file`, `batch_size`, `model_dir`, and `outputfile`. All other parameters will be adopted from the trained model.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "10cd73c8-8ea4-44cf-a2c5-6b2f4aecd53b",
   "metadata": {},
   "outputs": [],
   "source": [
    "from scGenAI.prediction.predict import run_prediction_from_config"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "6d8cab06-6cec-49b1-8fa8-53aa426c38df",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "model_dir: /home/ubuntu/fulldataset/model/Finetune_GSE135194BW/best_model/\n",
      "predict_file: /home/ubuntu/fulldataset/BoneMarrowRefine2_val.h5ad\n",
      "log_dir: /home/ubuntu/scgenai/examples/logs\n",
      "output_dir: /home/ubuntu/scgenai/examples/output\n",
      "seed: 1314521\n",
      "mode: Predict\n",
      "min_cells: 50\n",
      "multiomics: No\n",
      "keepIntermediateFiles: Yes\n",
      "savelog: No\n",
      "savesetting: Yes\n",
      "model_backbone_name: llama\n",
      "model_backbone_size: normal\n",
      "context_method: random\n",
      "num_bins: 10\n",
      "optimizer: AdamW\n",
      "weight_decay: 0.01\n",
      "depth: 2\n",
      "max_length: 1024\n",
      "batch_size: 8\n",
      "learning_rate: 1e-05\n",
      "num_epochs: 30\n",
      "world_size: 1\n",
      "evaluate_during_training: False\n",
      "hidden_size: 768\n",
      "num_layers: 12\n",
      "num_heads: 12\n",
      "intermediate_size: 3072\n",
      "stride: 512\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2024-11-02 18:20:51.393 | INFO     | scGenAI.utils.distributed:setup_distributed:26 - Distributed setup complete for GPU 3, Total GPUs used: 4\n",
      "2024-11-02 18:20:51.405 | INFO     | scGenAI.utils.distributed:setup_distributed:26 - Distributed setup complete for GPU 0, Total GPUs used: 4\n",
      "2024-11-02 18:20:51.416 | INFO     | scGenAI.utils.distributed:setup_distributed:26 - Distributed setup complete for GPU 1, Total GPUs used: 4\n",
      "2024-11-02 18:20:51.421 | INFO     | scGenAI.utils.distributed:setup_distributed:26 - Distributed setup complete for GPU 2, Total GPUs used: 4\n",
      "GPU 2 Predicting: 100%|██████████| 371/371 [01:18<00:00,  4.72it/s]\n",
      "GPU 1 Predicting: 100%|██████████| 371/371 [01:18<00:00,  4.71it/s]\n",
      "GPU 3 Predicting: 100%|██████████| 371/371 [01:19<00:00,  4.65it/s]\n",
      "GPU 0 Predicting: 100%|██████████| 371/371 [01:21<00:00,  4.57it/s]\n",
      "2024-11-02 18:22:18.397 | INFO     | scGenAI.utils.distributed:cleanup:33 - Cleaned up DDP environment\n",
      "2024-11-02 18:22:18.397 | INFO     | scGenAI.utils.distributed:cleanup:33 - Cleaned up DDP environment\n",
      "2024-11-02 18:22:18.397 | INFO     | scGenAI.utils.distributed:cleanup:33 - Cleaned up DDP environment\n",
      "2024-11-02 18:22:18.397 | INFO     | scGenAI.utils.distributed:cleanup:33 - Cleaned up DDP environment\n",
      "\u001b[32m2024-11-02 18:22:20.192\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mscGenAI.data.postprocess\u001b[0m:\u001b[36mcombine_prediction_results\u001b[0m:\u001b[36m103\u001b[0m - \u001b[1mCompleted post-processing for prediction\u001b[0m\n"
     ]
    }
   ],
   "source": [
    "pred_config_file = '/home/ubuntu/scgenai/tutorials/yaml/config_prediction_GSE135194BW_random_context.yaml'\n",
    "run_prediction_from_config(pred_config_file)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7fa3e6fc-55c8-4ef4-b64e-f6347b84fbb4",
   "metadata": {},
   "source": [
    "The prediction output is a CSV file, as defined in the configuration file. It contains the original metadata extracted from the input prediction file (`obs` slot) along with three additional columns: `context_id`, `PredictedFeature`, and `prediction_score`. \n",
    "\n",
    "- `context_id` represents the context used to determine the prediction for the corresponding cell. \n",
    "- `PredictedFeature` is the final predicted feature for the cell using the trained model. \n",
    "- `prediction_score` indicates the confidence level of the prediction, with a maximum value of 1.\n",
    "\n",
    "For example in this tutorials:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "b4c89ab7-232d-41a1-8cb4-f59851b5fe6d",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>index</th>\n",
       "      <th>orig.ident</th>\n",
       "      <th>nCount_RNA</th>\n",
       "      <th>nFeature_RNA</th>\n",
       "      <th>RNA_snn_res.0.8</th>\n",
       "      <th>seurat_clusters</th>\n",
       "      <th>sample</th>\n",
       "      <th>sampleId</th>\n",
       "      <th>projectId</th>\n",
       "      <th>sampleType</th>\n",
       "      <th>...</th>\n",
       "      <th>age</th>\n",
       "      <th>gender</th>\n",
       "      <th>processStatus</th>\n",
       "      <th>predicted_cell_type</th>\n",
       "      <th>ct</th>\n",
       "      <th>Barcode</th>\n",
       "      <th>cell_barcode</th>\n",
       "      <th>context_id</th>\n",
       "      <th>PredictedFeature</th>\n",
       "      <th>prediction_score</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>CATTCGCGTTACTGAC-1--GSM3993353</td>\n",
       "      <td>SeuratProject</td>\n",
       "      <td>11849</td>\n",
       "      <td>2853</td>\n",
       "      <td>15</td>\n",
       "      <td>15</td>\n",
       "      <td>GSM3993353</td>\n",
       "      <td>GSM3993353</td>\n",
       "      <td>GSE135194</td>\n",
       "      <td>Normal</td>\n",
       "      <td>...</td>\n",
       "      <td>31.0</td>\n",
       "      <td>F</td>\n",
       "      <td>QC pass</td>\n",
       "      <td>22:myeloid progenitor</td>\n",
       "      <td>Megakaryocytic progenitor cell</td>\n",
       "      <td>CATTCGCGTTACTGAC-1--GSM3993353</td>\n",
       "      <td>CATTCGCGTTACTGAC-1--GSM3993353</td>\n",
       "      <td>3_CATTCGCGTTACTGAC-1--GSM3993353_window_3</td>\n",
       "      <td>Haematopoietic stem cell</td>\n",
       "      <td>0.999969</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>TAGTTGGAGATGTGTA-1--GSM5515744</td>\n",
       "      <td>SeuratProject</td>\n",
       "      <td>1800</td>\n",
       "      <td>799</td>\n",
       "      <td>15</td>\n",
       "      <td>15</td>\n",
       "      <td>GSM5515744</td>\n",
       "      <td>GSM5515744</td>\n",
       "      <td>GSE181989</td>\n",
       "      <td>Normal</td>\n",
       "      <td>...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>QC pass</td>\n",
       "      <td>4:hematopoietic stem cell</td>\n",
       "      <td>Megakaryocytic progenitor cell</td>\n",
       "      <td>TAGTTGGAGATGTGTA-1--GSM5515744</td>\n",
       "      <td>TAGTTGGAGATGTGTA-1--GSM5515744</td>\n",
       "      <td>0_TAGTTGGAGATGTGTA-1--GSM5515744_window_14</td>\n",
       "      <td>Common lymphoid progenitor (CLP)</td>\n",
       "      <td>0.995315</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>CTCACACAGTGGCACA-1--GSM3993355</td>\n",
       "      <td>SeuratProject</td>\n",
       "      <td>8191</td>\n",
       "      <td>2212</td>\n",
       "      <td>20</td>\n",
       "      <td>20</td>\n",
       "      <td>GSM3993355</td>\n",
       "      <td>GSM3993355</td>\n",
       "      <td>GSE135194</td>\n",
       "      <td>Normal</td>\n",
       "      <td>...</td>\n",
       "      <td>44.0</td>\n",
       "      <td>M</td>\n",
       "      <td>QC pass</td>\n",
       "      <td>18:hematopoietic stem cell</td>\n",
       "      <td>Pro-B cell</td>\n",
       "      <td>CTCACACAGTGGCACA-1--GSM3993355</td>\n",
       "      <td>CTCACACAGTGGCACA-1--GSM3993355</td>\n",
       "      <td>2_CTCACACAGTGGCACA-1--GSM3993355_window_21</td>\n",
       "      <td>Pro-B cell</td>\n",
       "      <td>0.999961</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>GCCAAATAGTTACCCA-1--GSM3993355</td>\n",
       "      <td>SeuratProject</td>\n",
       "      <td>8839</td>\n",
       "      <td>2353</td>\n",
       "      <td>19</td>\n",
       "      <td>19</td>\n",
       "      <td>GSM3993355</td>\n",
       "      <td>GSM3993355</td>\n",
       "      <td>GSE135194</td>\n",
       "      <td>Normal</td>\n",
       "      <td>...</td>\n",
       "      <td>44.0</td>\n",
       "      <td>M</td>\n",
       "      <td>QC pass</td>\n",
       "      <td>13:cDC2</td>\n",
       "      <td>Conventional dendritic cell</td>\n",
       "      <td>GCCAAATAGTTACCCA-1--GSM3993355</td>\n",
       "      <td>GCCAAATAGTTACCCA-1--GSM3993355</td>\n",
       "      <td>2_GCCAAATAGTTACCCA-1--GSM3993355_window_30</td>\n",
       "      <td>Conventional dendritic cell</td>\n",
       "      <td>0.999884</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>TATTACCTCAGCATGT-1--GSM5515743</td>\n",
       "      <td>SeuratProject</td>\n",
       "      <td>3597</td>\n",
       "      <td>1248</td>\n",
       "      <td>3</td>\n",
       "      <td>3</td>\n",
       "      <td>GSM5515743</td>\n",
       "      <td>GSM5515743</td>\n",
       "      <td>GSE181989</td>\n",
       "      <td>Normal</td>\n",
       "      <td>...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>QC pass</td>\n",
       "      <td>5:effector memory CD8 T cell</td>\n",
       "      <td>Effector CD8 T cell</td>\n",
       "      <td>TATTACCTCAGCATGT-1--GSM5515743</td>\n",
       "      <td>TATTACCTCAGCATGT-1--GSM5515743</td>\n",
       "      <td>2_TATTACCTCAGCATGT-1--GSM5515743_window_34</td>\n",
       "      <td>Plasmacytoid dendritic cell</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 24 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                            index     orig.ident  nCount_RNA  nFeature_RNA   \n",
       "0  CATTCGCGTTACTGAC-1--GSM3993353  SeuratProject       11849          2853  \\\n",
       "1  TAGTTGGAGATGTGTA-1--GSM5515744  SeuratProject        1800           799   \n",
       "2  CTCACACAGTGGCACA-1--GSM3993355  SeuratProject        8191          2212   \n",
       "3  GCCAAATAGTTACCCA-1--GSM3993355  SeuratProject        8839          2353   \n",
       "4  TATTACCTCAGCATGT-1--GSM5515743  SeuratProject        3597          1248   \n",
       "\n",
       "   RNA_snn_res.0.8  seurat_clusters      sample    sampleId  projectId   \n",
       "0               15               15  GSM3993353  GSM3993353  GSE135194  \\\n",
       "1               15               15  GSM5515744  GSM5515744  GSE181989   \n",
       "2               20               20  GSM3993355  GSM3993355  GSE135194   \n",
       "3               19               19  GSM3993355  GSM3993355  GSE135194   \n",
       "4                3                3  GSM5515743  GSM5515743  GSE181989   \n",
       "\n",
       "  sampleType  ...   age gender processStatus           predicted_cell_type   \n",
       "0     Normal  ...  31.0      F       QC pass         22:myeloid progenitor  \\\n",
       "1     Normal  ...   NaN    NaN       QC pass     4:hematopoietic stem cell   \n",
       "2     Normal  ...  44.0      M       QC pass    18:hematopoietic stem cell   \n",
       "3     Normal  ...  44.0      M       QC pass                       13:cDC2   \n",
       "4     Normal  ...   NaN    NaN       QC pass  5:effector memory CD8 T cell   \n",
       "\n",
       "                               ct                         Barcode   \n",
       "0  Megakaryocytic progenitor cell  CATTCGCGTTACTGAC-1--GSM3993353  \\\n",
       "1  Megakaryocytic progenitor cell  TAGTTGGAGATGTGTA-1--GSM5515744   \n",
       "2                      Pro-B cell  CTCACACAGTGGCACA-1--GSM3993355   \n",
       "3     Conventional dendritic cell  GCCAAATAGTTACCCA-1--GSM3993355   \n",
       "4             Effector CD8 T cell  TATTACCTCAGCATGT-1--GSM5515743   \n",
       "\n",
       "                     cell_barcode                                  context_id   \n",
       "0  CATTCGCGTTACTGAC-1--GSM3993353   3_CATTCGCGTTACTGAC-1--GSM3993353_window_3  \\\n",
       "1  TAGTTGGAGATGTGTA-1--GSM5515744  0_TAGTTGGAGATGTGTA-1--GSM5515744_window_14   \n",
       "2  CTCACACAGTGGCACA-1--GSM3993355  2_CTCACACAGTGGCACA-1--GSM3993355_window_21   \n",
       "3  GCCAAATAGTTACCCA-1--GSM3993355  2_GCCAAATAGTTACCCA-1--GSM3993355_window_30   \n",
       "4  TATTACCTCAGCATGT-1--GSM5515743  2_TATTACCTCAGCATGT-1--GSM5515743_window_34   \n",
       "\n",
       "                   PredictedFeature prediction_score  \n",
       "0          Haematopoietic stem cell         0.999969  \n",
       "1  Common lymphoid progenitor (CLP)         0.995315  \n",
       "2                        Pro-B cell         0.999961  \n",
       "3       Conventional dendritic cell         0.999884  \n",
       "4       Plasmacytoid dendritic cell         1.000000  \n",
       "\n",
       "[5 rows x 24 columns]"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import pandas as pd\n",
    "preconfig = Config(pred_config_file)\n",
    "dfprediction = pd.read_csv(preconfig.outputfile)\n",
    "dfprediction.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "0cd6dab7-bb05-4204-bc26-b730ad4461a4",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.8"
  },
  "toc": {
   "base_numbering": 1,
   "nav_menu": {},
   "number_sections": false,
   "sideBar": true,
   "skip_h1_title": false,
   "title_cell": "Table of Contents",
   "title_sidebar": "Contents",
   "toc_cell": false,
   "toc_position": {},
   "toc_section_display": true,
   "toc_window_display": true
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}