fastsurfer.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Environment variables set for FreeSurfer:\n",
      "FREESURFER_HOME: /Applications/freesurfer/7.4.1\n",
      "SUBJECTS_DIR: /Applications/freesurfer/7.4.1/subjects\n",
      "FSFAST_HOME: /Applications/freesurfer/7.4.1/fsfast\n",
      "FSF_OUTPUT_FORMAT: nii.gz\n",
      "FSL_DIR: /usr/local/fsl\n",
      "MINC_BIN_DIR: /Applications/freesurfer/7.4.1/mni/bin\n",
      "MINC_LIB_DIR: /Applications/freesurfer/7.4.1/mni/lib\n",
      "MNI_DIR: /Applications/freesurfer/7.4.1/mni\n",
      "PYTHONPATH: /Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer:/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer:/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer:/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer:/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer:/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer:/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer:/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer/FastSurferCNN:/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer/FastSurferCNN:/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer/FastSurferCNN:/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer/FastSurferCNN:/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer/FastSurferCNN:/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer/FastSurferCNN:/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer/FastSurferCNN:/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer/FastSurferCNN:\n",
      "PATH: /usr/local/fsl/bin:/Applications/freesurfer/7.4.1/mni/bin:/Applications/freesurfer/7.4.1/bin:/Applications/freesurfer/7.4.1/fsfast/bin:/usr/local/fsl/bin:/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer/fastsurferenv/bin:/Users/suhanasaigoli/miniconda3/condabin:/Library/Frameworks/Python.framework/Versions/3.7/bin:/usr/local/bin:/System/Cryptexes/App/usr/bin:/usr/bin:/bin:/usr/sbin:/sbin:/var/run/com.apple.security.cryptexd/codex.system/bootstrap/usr/local/bin:/var/run/com.apple.security.cryptexd/codex.system/bootstrap/usr/bin:/var/run/com.apple.security.cryptexd/codex.system/bootstrap/usr/appleinternal/bin:/opt/X11/bin:/Library/Apple/usr/bin\n",
      "DYLD_LIBRARY_PATH: /Applications/freesurfer/7.4.1/lib/misc/lib:\n",
      "PERL5LIB: /Applications/freesurfer/7.4.1/mni/share/perl5\n"
     ]
    }
   ],
   "source": [
    "# USAGE - PIPELINE\n",
    "\n",
    "import os\n",
    "import subprocess\n",
    "import sys\n",
    "\n",
    "# Data directories\n",
    "datadir = \"/Volumes/drive/disc1/OAS1_0003_MR1/mri/\"\n",
    "fastsurferdir = \"/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer\"\n",
    "# Subject-specific variables\n",
    "subject = \"OAS_0003-1\"\n",
    "t1 = os.path.join(datadir, \"T1.mgz\")\n",
    "sd = os.path.join(fastsurferdir, subject)\n",
    "\n",
    "# DO NOT CHANGE BELOW\n",
    "\n",
    "# Dynamic FastSurfer directories \n",
    "fastsurfercnndir = os.path.join(fastsurferdir, \"FastSurferCNN\")\n",
    "reconsurfdir = os.path.join(fastsurferdir, \"recon_surf\")\n",
    "\n",
    "# Add FastSurferCNN directory to PYTHONPATH\n",
    "os.environ[\"PYTHONPATH\"] = f\"{fastsurferdir}:{os.environ.get('PYTHONPATH', '')}\"\n",
    "sys.path.append(fastsurferdir)\n",
    "\n",
    "# # Verify if FastSurferCNN module can be imported\n",
    "# try:\n",
    "#     import FastSurferCNN.reduce_to_aseg as rta\n",
    "#     print(\"FastSurferCNN module imported successfully\")\n",
    "# except ImportError as e:\n",
    "#     print(\"Failed to import FastSurferCNN module:\", e)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import sys\n",
    "\n",
    "# Set FreeSurfer environment variables\n",
    "os.environ[\"FREESURFER_HOME\"] = \"/Applications/freesurfer/7.4.1\"\n",
    "os.environ[\"SUBJECTS_DIR\"] = os.environ[\"FREESURFER_HOME\"] + \"/subjects\"\n",
    "os.environ[\"FUNCTIONALS_DIR\"] = os.environ[\"FREESURFER_HOME\"] + \"/sessions\"\n",
    "os.environ[\"FSFAST_HOME\"] = os.environ[\"FREESURFER_HOME\"] + \"/fsfast\"\n",
    "os.environ[\"FSF_OUTPUT_FORMAT\"] = \"nii.gz\"\n",
    "os.environ[\"LOCAL_DIR\"] = os.environ[\"FREESURFER_HOME\"] + \"/local\"\n",
    "os.environ[\"FSL_DIR\"] = \"/usr/local/fsl\"\n",
    "os.environ[\"FSL_BIN\"] = os.environ[\"FSL_DIR\"] + \"/bin\"\n",
    "os.environ[\"PATH\"] = f\"{os.environ['FREESURFER_HOME']}/bin:{os.environ['FSFAST_HOME']}/bin:{os.environ['FSL_BIN']}:{os.environ['PATH']}\"\n",
    "\n",
    "# Optional MINC toolkit setup\n",
    "if \"NO_MINC\" not in os.environ:\n",
    "    os.environ[\"MINC_BIN_DIR\"] = os.path.join(os.environ[\"FREESURFER_HOME\"], \"mni\", \"bin\")\n",
    "    os.environ[\"MINC_LIB_DIR\"] = os.path.join(os.environ[\"FREESURFER_HOME\"], \"mni\", \"lib\")\n",
    "    os.environ[\"MNI_DIR\"] = os.path.join(os.environ[\"FREESURFER_HOME\"], \"mni\")\n",
    "    os.environ[\"PATH\"] = f\"{os.environ['MINC_BIN_DIR']}:{os.environ['PATH']}\"\n",
    "\n",
    "# Optional: Set Perl library path for MINC toolkit\n",
    "if \"MNI_PERL5LIB\" not in os.environ:\n",
    "    mni_perl5lib = os.path.join(os.environ[\"FREESURFER_HOME\"], \"mni\", \"share\", \"perl5\")\n",
    "    os.environ[\"MNI_PERL5LIB\"] = mni_perl5lib\n",
    "    os.environ[\"PERL5LIB\"] = mni_perl5lib\n",
    "\n",
    "# Library paths for MacOS\n",
    "if sys.platform == \"darwin\":\n",
    "    misc_lib = os.path.join(os.environ[\"FREESURFER_HOME\"], \"lib\", \"misc\", \"lib\")\n",
    "    os.environ[\"DYLD_LIBRARY_PATH\"] = f\"{misc_lib}:{os.environ.get('DYLD_LIBRARY_PATH', '')}\"\n",
    "\n",
    "# Set up FSL environment\n",
    "if \"FSL_DIR\" in os.environ:\n",
    "    os.environ[\"FSLDIR\"] = os.environ[\"FSL_DIR\"]\n",
    "    os.environ[\"FSLOUTPUTTYPE\"] = \"NIFTI_GZ\"\n",
    "    if \"FSL_BIN\" in os.environ:\n",
    "        os.environ[\"PATH\"] = f\"{os.environ['FSL_BIN']}:{os.environ['PATH']}\"\n",
    "\n",
    "# Add FastSurferCNN directory to PYTHONPATH\n",
    "fastsurferdir = \"/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer\"\n",
    "os.environ[\"PYTHONPATH\"] = f\"{fastsurferdir}:{os.environ.get('PYTHONPATH', '')}\"\n",
    "sys.path.append(fastsurferdir)\n",
    "\n",
    "# Verify the environment setup\n",
    "print(\"Environment variables set for FreeSurfer:\")\n",
    "print(\"FREESURFER_HOME:\", os.environ[\"FREESURFER_HOME\"])\n",
    "print(\"SUBJECTS_DIR:\", os.environ[\"SUBJECTS_DIR\"])\n",
    "print(\"FSFAST_HOME:\", os.environ[\"FSFAST_HOME\"])\n",
    "print(\"FSF_OUTPUT_FORMAT:\", os.environ[\"FSF_OUTPUT_FORMAT\"])\n",
    "print(\"FSL_DIR:\", os.environ.get(\"FSL_DIR\"))\n",
    "print(\"MINC_BIN_DIR:\", os.environ.get(\"MINC_BIN_DIR\"))\n",
    "print(\"MINC_LIB_DIR:\", os.environ.get(\"MINC_LIB_DIR\"))\n",
    "print(\"MNI_DIR:\", os.environ.get(\"MNI_DIR\"))\n",
    "print(\"PYTHONPATH:\", os.environ[\"PYTHONPATH\"])\n",
    "print(\"PATH:\", os.environ[\"PATH\"])\n",
    "print(\"DYLD_LIBRARY_PATH:\", os.environ.get(\"DYLD_LIBRARY_PATH\"))\n",
    "print(\"PERL5LIB:\", os.environ.get(\"PERL5LIB\"))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "FREESURFER_HOME: /Applications/freesurfer/7.4.1\n",
      "PYTHONPATH: /Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer:/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer:/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer:/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer:/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer:/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer:/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer:/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer:/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer:/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer/FastSurferCNN:/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer/FastSurferCNN:/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer/FastSurferCNN:/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer/FastSurferCNN:/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer/FastSurferCNN:/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer/FastSurferCNN:/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer/FastSurferCNN:/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer/FastSurferCNN:\n",
      "sys.path: ['/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH', '/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer/fastsurferenv/lib/python310.zip', '/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer/fastsurferenv/lib/python3.10', '/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer/fastsurferenv/lib/python3.10/lib-dynload', '', '/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer/fastsurferenv/lib/python3.10/site-packages', '/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer', '/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer', '/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer', '/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer', '/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer', '/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer', '/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer', '/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer', '/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer', '/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer', '/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer', '/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer', '/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer', '/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer', '/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer', '/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer', '/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer']\n",
      "Data Directory: /Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer/OAS1_0001/mri\n",
      "FastSurfer Directory: /Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer\n",
      "FastSurferCNN Directory: /Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer/FastSurferCNN\n",
      "ReconSurf Directory: /Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer/recon_surf\n",
      "Segmentation completed successfully.\n",
      "stdout: [INFO: run_prediction.py:  653]: Checking or downloading default checkpoints ...\n",
      "[INFO: common.py:  831]: Single subject with absolute file path for input.\n",
      "[INFO: common.py:  889]: Analyzing single subject /Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer/OAS1_0001/mri/t1.mgz\n",
      "[INFO: common.py:  990]: Output will be stored in Subjects Directory: /Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer\n",
      "[INFO: common.py:  114]: Using device: cpu\n",
      "[INFO: run_prediction.py:  246]: Running view aggregation on cpu\n",
      "[INFO: inference.py:  200]: Loading checkpoint /Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer/checkpoints/aparc_vinn_coronal_v2.0.0.pkl\n",
      "[INFO: inference.py:  200]: Loading checkpoint /Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer/checkpoints/aparc_vinn_sagittal_v2.0.0.pkl\n",
      "[INFO: inference.py:  200]: Loading checkpoint /Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer/checkpoints/aparc_vinn_axial_v2.0.0.pkl\n",
      "[INFO: run_prediction.py:  325]: Successfully loaded image from /Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer/OAS1_0001/mri/t1.mgz.\n",
      "[INFO: run_prediction.py:  455]: Successfully saved image   as /Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer/OAS_0001-2/mri/orig/001.mgz.\n",
      "A min-conformed image must satisfy the following criteria:\n",
      " - Number of Dimensions 3        : True\n",
      " - Dimensions 256x256x256        : True\n",
      " - Voxel Size 1.0x1.0x1.0        : True\n",
      " - Orientation LIA               : True\n",
      " - Dtype uint8                   : True\n",
      "[INFO: run_prediction.py:  455]: Successfully saved image   as /Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer/OAS_0001-2/mri/orig.mgz.\n",
      "[INFO: run_prediction.py:  401]: Run coronal prediction\n",
      "[INFO: dataset.py:   78]: Loading Coronal with input voxelsize (1.0, 1.0)\n",
      "[INFO: inference.py:  404]: Inference on 256 batches for coronal successful\n",
      "[INFO: inference.py:  466]: Coronal inference on /Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer/OAS1_0001/mri/t1.mgz finished in 328.2420 seconds\n",
      "[INFO: run_prediction.py:  401]: Run sagittal prediction\n",
      "[INFO: dataset.py:   69]: Loading Sagittal with input voxelsize (1.0, 1.0)\n",
      "[INFO: inference.py:  404]: Inference on 256 batches for sagittal successful\n",
      "[INFO: inference.py:  466]: Sagittal inference on /Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer/OAS1_0001/mri/t1.mgz finished in 326.7562 seconds\n",
      "[INFO: run_prediction.py:  401]: Run axial prediction\n",
      "[INFO: dataset.py:   74]: Loading Axial with input voxelsize (1.0, 1.0)\n",
      "[INFO: inference.py:  404]: Inference on 256 batches for axial successful\n",
      "[INFO: inference.py:  466]: Axial inference on /Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer/OAS1_0001/mri/t1.mgz finished in 363.8923 seconds\n",
      "[INFO: run_prediction.py:  455]: Successfully saved image   as /Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer/OAS_0001-2/mri/aparc.DKTatlas+aseg.deep.mgz.\n",
      "[INFO: run_prediction.py:  730]: Creating brainmask based on segmentation...\n",
      "Creating dilated mask ...\n",
      "Frontal region special treatment:  18347\n",
      "  Found 1 connected component(s)!\n",
      "[INFO: run_prediction.py:  455]: Successfully saved image   as /Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer/OAS_0001-2/mri/mask.mgz.\n",
      "[INFO: run_prediction.py:  749]: Creating aseg based on segmentation...\n",
      "Reducing to aseg ...\n",
      "FlipWM: rh 0 and lh 0 flipped.\n",
      "[INFO: run_prediction.py:  455]: Successfully saved image   as /Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer/OAS_0001-2/mri/aseg.auto_noCCseg.mgz.\n",
      "[INFO: run_prediction.py:  768]: Running volume-based QC check on segmentation...\n",
      "Checking total volume ...\n",
      "Voxel size in mm3: 1.0\n",
      "Total segmentation volume in liter: 0.98\n",
      "\n",
      "stderr: \n",
      "  0%|          | 0/256 [00:00<?, ?batch/s]\n",
      "  0%|          | 1/256 [00:01<06:04,  1.43s/batch]\n",
      "  1%|          | 2/256 [00:02<05:28,  1.29s/batch]\n",
      "  1%|          | 3/256 [00:03<05:05,  1.21s/batch]\n",
      "  2%|▏         | 4/256 [00:04<04:55,  1.17s/batch]\n",
      "  2%|▏         | 5/256 [00:05<04:48,  1.15s/batch]\n",
      "  2%|▏         | 6/256 [00:07<04:44,  1.14s/batch]\n",
      "  3%|▎         | 7/256 [00:08<04:58,  1.20s/batch]\n",
      "  3%|▎         | 8/256 [00:10<05:52,  1.42s/batch]\n",
      "  4%|▎         | 9/256 [00:11<05:59,  1.46s/batch]\n",
      "  4%|▍         | 10/256 [00:13<05:52,  1.43s/batch]\n",
      "  4%|▍         | 11/256 [00:14<05:45,  1.41s/batch]\n",
      "  5%|▍         | 12/256 [00:15<05:32,  1.36s/batch]\n",
      "  5%|▌         | 13/256 [00:17<05:20,  1.32s/batch]\n",
      "  5%|▌         | 14/256 [00:18<05:20,  1.32s/batch]\n",
      "  6%|▌         | 15/256 [00:19<05:08,  1.28s/batch]\n",
      "  6%|▋         | 16/256 [00:21<05:32,  1.38s/batch]\n",
      "  7%|▋         | 17/256 [00:22<05:25,  1.36s/batch]\n",
      "  7%|▋         | 18/256 [00:23<05:18,  1.34s/batch]\n",
      "  7%|▋         | 19/256 [00:24<05:01,  1.27s/batch]\n",
      "  8%|▊         | 20/256 [00:26<05:00,  1.27s/batch]\n",
      "  8%|▊         | 21/256 [00:27<04:56,  1.26s/batch]\n",
      "  9%|▊         | 22/256 [00:28<04:54,  1.26s/batch]\n",
      "  9%|▉         | 23/256 [00:30<05:05,  1.31s/batch]\n",
      "  9%|▉         | 24/256 [00:31<05:06,  1.32s/batch]\n",
      " 10%|▉         | 25/256 [00:32<05:11,  1.35s/batch]\n",
      " 10%|█         | 26/256 [00:34<05:09,  1.34s/batch]\n",
      " 11%|█         | 27/256 [00:35<04:54,  1.29s/batch]\n",
      " 11%|█         | 28/256 [00:36<04:39,  1.22s/batch]\n",
      " 11%|█▏        | 29/256 [00:37<04:25,  1.17s/batch]\n",
      " 12%|█▏        | 30/256 [00:38<04:14,  1.13s/batch]\n",
      " 12%|█▏        | 31/256 [00:39<04:05,  1.09s/batch]\n",
      " 12%|█▎        | 32/256 [00:40<04:03,  1.09s/batch]\n",
      " 13%|█▎        | 33/256 [00:41<04:01,  1.08s/batch]\n",
      " 13%|█▎        | 34/256 [00:42<04:02,  1.09s/batch]\n",
      " 14%|█▎        | 35/256 [00:43<03:57,  1.08s/batch]\n",
      " 14%|█▍        | 36/256 [00:44<03:52,  1.06s/batch]\n",
      " 14%|█▍        | 37/256 [00:45<03:48,  1.04s/batch]\n",
      " 15%|█▍        | 38/256 [00:46<03:43,  1.03s/batch]\n",
      " 15%|█▌        | 39/256 [00:47<03:43,  1.03s/batch]\n",
      " 16%|█▌        | 40/256 [00:48<03:45,  1.04s/batch]\n",
      " 16%|█▌        | 41/256 [00:49<03:46,  1.05s/batch]\n",
      " 16%|█▋        | 42/256 [00:51<03:46,  1.06s/batch]\n",
      " 17%|█▋        | 43/256 [00:52<03:47,  1.07s/batch]\n",
      " 17%|█▋        | 44/256 [00:53<03:45,  1.06s/batch]\n",
      " 18%|█▊        | 45/256 [00:54<03:40,  1.05s/batch]\n",
      " 18%|█▊        | 46/256 [00:55<03:38,  1.04s/batch]\n",
      " 18%|█▊        | 47/256 [00:56<03:37,  1.04s/batch]\n",
      " 19%|█▉        | 48/256 [00:57<03:37,  1.05s/batch]\n",
      " 19%|█▉        | 49/256 [00:58<03:37,  1.05s/batch]\n",
      " 20%|█▉        | 50/256 [00:59<03:35,  1.04s/batch]\n",
      " 20%|█▉        | 51/256 [01:00<03:34,  1.05s/batch]\n",
      " 20%|██        | 52/256 [01:01<03:31,  1.03s/batch]\n",
      " 21%|██        | 53/256 [01:02<03:28,  1.02s/batch]\n",
      " 21%|██        | 54/256 [01:03<03:26,  1.02s/batch]\n",
      " 21%|██▏       | 55/256 [01:04<03:25,  1.02s/batch]\n",
      " 22%|██▏       | 56/256 [01:05<03:22,  1.01s/batch]\n",
      " 22%|██▏       | 57/256 [01:06<03:20,  1.01s/batch]\n",
      " 23%|██▎       | 58/256 [01:07<03:35,  1.09s/batch]\n",
      " 23%|██▎       | 59/256 [01:08<03:41,  1.12s/batch]\n",
      " 23%|██▎       | 60/256 [01:10<03:42,  1.14s/batch]\n",
      " 24%|██▍       | 61/256 [01:11<03:50,  1.18s/batch]\n",
      " 24%|██▍       | 62/256 [01:12<03:57,  1.23s/batch]\n",
      " 25%|██▍       | 63/256 [01:14<04:10,  1.30s/batch]\n",
      " 25%|██▌       | 64/256 [01:15<04:29,  1.40s/batch]\n",
      " 25%|██▌       | 65/256 [01:17<04:46,  1.50s/batch]\n",
      " 26%|██▌       | 66/256 [01:19<05:04,  1.60s/batch]\n",
      " 26%|██▌       | 67/256 [01:21<05:22,  1.71s/batch]\n",
      " 27%|██▋       | 68/256 [01:23<05:14,  1.67s/batch]\n",
      " 27%|██▋       | 69/256 [01:24<05:18,  1.70s/batch]\n",
      " 27%|██▋       | 70/256 [01:26<05:07,  1.65s/batch]\n",
      " 28%|██▊       | 71/256 [01:28<05:13,  1.70s/batch]\n",
      " 28%|██▊       | 72/256 [01:30<05:24,  1.77s/batch]\n",
      " 29%|██▊       | 73/256 [01:31<05:25,  1.78s/batch]\n",
      " 29%|██▉       | 74/256 [01:33<05:15,  1.74s/batch]\n",
      " 29%|██▉       | 75/256 [01:35<05:11,  1.72s/batch]\n",
      " 30%|██▉       | 76/256 [01:36<05:07,  1.71s/batch]\n",
      " 30%|███       | 77/256 [01:38<04:57,  1.66s/batch]\n",
      " 30%|███       | 78/256 [01:40<05:10,  1.74s/batch]\n",
      " 31%|███       | 79/256 [01:41<04:53,  1.66s/batch]\n",
      " 31%|███▏      | 80/256 [01:43<04:53,  1.67s/batch]\n",
      " 32%|███▏      | 81/256 [01:44<04:36,  1.58s/batch]\n",
      " 32%|███▏      | 82/256 [01:46<04:25,  1.53s/batch]\n",
      " 32%|███▏      | 83/256 [01:47<04:08,  1.43s/batch]\n",
      " 33%|███▎      | 84/256 [01:48<04:02,  1.41s/batch]\n",
      " 33%|███▎      | 85/256 [01:50<03:55,  1.38s/batch]\n",
      " 34%|███▎      | 86/256 [01:51<03:51,  1.36s/batch]\n",
      " 34%|███▍      | 87/256 [01:52<03:43,  1.32s/batch]\n",
      " 34%|███▍      | 88/256 [01:54<03:44,  1.33s/batch]\n",
      " 35%|███▍      | 89/256 [01:55<03:37,  1.30s/batch]\n",
      " 35%|███▌      | 90/256 [01:56<03:32,  1.28s/batch]\n",
      " 36%|███▌      | 91/256 [01:57<03:28,  1.26s/batch]\n",
      " 36%|███▌      | 92/256 [01:59<03:29,  1.28s/batch]\n",
      " 36%|███▋      | 93/256 [02:00<03:17,  1.21s/batch]\n",
      " 37%|███▋      | 94/256 [02:01<03:08,  1.17s/batch]\n",
      " 37%|███▋      | 95/256 [02:02<03:08,  1.17s/batch]\n",
      " 38%|███▊      | 96/256 [02:03<03:03,  1.14s/batch]\n",
      " 38%|███▊      | 97/256 [02:04<03:04,  1.16s/batch]\n",
      " 38%|███▊      | 98/256 [02:05<03:03,  1.16s/batch]\n",
      " 39%|███▊      | 99/256 [02:06<02:59,  1.14s/batch]\n",
      " 39%|███▉      | 100/256 [02:08<02:56,  1.13s/batch]\n",
      " 39%|███▉      | 101/256 [02:09<02:54,  1.13s/batch]\n",
      " 40%|███▉      | 102/256 [02:10<02:58,  1.16s/batch]\n",
      " 40%|████      | 103/256 [02:11<03:05,  1.22s/batch]\n",
      " 41%|████      | 104/256 [02:12<03:00,  1.19s/batch]\n",
      " 41%|████      | 105/256 [02:14<02:59,  1.19s/batch]\n",
      " 41%|████▏     | 106/256 [02:15<03:04,  1.23s/batch]\n",
      " 42%|████▏     | 107/256 [02:16<03:02,  1.22s/batch]\n",
      " 42%|████▏     | 108/256 [02:17<03:00,  1.22s/batch]\n",
      " 43%|████▎     | 109/256 [02:18<02:59,  1.22s/batch]\n",
      " 43%|████▎     | 110/256 [02:20<02:54,  1.19s/batch]\n",
      " 43%|████▎     | 111/256 [02:21<02:50,  1.18s/batch]\n",
      " 44%|████▍     | 112/256 [02:22<03:01,  1.26s/batch]\n",
      " 44%|████▍     | 113/256 [02:24<03:10,  1.33s/batch]\n",
      " 45%|████▍     | 114/256 [02:25<03:11,  1.35s/batch]\n",
      " 45%|████▍     | 115/256 [02:26<03:07,  1.33s/batch]\n",
      " 45%|████▌     | 116/256 [02:28<03:36,  1.55s/batch]\n",
      " 46%|████▌     | 117/256 [02:30<03:44,  1.62s/batch]\n",
      " 46%|████▌     | 118/256 [02:32<03:55,  1.71s/batch]\n",
      " 46%|████▋     | 119/256 [02:34<03:54,  1.71s/batch]\n",
      " 47%|████▋     | 120/256 [02:36<04:08,  1.83s/batch]\n",
      " 47%|████▋     | 121/256 [02:38<04:20,  1.93s/batch]\n",
      " 48%|████▊     | 122/256 [02:41<04:41,  2.10s/batch]\n",
      " 48%|████▊     | 123/256 [02:42<03:59,  1.80s/batch]\n",
      " 48%|████▊     | 124/256 [02:43<03:37,  1.65s/batch]\n",
      " 49%|████▉     | 125/256 [02:45<03:50,  1.76s/batch]\n",
      " 49%|████▉     | 126/256 [02:47<03:51,  1.78s/batch]\n",
      " 50%|████▉     | 127/256 [02:48<03:39,  1.70s/batch]\n",
      " 50%|█████     | 128/256 [02:50<03:26,  1.61s/batch]\n",
      " 50%|█████     | 129/256 [02:51<03:16,  1.54s/batch]\n",
      " 51%|█████     | 130/256 [02:53<03:13,  1.54s/batch]\n",
      " 51%|█████     | 131/256 [02:54<03:09,  1.52s/batch]\n",
      " 52%|█████▏    | 132/256 [02:56<03:03,  1.48s/batch]\n",
      " 52%|█████▏    | 133/256 [02:57<03:14,  1.58s/batch]\n",
      " 52%|█████▏    | 134/256 [02:59<03:09,  1.55s/batch]\n",
      " 53%|█████▎    | 135/256 [03:00<03:05,  1.53s/batch]\n",
      " 53%|█████▎    | 136/256 [03:02<03:22,  1.69s/batch]\n",
      " 54%|█████▎    | 137/256 [03:04<03:01,  1.53s/batch]\n",
      " 54%|█████▍    | 138/256 [03:05<02:44,  1.39s/batch]\n",
      " 54%|█████▍    | 139/256 [03:06<02:30,  1.29s/batch]\n",
      " 55%|█████▍    | 140/256 [03:07<02:21,  1.22s/batch]\n",
      " 55%|█████▌    | 141/256 [03:08<02:17,  1.20s/batch]\n",
      " 55%|█████▌    | 142/256 [03:09<02:13,  1.18s/batch]\n",
      " 56%|█████▌    | 143/256 [03:10<02:17,  1.22s/batch]\n",
      " 56%|█████▋    | 144/256 [03:11<02:13,  1.19s/batch]\n",
      " 57%|█████▋    | 145/256 [03:13<02:16,  1.23s/batch]\n",
      " 57%|█████▋    | 146/256 [03:14<02:22,  1.30s/batch]\n",
      " 57%|█████▋    | 147/256 [03:15<02:17,  1.26s/batch]\n",
      " 58%|█████▊    | 148/256 [03:17<02:19,  1.29s/batch]\n",
      " 58%|█████▊    | 149/256 [03:18<02:30,  1.41s/batch]\n",
      " 59%|█████▊    | 150/256 [03:20<02:27,  1.39s/batch]\n",
      " 59%|█████▉    | 151/256 [03:21<02:21,  1.35s/batch]\n",
      " 59%|█████▉    | 152/256 [03:23<02:25,  1.40s/batch]\n",
      " 60%|█████▉    | 153/256 [03:24<02:24,  1.40s/batch]\n",
      " 60%|██████    | 154/256 [03:25<02:23,  1.40s/batch]\n",
      " 61%|██████    | 155/256 [03:27<02:30,  1.49s/batch]\n",
      " 61%|██████    | 156/256 [03:28<02:19,  1.40s/batch]\n",
      " 61%|██████▏   | 157/256 [03:30<02:19,  1.41s/batch]\n",
      " 62%|██████▏   | 158/256 [03:31<02:19,  1.43s/batch]\n",
      " 62%|██████▏   | 159/256 [03:33<02:16,  1.41s/batch]\n",
      " 62%|██████▎   | 160/256 [03:34<02:07,  1.33s/batch]\n",
      " 63%|██████▎   | 161/256 [03:35<02:00,  1.27s/batch]\n",
      " 63%|██████▎   | 162/256 [03:36<01:53,  1.20s/batch]\n",
      " 64%|██████▎   | 163/256 [03:37<02:01,  1.31s/batch]\n",
      " 64%|██████▍   | 164/256 [03:39<01:58,  1.28s/batch]\n",
      " 64%|██████▍   | 165/256 [03:40<01:59,  1.31s/batch]\n",
      " 65%|██████▍   | 166/256 [03:41<01:51,  1.24s/batch]\n",
      " 65%|██████▌   | 167/256 [03:42<01:46,  1.20s/batch]\n",
      " 66%|██████▌   | 168/256 [03:43<01:41,  1.16s/batch]\n",
      " 66%|██████▌   | 169/256 [03:44<01:37,  1.13s/batch]\n",
      " 66%|██████▋   | 170/256 [03:45<01:35,  1.11s/batch]\n",
      " 67%|██████▋   | 171/256 [03:46<01:33,  1.10s/batch]\n",
      " 67%|██████▋   | 172/256 [03:48<01:38,  1.17s/batch]\n",
      " 68%|██████▊   | 173/256 [03:49<01:36,  1.17s/batch]\n",
      " 68%|██████▊   | 174/256 [03:50<01:33,  1.14s/batch]\n",
      " 68%|██████▊   | 175/256 [03:51<01:30,  1.12s/batch]\n",
      " 69%|██████▉   | 176/256 [03:52<01:27,  1.10s/batch]\n",
      " 69%|██████▉   | 177/256 [03:54<01:33,  1.18s/batch]\n",
      " 70%|██████▉   | 178/256 [03:55<01:31,  1.18s/batch]\n",
      " 70%|██████▉   | 179/256 [03:56<01:28,  1.15s/batch]\n",
      " 70%|███████   | 180/256 [03:57<01:24,  1.12s/batch]\n",
      " 71%|███████   | 181/256 [03:58<01:26,  1.16s/batch]\n",
      " 71%|███████   | 182/256 [03:59<01:25,  1.16s/batch]\n",
      " 71%|███████▏  | 183/256 [04:00<01:23,  1.14s/batch]\n",
      " 72%|███████▏  | 184/256 [04:01<01:20,  1.11s/batch]\n",
      " 72%|███████▏  | 185/256 [04:02<01:18,  1.10s/batch]\n",
      " 73%|███████▎  | 186/256 [04:03<01:16,  1.09s/batch]\n",
      " 73%|███████▎  | 187/256 [04:05<01:14,  1.09s/batch]\n",
      " 73%|███████▎  | 188/256 [04:06<01:14,  1.09s/batch]\n",
      " 74%|███████▍  | 189/256 [04:07<01:13,  1.10s/batch]\n",
      " 74%|███████▍  | 190/256 [04:08<01:13,  1.11s/batch]\n",
      " 75%|███████▍  | 191/256 [04:09<01:11,  1.10s/batch]\n",
      " 75%|███████▌  | 192/256 [04:10<01:11,  1.12s/batch]\n",
      " 75%|███████▌  | 193/256 [04:11<01:11,  1.14s/batch]\n",
      " 76%|███████▌  | 194/256 [04:12<01:10,  1.13s/batch]\n",
      " 76%|███████▌  | 195/256 [04:14<01:11,  1.17s/batch]\n",
      " 77%|███████▋  | 196/256 [04:15<01:10,  1.17s/batch]\n",
      " 77%|███████▋  | 197/256 [04:16<01:08,  1.16s/batch]\n",
      " 77%|███████▋  | 198/256 [04:17<01:06,  1.15s/batch]\n",
      " 78%|███████▊  | 199/256 [04:18<01:06,  1.17s/batch]\n",
      " 78%|███████▊  | 200/256 [04:19<01:04,  1.14s/batch]\n",
      " 79%|███████▊  | 201/256 [04:21<01:02,  1.13s/batch]\n",
      " 79%|███████▉  | 202/256 [04:22<01:01,  1.13s/batch]\n",
      " 79%|███████▉  | 203/256 [04:23<00:59,  1.12s/batch]\n",
      " 80%|███████▉  | 204/256 [04:24<00:57,  1.11s/batch]\n",
      " 80%|████████  | 205/256 [04:25<00:56,  1.10s/batch]\n",
      " 80%|████████  | 206/256 [04:26<00:55,  1.10s/batch]\n",
      " 81%|████████  | 207/256 [04:27<00:54,  1.11s/batch]\n",
      " 81%|████████▏ | 208/256 [04:28<00:53,  1.11s/batch]\n",
      " 82%|████████▏ | 209/256 [04:29<00:52,  1.11s/batch]\n",
      " 82%|████████▏ | 210/256 [04:30<00:50,  1.09s/batch]\n",
      " 82%|████████▏ | 211/256 [04:32<00:48,  1.08s/batch]\n",
      " 83%|████████▎ | 212/256 [04:33<00:48,  1.11s/batch]\n",
      " 83%|████████▎ | 213/256 [04:34<00:51,  1.19s/batch]\n",
      " 84%|████████▎ | 214/256 [04:35<00:51,  1.22s/batch]\n",
      " 84%|████████▍ | 215/256 [04:37<00:54,  1.33s/batch]\n",
      " 84%|████████▍ | 216/256 [04:38<00:51,  1.30s/batch]\n",
      " 85%|████████▍ | 217/256 [04:39<00:49,  1.28s/batch]\n",
      " 85%|████████▌ | 218/256 [04:41<00:48,  1.27s/batch]\n",
      " 86%|████████▌ | 219/256 [04:42<00:46,  1.27s/batch]\n",
      " 86%|████████▌ | 220/256 [04:43<00:46,  1.29s/batch]\n",
      " 86%|████████▋ | 221/256 [04:45<00:45,  1.29s/batch]\n",
      " 87%|████████▋ | 222/256 [04:46<00:42,  1.24s/batch]\n",
      " 87%|████████▋ | 223/256 [04:47<00:40,  1.21s/batch]\n",
      " 88%|████████▊ | 224/256 [04:48<00:39,  1.22s/batch]\n",
      " 88%|████████▊ | 225/256 [04:49<00:38,  1.23s/batch]\n",
      " 88%|████████▊ | 226/256 [04:51<00:37,  1.24s/batch]\n",
      " 89%|████████▊ | 227/256 [04:52<00:36,  1.26s/batch]\n",
      " 89%|████████▉ | 228/256 [04:53<00:36,  1.29s/batch]\n",
      " 89%|████████▉ | 229/256 [04:54<00:33,  1.25s/batch]\n",
      " 90%|████████▉ | 230/256 [04:56<00:32,  1.24s/batch]\n",
      " 90%|█████████ | 231/256 [04:57<00:34,  1.36s/batch]\n",
      " 91%|█████████ | 232/256 [04:59<00:32,  1.35s/batch]\n",
      " 91%|█████████ | 233/256 [05:00<00:30,  1.34s/batch]\n",
      " 91%|█████████▏| 234/256 [05:01<00:30,  1.38s/batch]\n",
      " 92%|█████████▏| 235/256 [05:03<00:29,  1.38s/batch]\n",
      " 92%|█████████▏| 236/256 [05:04<00:27,  1.38s/batch]\n",
      " 93%|█████████▎| 237/256 [05:05<00:25,  1.36s/batch]\n",
      " 93%|█████████▎| 238/256 [05:07<00:23,  1.31s/batch]\n",
      " 93%|█████████▎| 239/256 [05:08<00:22,  1.30s/batch]\n",
      " 94%|█████████▍| 240/256 [05:09<00:20,  1.29s/batch]\n",
      " 94%|█████████▍| 241/256 [05:11<00:20,  1.36s/batch]\n",
      " 95%|█████████▍| 242/256 [05:12<00:17,  1.27s/batch]\n",
      " 95%|█████████▍| 243/256 [05:13<00:16,  1.25s/batch]\n",
      " 95%|█████████▌| 244/256 [05:14<00:14,  1.24s/batch]\n",
      " 96%|█████████▌| 245/256 [05:15<00:13,  1.23s/batch]\n",
      " 96%|█████████▌| 246/256 [05:16<00:11,  1.18s/batch]\n",
      " 96%|█████████▋| 247/256 [05:18<00:10,  1.15s/batch]\n",
      " 97%|█████████▋| 248/256 [05:19<00:09,  1.13s/batch]\n",
      " 97%|█████████▋| 249/256 [05:20<00:07,  1.13s/batch]\n",
      " 98%|█████████▊| 250/256 [05:21<00:06,  1.13s/batch]\n",
      " 98%|█████████▊| 251/256 [05:22<00:05,  1.12s/batch]\n",
      " 98%|█████████▊| 252/256 [05:23<00:04,  1.12s/batch]\n",
      " 99%|█████████▉| 253/256 [05:24<00:03,  1.13s/batch]\n",
      " 99%|█████████▉| 254/256 [05:25<00:02,  1.15s/batch]\n",
      "100%|█████████▉| 255/256 [05:27<00:01,  1.13s/batch]\n",
      "100%|██████████| 256/256 [05:28<00:00,  1.13s/batch]\n",
      "100%|██████████| 256/256 [05:28<00:00,  1.28s/batch]\n",
      "\n",
      "  0%|          | 0/256 [00:00<?, ?batch/s]\n",
      "  0%|          | 1/256 [00:01<04:47,  1.13s/batch]\n",
      "  1%|          | 2/256 [00:02<04:33,  1.08s/batch]\n",
      "  1%|          | 3/256 [00:03<04:30,  1.07s/batch]\n",
      "  2%|▏         | 4/256 [00:04<05:10,  1.23s/batch]\n",
      "  2%|▏         | 5/256 [00:06<05:48,  1.39s/batch]\n",
      "  2%|▏         | 6/256 [00:08<06:19,  1.52s/batch]\n",
      "  3%|▎         | 7/256 [00:09<05:57,  1.44s/batch]\n",
      "  3%|▎         | 8/256 [00:10<05:43,  1.39s/batch]\n",
      "  4%|▎         | 9/256 [00:11<05:30,  1.34s/batch]\n",
      "  4%|▍         | 10/256 [00:13<05:13,  1.28s/batch]\n",
      "  4%|▍         | 11/256 [00:14<05:23,  1.32s/batch]\n",
      "  5%|▍         | 12/256 [00:15<05:27,  1.34s/batch]\n",
      "  5%|▌         | 13/256 [00:17<05:23,  1.33s/batch]\n",
      "  5%|▌         | 14/256 [00:18<05:27,  1.35s/batch]\n",
      "  6%|▌         | 15/256 [00:19<05:15,  1.31s/batch]\n",
      "  6%|▋         | 16/256 [00:21<05:12,  1.30s/batch]\n",
      "  7%|▋         | 17/256 [00:22<05:06,  1.28s/batch]\n",
      "  7%|▋         | 18/256 [00:23<05:31,  1.39s/batch]\n",
      "  7%|▋         | 19/256 [00:25<05:12,  1.32s/batch]\n",
      "  8%|▊         | 20/256 [00:26<04:57,  1.26s/batch]\n",
      "  8%|▊         | 21/256 [00:27<05:04,  1.30s/batch]\n",
      "  9%|▊         | 22/256 [00:28<05:04,  1.30s/batch]\n",
      "  9%|▉         | 23/256 [00:30<05:07,  1.32s/batch]\n",
      "  9%|▉         | 24/256 [00:31<05:14,  1.35s/batch]\n",
      " 10%|▉         | 25/256 [00:33<05:32,  1.44s/batch]\n",
      " 10%|█         | 26/256 [00:34<05:28,  1.43s/batch]\n",
      " 11%|█         | 27/256 [00:36<05:28,  1.43s/batch]\n",
      " 11%|█         | 28/256 [00:37<05:28,  1.44s/batch]\n",
      " 11%|█▏        | 29/256 [00:39<05:41,  1.51s/batch]\n",
      " 12%|█▏        | 30/256 [00:40<05:46,  1.53s/batch]\n",
      " 12%|█▏        | 31/256 [00:42<06:18,  1.68s/batch]\n",
      " 12%|█▎        | 32/256 [00:44<06:02,  1.62s/batch]\n",
      " 13%|█▎        | 33/256 [00:45<05:40,  1.53s/batch]\n",
      " 13%|█▎        | 34/256 [00:47<05:33,  1.50s/batch]\n",
      " 14%|█▎        | 35/256 [00:48<05:09,  1.40s/batch]\n",
      " 14%|█▍        | 36/256 [00:49<04:55,  1.34s/batch]\n",
      " 14%|█▍        | 37/256 [00:51<05:08,  1.41s/batch]\n",
      " 15%|█▍        | 38/256 [00:52<05:08,  1.42s/batch]\n",
      " 15%|█▌        | 39/256 [00:54<05:14,  1.45s/batch]\n",
      " 16%|█▌        | 40/256 [00:56<05:56,  1.65s/batch]\n",
      " 16%|█▌        | 41/256 [00:58<06:37,  1.85s/batch]\n",
      " 16%|█▋        | 42/256 [01:00<06:27,  1.81s/batch]\n",
      " 17%|█▋        | 43/256 [01:01<06:15,  1.76s/batch]\n",
      " 17%|█▋        | 44/256 [01:03<05:57,  1.69s/batch]\n",
      " 18%|█▊        | 45/256 [01:05<06:23,  1.82s/batch]\n",
      " 18%|█▊        | 46/256 [01:07<06:12,  1.78s/batch]\n",
      " 18%|█▊        | 47/256 [01:09<06:21,  1.83s/batch]\n",
      " 19%|█▉        | 48/256 [01:10<06:04,  1.75s/batch]\n",
      " 19%|█▉        | 49/256 [01:12<05:51,  1.70s/batch]\n",
      " 20%|█▉        | 50/256 [01:13<05:34,  1.62s/batch]\n",
      " 20%|█▉        | 51/256 [01:15<06:10,  1.81s/batch]\n",
      " 20%|██        | 52/256 [01:17<06:00,  1.77s/batch]\n",
      " 21%|██        | 53/256 [01:19<06:12,  1.84s/batch]\n",
      " 21%|██        | 54/256 [01:22<06:43,  2.00s/batch]\n",
      " 21%|██▏       | 55/256 [01:23<06:20,  1.89s/batch]\n",
      " 22%|██▏       | 56/256 [01:25<05:53,  1.77s/batch]\n",
      " 22%|██▏       | 57/256 [01:26<05:35,  1.68s/batch]\n",
      " 23%|██▎       | 58/256 [01:28<05:58,  1.81s/batch]\n",
      " 23%|██▎       | 59/256 [01:30<06:17,  1.92s/batch]\n",
      " 23%|██▎       | 60/256 [01:32<05:56,  1.82s/batch]\n",
      " 24%|██▍       | 61/256 [01:34<05:39,  1.74s/batch]\n",
      " 24%|██▍       | 62/256 [01:35<05:37,  1.74s/batch]\n",
      " 25%|██▍       | 63/256 [01:37<05:22,  1.67s/batch]\n",
      " 25%|██▌       | 64/256 [01:38<05:00,  1.57s/batch]\n",
      " 25%|██▌       | 65/256 [01:39<04:44,  1.49s/batch]\n",
      " 26%|██▌       | 66/256 [01:41<04:34,  1.45s/batch]\n",
      " 26%|██▌       | 67/256 [01:42<04:27,  1.41s/batch]\n",
      " 27%|██▋       | 68/256 [01:43<04:18,  1.38s/batch]\n",
      " 27%|██▋       | 69/256 [01:45<04:16,  1.37s/batch]\n",
      " 27%|██▋       | 70/256 [01:46<04:07,  1.33s/batch]\n",
      " 28%|██▊       | 71/256 [01:47<03:59,  1.29s/batch]\n",
      " 28%|██▊       | 72/256 [01:48<03:57,  1.29s/batch]\n",
      " 29%|██▊       | 73/256 [01:50<03:55,  1.29s/batch]\n",
      " 29%|██▉       | 74/256 [01:51<03:57,  1.31s/batch]\n",
      " 29%|██▉       | 75/256 [01:52<03:51,  1.28s/batch]\n",
      " 30%|██▉       | 76/256 [01:54<03:48,  1.27s/batch]\n",
      " 30%|███       | 77/256 [01:55<03:44,  1.25s/batch]\n",
      " 30%|███       | 78/256 [01:56<03:46,  1.27s/batch]\n",
      " 31%|███       | 79/256 [01:57<03:40,  1.25s/batch]\n",
      " 31%|███▏      | 80/256 [01:59<03:38,  1.24s/batch]\n",
      " 32%|███▏      | 81/256 [02:00<03:35,  1.23s/batch]\n",
      " 32%|███▏      | 82/256 [02:01<03:37,  1.25s/batch]\n",
      " 32%|███▏      | 83/256 [02:02<03:34,  1.24s/batch]\n",
      " 33%|███▎      | 84/256 [02:04<03:34,  1.25s/batch]\n",
      " 33%|███▎      | 85/256 [02:05<03:31,  1.24s/batch]\n",
      " 34%|███▎      | 86/256 [02:06<03:31,  1.24s/batch]\n",
      " 34%|███▍      | 87/256 [02:07<03:29,  1.24s/batch]\n",
      " 34%|███▍      | 88/256 [02:08<03:28,  1.24s/batch]\n",
      " 35%|███▍      | 89/256 [02:10<03:29,  1.25s/batch]\n",
      " 35%|███▌      | 90/256 [02:11<03:35,  1.30s/batch]\n",
      " 36%|███▌      | 91/256 [02:13<03:37,  1.32s/batch]\n",
      " 36%|███▌      | 92/256 [02:14<03:41,  1.35s/batch]\n",
      " 36%|███▋      | 93/256 [02:15<03:37,  1.33s/batch]\n",
      " 37%|███▋      | 94/256 [02:17<03:37,  1.35s/batch]\n",
      " 37%|███▋      | 95/256 [02:18<03:33,  1.33s/batch]\n",
      " 38%|███▊      | 96/256 [02:20<03:46,  1.42s/batch]\n",
      " 38%|███▊      | 97/256 [02:21<03:49,  1.45s/batch]\n",
      " 38%|███▊      | 98/256 [02:23<03:57,  1.50s/batch]\n",
      " 39%|███▊      | 99/256 [02:24<03:58,  1.52s/batch]\n",
      " 39%|███▉      | 100/256 [02:26<03:54,  1.50s/batch]\n",
      " 39%|███▉      | 101/256 [02:27<03:41,  1.43s/batch]\n",
      " 40%|███▉      | 102/256 [02:28<03:38,  1.42s/batch]\n",
      " 40%|████      | 103/256 [02:29<03:25,  1.34s/batch]\n",
      " 41%|████      | 104/256 [02:31<03:13,  1.27s/batch]\n",
      " 41%|████      | 105/256 [02:32<03:07,  1.24s/batch]\n",
      " 41%|████▏     | 106/256 [02:33<03:07,  1.25s/batch]\n",
      " 42%|████▏     | 107/256 [02:34<03:09,  1.27s/batch]\n",
      " 42%|████▏     | 108/256 [02:36<03:23,  1.37s/batch]\n",
      " 43%|████▎     | 109/256 [02:37<03:17,  1.35s/batch]\n",
      " 43%|████▎     | 110/256 [02:39<03:18,  1.36s/batch]\n",
      " 43%|████▎     | 111/256 [02:40<03:16,  1.35s/batch]\n",
      " 44%|████▍     | 112/256 [02:41<03:07,  1.30s/batch]\n",
      " 44%|████▍     | 113/256 [02:43<03:09,  1.33s/batch]\n",
      " 45%|████▍     | 114/256 [02:44<03:16,  1.39s/batch]\n",
      " 45%|████▍     | 115/256 [02:46<03:20,  1.42s/batch]\n",
      " 45%|████▌     | 116/256 [02:47<03:14,  1.39s/batch]\n",
      " 46%|████▌     | 117/256 [02:48<03:21,  1.45s/batch]\n",
      " 46%|████▌     | 118/256 [02:50<03:17,  1.43s/batch]\n",
      " 46%|████▋     | 119/256 [02:51<03:12,  1.40s/batch]\n",
      " 47%|████▋     | 120/256 [02:52<03:04,  1.36s/batch]\n",
      " 47%|████▋     | 121/256 [02:54<02:57,  1.32s/batch]\n",
      " 48%|████▊     | 122/256 [02:55<02:54,  1.30s/batch]\n",
      " 48%|████▊     | 123/256 [02:56<02:49,  1.28s/batch]\n",
      " 48%|████▊     | 124/256 [02:57<02:49,  1.28s/batch]\n",
      " 49%|████▉     | 125/256 [02:59<02:44,  1.26s/batch]\n",
      " 49%|████▉     | 126/256 [03:00<02:44,  1.26s/batch]\n",
      " 50%|████▉     | 127/256 [03:01<02:41,  1.25s/batch]\n",
      " 50%|█████     | 128/256 [03:02<02:39,  1.25s/batch]\n",
      " 50%|█████     | 129/256 [03:04<02:42,  1.28s/batch]\n",
      " 51%|█████     | 130/256 [03:05<02:53,  1.37s/batch]\n",
      " 51%|█████     | 131/256 [03:07<02:49,  1.36s/batch]\n",
      " 52%|█████▏    | 132/256 [03:08<02:46,  1.35s/batch]\n",
      " 52%|█████▏    | 133/256 [03:09<02:40,  1.30s/batch]\n",
      " 52%|█████▏    | 134/256 [03:11<02:43,  1.34s/batch]\n",
      " 53%|█████▎    | 135/256 [03:12<02:43,  1.35s/batch]\n",
      " 53%|█████▎    | 136/256 [03:13<02:45,  1.38s/batch]\n",
      " 54%|█████▎    | 137/256 [03:15<02:45,  1.39s/batch]\n",
      " 54%|█████▍    | 138/256 [03:16<02:44,  1.39s/batch]\n",
      " 54%|█████▍    | 139/256 [03:18<02:38,  1.36s/batch]\n",
      " 55%|█████▍    | 140/256 [03:19<02:36,  1.35s/batch]\n",
      " 55%|█████▌    | 141/256 [03:20<02:33,  1.33s/batch]\n",
      " 55%|█████▌    | 142/256 [03:21<02:30,  1.32s/batch]\n",
      " 56%|█████▌    | 143/256 [03:23<02:28,  1.31s/batch]\n",
      " 56%|█████▋    | 144/256 [03:24<02:25,  1.30s/batch]\n",
      " 57%|█████▋    | 145/256 [03:25<02:24,  1.30s/batch]\n",
      " 57%|█████▋    | 146/256 [03:26<02:19,  1.27s/batch]\n",
      " 57%|█████▋    | 147/256 [03:28<02:17,  1.26s/batch]\n",
      " 58%|█████▊    | 148/256 [03:29<02:09,  1.20s/batch]\n",
      " 58%|█████▊    | 149/256 [03:30<02:03,  1.16s/batch]\n",
      " 59%|█████▊    | 150/256 [03:31<01:58,  1.12s/batch]\n",
      " 59%|█████▉    | 151/256 [03:32<01:54,  1.09s/batch]\n",
      " 59%|█████▉    | 152/256 [03:33<01:51,  1.07s/batch]\n",
      " 60%|█████▉    | 153/256 [03:34<01:49,  1.07s/batch]\n",
      " 60%|██████    | 154/256 [03:35<01:50,  1.08s/batch]\n",
      " 61%|██████    | 155/256 [03:36<01:50,  1.09s/batch]\n",
      " 61%|██████    | 156/256 [03:37<01:49,  1.09s/batch]\n",
      " 61%|██████▏   | 157/256 [03:39<01:51,  1.13s/batch]\n",
      " 62%|██████▏   | 158/256 [03:40<01:48,  1.10s/batch]\n",
      " 62%|██████▏   | 159/256 [03:41<01:45,  1.09s/batch]\n",
      " 62%|██████▎   | 160/256 [03:42<01:49,  1.14s/batch]\n",
      " 63%|██████▎   | 161/256 [03:43<01:49,  1.16s/batch]\n",
      " 63%|██████▎   | 162/256 [03:44<01:48,  1.16s/batch]\n",
      " 64%|██████▎   | 163/256 [03:45<01:45,  1.13s/batch]\n",
      " 64%|██████▍   | 164/256 [03:46<01:42,  1.11s/batch]\n",
      " 64%|██████▍   | 165/256 [03:48<01:41,  1.12s/batch]\n",
      " 65%|██████▍   | 166/256 [03:49<01:37,  1.09s/batch]\n",
      " 65%|██████▌   | 167/256 [03:50<01:35,  1.08s/batch]\n",
      " 66%|██████▌   | 168/256 [03:51<01:33,  1.07s/batch]\n",
      " 66%|██████▌   | 169/256 [03:52<01:32,  1.06s/batch]\n",
      " 66%|██████▋   | 170/256 [03:53<01:29,  1.04s/batch]\n",
      " 67%|██████▋   | 171/256 [03:54<01:28,  1.05s/batch]\n",
      " 67%|██████▋   | 172/256 [03:55<01:29,  1.06s/batch]\n",
      " 68%|██████▊   | 173/256 [03:56<01:32,  1.11s/batch]\n",
      " 68%|██████▊   | 174/256 [03:57<01:35,  1.17s/batch]\n",
      " 68%|██████▊   | 175/256 [03:58<01:31,  1.13s/batch]\n",
      " 69%|██████▉   | 176/256 [03:59<01:27,  1.09s/batch]\n",
      " 69%|██████▉   | 177/256 [04:00<01:25,  1.08s/batch]\n",
      " 70%|██████▉   | 178/256 [04:01<01:22,  1.05s/batch]\n",
      " 70%|██████▉   | 179/256 [04:02<01:20,  1.04s/batch]\n",
      " 70%|███████   | 180/256 [04:03<01:19,  1.04s/batch]\n",
      " 71%|███████   | 181/256 [04:05<01:18,  1.05s/batch]\n",
      " 71%|███████   | 182/256 [04:06<01:18,  1.06s/batch]\n",
      " 71%|███████▏  | 183/256 [04:07<01:17,  1.06s/batch]\n",
      " 72%|███████▏  | 184/256 [04:08<01:15,  1.05s/batch]\n",
      " 72%|███████▏  | 185/256 [04:09<01:13,  1.04s/batch]\n",
      " 73%|███████▎  | 186/256 [04:10<01:12,  1.04s/batch]\n",
      " 73%|███████▎  | 187/256 [04:11<01:11,  1.04s/batch]\n",
      " 73%|███████▎  | 188/256 [04:12<01:10,  1.03s/batch]\n",
      " 74%|███████▍  | 189/256 [04:13<01:09,  1.04s/batch]\n",
      " 74%|███████▍  | 190/256 [04:14<01:10,  1.07s/batch]\n",
      " 75%|███████▍  | 191/256 [04:15<01:12,  1.11s/batch]\n",
      " 75%|███████▌  | 192/256 [04:16<01:11,  1.12s/batch]\n",
      " 75%|███████▌  | 193/256 [04:17<01:09,  1.11s/batch]\n",
      " 76%|███████▌  | 194/256 [04:19<01:07,  1.09s/batch]\n",
      " 76%|███████▌  | 195/256 [04:20<01:07,  1.10s/batch]\n",
      " 77%|███████▋  | 196/256 [04:21<01:08,  1.15s/batch]\n",
      " 77%|███████▋  | 197/256 [04:22<01:08,  1.16s/batch]\n",
      " 77%|███████▋  | 198/256 [04:23<01:05,  1.13s/batch]\n",
      " 78%|███████▊  | 199/256 [04:24<01:04,  1.12s/batch]\n",
      " 78%|███████▊  | 200/256 [04:25<01:03,  1.13s/batch]\n",
      " 79%|███████▊  | 201/256 [04:26<01:00,  1.10s/batch]\n",
      " 79%|███████▉  | 202/256 [04:27<00:57,  1.07s/batch]\n",
      " 79%|███████▉  | 203/256 [04:28<00:55,  1.05s/batch]\n",
      " 80%|███████▉  | 204/256 [04:29<00:53,  1.04s/batch]\n",
      " 80%|████████  | 205/256 [04:30<00:52,  1.04s/batch]\n",
      " 80%|████████  | 206/256 [04:31<00:51,  1.03s/batch]\n",
      " 81%|████████  | 207/256 [04:33<00:50,  1.03s/batch]\n",
      " 81%|████████▏ | 208/256 [04:34<00:49,  1.02s/batch]\n",
      " 82%|████████▏ | 209/256 [04:35<00:48,  1.03s/batch]\n",
      " 82%|████████▏ | 210/256 [04:36<00:49,  1.07s/batch]\n",
      " 82%|████████▏ | 211/256 [04:37<00:47,  1.07s/batch]\n",
      " 83%|████████▎ | 212/256 [04:38<00:46,  1.05s/batch]\n",
      " 83%|████████▎ | 213/256 [04:39<00:44,  1.04s/batch]\n",
      " 84%|████████▎ | 214/256 [04:40<00:43,  1.03s/batch]\n",
      " 84%|████████▍ | 215/256 [04:41<00:42,  1.03s/batch]\n",
      " 84%|████████▍ | 216/256 [04:42<00:43,  1.09s/batch]\n",
      " 85%|████████▍ | 217/256 [04:43<00:45,  1.16s/batch]\n",
      " 85%|████████▌ | 218/256 [04:45<00:43,  1.15s/batch]\n",
      " 86%|████████▌ | 219/256 [04:46<00:41,  1.12s/batch]\n",
      " 86%|████████▌ | 220/256 [04:47<00:38,  1.08s/batch]\n",
      " 86%|████████▋ | 221/256 [04:48<00:37,  1.06s/batch]\n",
      " 87%|████████▋ | 222/256 [04:49<00:35,  1.04s/batch]\n",
      " 87%|████████▋ | 223/256 [04:50<00:34,  1.03s/batch]\n",
      " 88%|████████▊ | 224/256 [04:51<00:32,  1.03s/batch]\n",
      " 88%|████████▊ | 225/256 [04:52<00:31,  1.03s/batch]\n",
      " 88%|████████▊ | 226/256 [04:53<00:31,  1.04s/batch]\n",
      " 89%|████████▊ | 227/256 [04:54<00:30,  1.05s/batch]\n",
      " 89%|████████▉ | 228/256 [04:55<00:29,  1.05s/batch]\n",
      " 89%|████████▉ | 229/256 [04:56<00:28,  1.05s/batch]\n",
      " 90%|████████▉ | 230/256 [04:57<00:27,  1.06s/batch]\n",
      " 90%|█████████ | 231/256 [04:58<00:26,  1.07s/batch]\n",
      " 91%|█████████ | 232/256 [04:59<00:25,  1.06s/batch]\n",
      " 91%|█████████ | 233/256 [05:00<00:24,  1.07s/batch]\n",
      " 91%|█████████▏| 234/256 [05:01<00:24,  1.10s/batch]\n",
      " 92%|█████████▏| 235/256 [05:03<00:24,  1.16s/batch]\n",
      " 92%|█████████▏| 236/256 [05:04<00:22,  1.13s/batch]\n",
      " 93%|█████████▎| 237/256 [05:05<00:21,  1.11s/batch]\n",
      " 93%|█████████▎| 238/256 [05:06<00:20,  1.14s/batch]\n",
      " 93%|█████████▎| 239/256 [05:07<00:19,  1.13s/batch]\n",
      " 94%|█████████▍| 240/256 [05:08<00:17,  1.11s/batch]\n",
      " 94%|█████████▍| 241/256 [05:09<00:16,  1.10s/batch]\n",
      " 95%|█████████▍| 242/256 [05:10<00:16,  1.15s/batch]\n",
      " 95%|█████████▍| 243/256 [05:12<00:15,  1.18s/batch]\n",
      " 95%|█████████▌| 244/256 [05:13<00:13,  1.15s/batch]\n",
      " 96%|█████████▌| 245/256 [05:14<00:12,  1.14s/batch]\n",
      " 96%|█████████▌| 246/256 [05:15<00:11,  1.13s/batch]\n",
      " 96%|█████████▋| 247/256 [05:16<00:09,  1.11s/batch]\n",
      " 97%|█████████▋| 248/256 [05:17<00:08,  1.09s/batch]\n",
      " 97%|█████████▋| 249/256 [05:18<00:07,  1.08s/batch]\n",
      " 98%|█████████▊| 250/256 [05:19<00:06,  1.08s/batch]\n",
      " 98%|█████████▊| 251/256 [05:21<00:05,  1.15s/batch]\n",
      " 98%|█████████▊| 252/256 [05:22<00:04,  1.14s/batch]\n",
      " 99%|█████████▉| 253/256 [05:23<00:03,  1.12s/batch]\n",
      " 99%|█████████▉| 254/256 [05:24<00:02,  1.11s/batch]\n",
      "100%|█████████▉| 255/256 [05:25<00:01,  1.10s/batch]\n",
      "100%|██████████| 256/256 [05:26<00:00,  1.16s/batch]\n",
      "100%|██████████| 256/256 [05:26<00:00,  1.28s/batch]\n",
      "\n",
      "  0%|          | 0/256 [00:00<?, ?batch/s]\n",
      "  0%|          | 1/256 [00:01<05:19,  1.25s/batch]\n",
      "  1%|          | 2/256 [00:02<04:51,  1.15s/batch]\n",
      "  1%|          | 3/256 [00:03<04:43,  1.12s/batch]\n",
      "  2%|▏         | 4/256 [00:04<04:44,  1.13s/batch]\n",
      "  2%|▏         | 5/256 [00:05<04:43,  1.13s/batch]\n",
      "  2%|▏         | 6/256 [00:06<04:33,  1.09s/batch]\n",
      "  3%|▎         | 7/256 [00:07<04:29,  1.08s/batch]\n",
      "  3%|▎         | 8/256 [00:08<04:26,  1.07s/batch]\n",
      "  4%|▎         | 9/256 [00:09<04:26,  1.08s/batch]\n",
      "  4%|▍         | 10/256 [00:10<04:24,  1.07s/batch]\n",
      "  4%|▍         | 11/256 [00:12<04:22,  1.07s/batch]\n",
      "  5%|▍         | 12/256 [00:13<04:22,  1.07s/batch]\n",
      "  5%|▌         | 13/256 [00:14<04:24,  1.09s/batch]\n",
      "  5%|▌         | 14/256 [00:15<04:21,  1.08s/batch]\n",
      "  6%|▌         | 15/256 [00:16<04:27,  1.11s/batch]\n",
      "  6%|▋         | 16/256 [00:17<04:25,  1.11s/batch]\n",
      "  7%|▋         | 17/256 [00:18<04:22,  1.10s/batch]\n",
      "  7%|▋         | 18/256 [00:19<04:16,  1.08s/batch]\n",
      "  7%|▋         | 19/256 [00:20<04:20,  1.10s/batch]\n",
      "  8%|▊         | 20/256 [00:21<04:21,  1.11s/batch]\n",
      "  8%|▊         | 21/256 [00:23<04:25,  1.13s/batch]\n",
      "  9%|▊         | 22/256 [00:24<04:23,  1.12s/batch]\n",
      "  9%|▉         | 23/256 [00:25<04:18,  1.11s/batch]\n",
      "  9%|▉         | 24/256 [00:26<04:11,  1.08s/batch]\n",
      " 10%|▉         | 25/256 [00:27<04:05,  1.06s/batch]\n",
      " 10%|█         | 26/256 [00:28<04:02,  1.05s/batch]\n",
      " 11%|█         | 27/256 [00:29<04:04,  1.07s/batch]\n",
      " 11%|█         | 28/256 [00:30<04:07,  1.09s/batch]\n",
      " 11%|█▏        | 29/256 [00:31<04:07,  1.09s/batch]\n",
      " 12%|█▏        | 30/256 [00:33<04:31,  1.20s/batch]\n",
      " 12%|█▏        | 31/256 [00:34<05:05,  1.36s/batch]\n",
      " 12%|█▎        | 32/256 [00:36<05:21,  1.44s/batch]\n",
      " 13%|█▎        | 33/256 [00:37<04:59,  1.34s/batch]\n",
      " 13%|█▎        | 34/256 [00:38<04:47,  1.30s/batch]\n",
      " 14%|█▎        | 35/256 [00:40<04:52,  1.32s/batch]\n",
      " 14%|█▍        | 36/256 [00:41<04:33,  1.25s/batch]\n",
      " 14%|█▍        | 37/256 [00:42<04:20,  1.19s/batch]\n",
      " 15%|█▍        | 38/256 [00:43<04:11,  1.15s/batch]\n",
      " 15%|█▌        | 39/256 [00:44<04:10,  1.15s/batch]\n",
      " 16%|█▌        | 40/256 [00:45<04:09,  1.15s/batch]\n",
      " 16%|█▌        | 41/256 [00:46<04:04,  1.14s/batch]\n",
      " 16%|█▋        | 42/256 [00:48<04:20,  1.22s/batch]\n",
      " 17%|█▋        | 43/256 [00:49<04:41,  1.32s/batch]\n",
      " 17%|█▋        | 44/256 [00:50<04:28,  1.27s/batch]\n",
      " 18%|█▊        | 45/256 [00:52<04:31,  1.29s/batch]\n",
      " 18%|█▊        | 46/256 [00:53<04:17,  1.23s/batch]\n",
      " 18%|█▊        | 47/256 [00:54<04:10,  1.20s/batch]\n",
      " 19%|█▉        | 48/256 [00:55<04:11,  1.21s/batch]\n",
      " 19%|█▉        | 49/256 [00:56<04:03,  1.18s/batch]\n",
      " 20%|█▉        | 50/256 [00:58<04:26,  1.30s/batch]\n",
      " 20%|█▉        | 51/256 [00:59<04:39,  1.37s/batch]\n",
      " 20%|██        | 52/256 [01:01<05:01,  1.48s/batch]\n",
      " 21%|██        | 53/256 [01:03<05:11,  1.53s/batch]\n",
      " 21%|██        | 54/256 [01:04<05:00,  1.49s/batch]\n",
      " 21%|██▏       | 55/256 [01:05<04:41,  1.40s/batch]\n",
      " 22%|██▏       | 56/256 [01:07<04:29,  1.35s/batch]\n",
      " 22%|██▏       | 57/256 [01:08<04:19,  1.30s/batch]\n",
      " 23%|██▎       | 58/256 [01:09<04:17,  1.30s/batch]\n",
      " 23%|██▎       | 59/256 [01:10<04:11,  1.28s/batch]\n",
      " 23%|██▎       | 60/256 [01:12<04:14,  1.30s/batch]\n",
      " 24%|██▍       | 61/256 [01:13<04:15,  1.31s/batch]\n",
      " 24%|██▍       | 62/256 [01:14<04:15,  1.32s/batch]\n",
      " 25%|██▍       | 63/256 [01:16<04:39,  1.45s/batch]\n",
      " 25%|██▌       | 64/256 [01:18<05:07,  1.60s/batch]\n",
      " 25%|██▌       | 65/256 [01:20<05:11,  1.63s/batch]\n",
      " 26%|██▌       | 66/256 [01:21<04:49,  1.53s/batch]\n",
      " 26%|██▌       | 67/256 [01:22<04:36,  1.46s/batch]\n",
      " 27%|██▋       | 68/256 [01:24<04:53,  1.56s/batch]\n",
      " 27%|██▋       | 69/256 [01:26<04:40,  1.50s/batch]\n",
      " 27%|██▋       | 70/256 [01:27<04:27,  1.44s/batch]\n",
      " 28%|██▊       | 71/256 [01:28<04:27,  1.44s/batch]\n",
      " 28%|██▊       | 72/256 [01:30<04:23,  1.43s/batch]\n",
      " 29%|██▊       | 73/256 [01:31<04:31,  1.48s/batch]\n",
      " 29%|██▉       | 74/256 [01:33<04:33,  1.50s/batch]\n",
      " 29%|██▉       | 75/256 [01:34<04:18,  1.43s/batch]\n",
      " 30%|██▉       | 76/256 [01:36<04:23,  1.46s/batch]\n",
      " 30%|███       | 77/256 [01:37<04:08,  1.39s/batch]\n",
      " 30%|███       | 78/256 [01:38<03:58,  1.34s/batch]\n",
      " 31%|███       | 79/256 [01:39<03:54,  1.32s/batch]\n",
      " 31%|███▏      | 80/256 [01:41<04:03,  1.38s/batch]\n",
      " 32%|███▏      | 81/256 [01:42<03:50,  1.32s/batch]\n",
      " 32%|███▏      | 82/256 [01:43<03:38,  1.26s/batch]\n",
      " 32%|███▏      | 83/256 [01:44<03:35,  1.25s/batch]\n",
      " 33%|███▎      | 84/256 [01:46<03:45,  1.31s/batch]\n",
      " 33%|███▎      | 85/256 [01:47<03:41,  1.29s/batch]\n",
      " 34%|███▎      | 86/256 [01:48<03:31,  1.25s/batch]\n",
      " 34%|███▍      | 87/256 [01:49<03:23,  1.20s/batch]\n",
      " 34%|███▍      | 88/256 [01:50<03:19,  1.19s/batch]\n",
      " 35%|███▍      | 89/256 [01:52<03:11,  1.15s/batch]\n",
      " 35%|███▌      | 90/256 [01:53<03:09,  1.14s/batch]\n",
      " 36%|███▌      | 91/256 [01:54<03:10,  1.15s/batch]\n",
      " 36%|███▌      | 92/256 [01:55<03:15,  1.19s/batch]\n",
      " 36%|███▋      | 93/256 [01:57<03:22,  1.24s/batch]\n",
      " 37%|███▋      | 94/256 [01:58<03:41,  1.37s/batch]\n",
      " 37%|███▋      | 95/256 [01:59<03:31,  1.31s/batch]\n",
      " 38%|███▊      | 96/256 [02:00<03:20,  1.25s/batch]\n",
      " 38%|███▊      | 97/256 [02:02<03:11,  1.21s/batch]\n",
      " 38%|███▊      | 98/256 [02:03<03:02,  1.15s/batch]\n",
      " 39%|███▊      | 99/256 [02:04<02:56,  1.13s/batch]\n",
      " 39%|███▉      | 100/256 [02:05<02:51,  1.10s/batch]\n",
      " 39%|███▉      | 101/256 [02:06<02:47,  1.08s/batch]\n",
      " 40%|███▉      | 102/256 [02:07<02:45,  1.08s/batch]\n",
      " 40%|████      | 103/256 [02:08<02:43,  1.07s/batch]\n",
      " 41%|████      | 104/256 [02:09<02:41,  1.07s/batch]\n",
      " 41%|████      | 105/256 [02:10<02:42,  1.08s/batch]\n",
      " 41%|████▏     | 106/256 [02:11<02:40,  1.07s/batch]\n",
      " 42%|████▏     | 107/256 [02:12<02:38,  1.06s/batch]\n",
      " 42%|████▏     | 108/256 [02:13<02:42,  1.10s/batch]\n",
      " 43%|████▎     | 109/256 [02:15<02:46,  1.13s/batch]\n",
      " 43%|████▎     | 110/256 [02:16<02:46,  1.14s/batch]\n",
      " 43%|████▎     | 111/256 [02:17<02:43,  1.13s/batch]\n",
      " 44%|████▍     | 112/256 [02:18<02:43,  1.13s/batch]\n",
      " 44%|████▍     | 113/256 [02:19<02:48,  1.18s/batch]\n",
      " 45%|████▍     | 114/256 [02:20<02:48,  1.19s/batch]\n",
      " 45%|████▍     | 115/256 [02:22<02:51,  1.22s/batch]\n",
      " 45%|████▌     | 116/256 [02:23<02:57,  1.27s/batch]\n",
      " 46%|████▌     | 117/256 [02:25<03:14,  1.40s/batch]\n",
      " 46%|████▌     | 118/256 [02:27<03:41,  1.60s/batch]\n",
      " 46%|████▋     | 119/256 [02:29<03:54,  1.71s/batch]\n",
      " 47%|████▋     | 120/256 [02:31<03:51,  1.70s/batch]\n",
      " 47%|████▋     | 121/256 [02:32<03:52,  1.72s/batch]\n",
      " 48%|████▊     | 122/256 [02:34<03:45,  1.68s/batch]\n",
      " 48%|████▊     | 123/256 [02:37<04:32,  2.05s/batch]\n",
      " 48%|████▊     | 124/256 [02:38<04:09,  1.89s/batch]\n",
      " 49%|████▉     | 125/256 [02:40<03:48,  1.75s/batch]\n",
      " 49%|████▉     | 126/256 [02:41<03:36,  1.66s/batch]\n",
      " 50%|████▉     | 127/256 [02:43<03:36,  1.68s/batch]\n",
      " 50%|█████     | 128/256 [02:44<03:31,  1.65s/batch]\n",
      " 50%|█████     | 129/256 [02:46<03:38,  1.72s/batch]\n",
      " 51%|█████     | 130/256 [02:48<03:34,  1.70s/batch]\n",
      " 51%|█████     | 131/256 [02:50<03:30,  1.69s/batch]\n",
      " 52%|█████▏    | 132/256 [02:52<03:46,  1.82s/batch]\n",
      " 52%|█████▏    | 133/256 [02:54<03:52,  1.89s/batch]\n",
      " 52%|█████▏    | 134/256 [02:56<03:46,  1.85s/batch]\n",
      " 53%|█████▎    | 135/256 [02:57<03:34,  1.78s/batch]\n",
      " 53%|█████▎    | 136/256 [02:59<03:30,  1.75s/batch]\n",
      " 54%|█████▎    | 137/256 [03:00<03:21,  1.69s/batch]\n",
      " 54%|█████▍    | 138/256 [03:02<03:11,  1.62s/batch]\n",
      " 54%|█████▍    | 139/256 [03:03<03:02,  1.56s/batch]\n",
      " 55%|█████▍    | 140/256 [03:05<02:57,  1.53s/batch]\n",
      " 55%|█████▌    | 141/256 [03:06<02:51,  1.49s/batch]\n",
      " 55%|█████▌    | 142/256 [03:08<02:51,  1.50s/batch]\n",
      " 56%|█████▌    | 143/256 [03:09<02:44,  1.46s/batch]\n",
      " 56%|█████▋    | 144/256 [03:11<02:51,  1.53s/batch]\n",
      " 57%|█████▋    | 145/256 [03:12<02:45,  1.49s/batch]\n",
      " 57%|█████▋    | 146/256 [03:14<02:42,  1.48s/batch]\n",
      " 57%|█████▋    | 147/256 [03:17<03:39,  2.01s/batch]\n",
      " 58%|█████▊    | 148/256 [03:18<03:22,  1.87s/batch]\n",
      " 58%|█████▊    | 149/256 [03:20<03:22,  1.90s/batch]\n",
      " 59%|█████▊    | 150/256 [03:22<03:17,  1.87s/batch]\n",
      " 59%|█████▉    | 151/256 [03:24<03:14,  1.85s/batch]\n",
      " 59%|█████▉    | 152/256 [03:26<03:10,  1.83s/batch]\n",
      " 60%|█████▉    | 153/256 [03:27<03:02,  1.77s/batch]\n",
      " 60%|██████    | 154/256 [03:29<03:08,  1.85s/batch]\n",
      " 61%|██████    | 155/256 [03:32<03:16,  1.94s/batch]\n",
      " 61%|██████    | 156/256 [03:33<03:04,  1.84s/batch]\n",
      " 61%|██████▏   | 157/256 [03:35<03:00,  1.82s/batch]\n",
      " 62%|██████▏   | 158/256 [03:37<02:52,  1.76s/batch]\n",
      " 62%|██████▏   | 159/256 [03:38<02:39,  1.65s/batch]\n",
      " 62%|██████▎   | 160/256 [03:40<02:41,  1.69s/batch]\n",
      " 63%|██████▎   | 161/256 [03:41<02:35,  1.64s/batch]\n",
      " 63%|██████▎   | 162/256 [03:43<02:30,  1.60s/batch]\n",
      " 64%|██████▎   | 163/256 [03:44<02:23,  1.54s/batch]\n",
      " 64%|██████▍   | 164/256 [03:46<02:27,  1.60s/batch]\n",
      " 64%|██████▍   | 165/256 [03:48<02:31,  1.66s/batch]\n",
      " 65%|██████▍   | 166/256 [03:50<02:35,  1.73s/batch]\n",
      " 65%|██████▌   | 167/256 [03:51<02:29,  1.68s/batch]\n",
      " 66%|██████▌   | 168/256 [03:53<02:23,  1.64s/batch]\n",
      " 66%|██████▌   | 169/256 [03:54<02:20,  1.62s/batch]\n",
      " 66%|██████▋   | 170/256 [03:56<02:17,  1.60s/batch]\n",
      " 67%|██████▋   | 171/256 [03:57<02:15,  1.59s/batch]\n",
      " 67%|██████▋   | 172/256 [03:59<02:19,  1.66s/batch]\n",
      " 68%|██████▊   | 173/256 [04:01<02:16,  1.64s/batch]\n",
      " 68%|██████▊   | 174/256 [04:02<02:10,  1.59s/batch]\n",
      " 68%|██████▊   | 175/256 [04:04<02:03,  1.52s/batch]\n",
      " 69%|██████▉   | 176/256 [04:06<02:11,  1.64s/batch]\n",
      " 69%|██████▉   | 177/256 [04:07<02:07,  1.62s/batch]\n",
      " 70%|██████▉   | 178/256 [04:09<02:07,  1.63s/batch]\n",
      " 70%|██████▉   | 179/256 [04:10<02:02,  1.60s/batch]\n",
      " 70%|███████   | 180/256 [04:12<02:07,  1.67s/batch]\n",
      " 71%|███████   | 181/256 [04:14<02:00,  1.60s/batch]\n",
      " 71%|███████   | 182/256 [04:16<02:07,  1.73s/batch]\n",
      " 71%|███████▏  | 183/256 [04:17<02:05,  1.72s/batch]\n",
      " 72%|███████▏  | 184/256 [04:19<01:56,  1.61s/batch]\n",
      " 72%|███████▏  | 185/256 [04:20<01:49,  1.55s/batch]\n",
      " 73%|███████▎  | 186/256 [04:22<01:53,  1.62s/batch]\n",
      " 73%|███████▎  | 187/256 [04:23<01:45,  1.53s/batch]\n",
      " 73%|███████▎  | 188/256 [04:25<01:44,  1.54s/batch]\n",
      " 74%|███████▍  | 189/256 [04:26<01:35,  1.43s/batch]\n",
      " 74%|███████▍  | 190/256 [04:27<01:31,  1.39s/batch]\n",
      " 75%|███████▍  | 191/256 [04:29<01:31,  1.40s/batch]\n",
      " 75%|███████▌  | 192/256 [04:30<01:27,  1.37s/batch]\n",
      " 75%|███████▌  | 193/256 [04:31<01:25,  1.35s/batch]\n",
      " 76%|███████▌  | 194/256 [04:33<01:23,  1.35s/batch]\n",
      " 76%|███████▌  | 195/256 [04:34<01:20,  1.32s/batch]\n",
      " 77%|███████▋  | 196/256 [04:35<01:18,  1.31s/batch]\n",
      " 77%|███████▋  | 197/256 [04:36<01:16,  1.30s/batch]\n",
      " 77%|███████▋  | 198/256 [04:38<01:15,  1.30s/batch]\n",
      " 78%|███████▊  | 199/256 [04:39<01:13,  1.30s/batch]\n",
      " 78%|███████▊  | 200/256 [04:40<01:15,  1.34s/batch]\n",
      " 79%|███████▊  | 201/256 [04:42<01:12,  1.32s/batch]\n",
      " 79%|███████▉  | 202/256 [04:43<01:10,  1.30s/batch]\n",
      " 79%|███████▉  | 203/256 [04:44<01:08,  1.30s/batch]\n",
      " 80%|███████▉  | 204/256 [04:46<01:07,  1.30s/batch]\n",
      " 80%|████████  | 205/256 [04:47<01:10,  1.38s/batch]\n",
      " 80%|████████  | 206/256 [04:49<01:13,  1.48s/batch]\n",
      " 81%|████████  | 207/256 [04:51<01:17,  1.59s/batch]\n",
      " 81%|████████▏ | 208/256 [04:52<01:15,  1.57s/batch]\n",
      " 82%|████████▏ | 209/256 [04:54<01:15,  1.60s/batch]\n",
      " 82%|████████▏ | 210/256 [04:56<01:15,  1.64s/batch]\n",
      " 82%|████████▏ | 211/256 [04:57<01:10,  1.57s/batch]\n",
      " 83%|████████▎ | 212/256 [04:58<01:05,  1.48s/batch]\n",
      " 83%|████████▎ | 213/256 [05:00<01:03,  1.47s/batch]\n",
      " 84%|████████▎ | 214/256 [05:01<00:59,  1.42s/batch]\n",
      " 84%|████████▍ | 215/256 [05:03<00:58,  1.43s/batch]\n",
      " 84%|████████▍ | 216/256 [05:04<01:00,  1.50s/batch]\n",
      " 85%|████████▍ | 217/256 [05:06<00:57,  1.48s/batch]\n",
      " 85%|████████▌ | 218/256 [05:07<00:57,  1.51s/batch]\n",
      " 86%|████████▌ | 219/256 [05:09<00:56,  1.53s/batch]\n",
      " 86%|████████▌ | 220/256 [05:10<00:53,  1.49s/batch]\n",
      " 86%|████████▋ | 221/256 [05:12<00:51,  1.47s/batch]\n",
      " 87%|████████▋ | 222/256 [05:13<00:52,  1.55s/batch]\n",
      " 87%|████████▋ | 223/256 [05:15<00:49,  1.49s/batch]\n",
      " 88%|████████▊ | 224/256 [05:16<00:47,  1.50s/batch]\n",
      " 88%|████████▊ | 225/256 [05:18<00:44,  1.43s/batch]\n",
      " 88%|████████▊ | 226/256 [05:19<00:41,  1.39s/batch]\n",
      " 89%|████████▊ | 227/256 [05:20<00:38,  1.34s/batch]\n",
      " 89%|████████▉ | 228/256 [05:21<00:37,  1.33s/batch]\n",
      " 89%|████████▉ | 229/256 [05:23<00:34,  1.29s/batch]\n",
      " 90%|████████▉ | 230/256 [05:24<00:33,  1.29s/batch]\n",
      " 90%|█████████ | 231/256 [05:25<00:32,  1.28s/batch]\n",
      " 91%|█████████ | 232/256 [05:26<00:31,  1.31s/batch]\n",
      " 91%|█████████ | 233/256 [05:28<00:30,  1.33s/batch]\n",
      " 91%|█████████▏| 234/256 [05:29<00:28,  1.31s/batch]\n",
      " 92%|█████████▏| 235/256 [05:30<00:27,  1.30s/batch]\n",
      " 92%|█████████▏| 236/256 [05:31<00:24,  1.23s/batch]\n",
      " 93%|█████████▎| 237/256 [05:32<00:22,  1.17s/batch]\n",
      " 93%|█████████▎| 238/256 [05:34<00:20,  1.15s/batch]\n",
      " 93%|█████████▎| 239/256 [05:35<00:19,  1.14s/batch]\n",
      " 94%|█████████▍| 240/256 [05:36<00:17,  1.12s/batch]\n",
      " 94%|█████████▍| 241/256 [05:37<00:16,  1.10s/batch]\n",
      " 95%|█████████▍| 242/256 [05:38<00:15,  1.09s/batch]\n",
      " 95%|█████████▍| 243/256 [05:39<00:14,  1.08s/batch]\n",
      " 95%|█████████▌| 244/256 [05:40<00:12,  1.06s/batch]\n",
      " 96%|█████████▌| 245/256 [05:41<00:11,  1.05s/batch]\n",
      " 96%|█████████▌| 246/256 [05:42<00:10,  1.04s/batch]\n",
      " 96%|█████████▋| 247/256 [05:43<00:10,  1.13s/batch]\n",
      " 97%|█████████▋| 248/256 [05:45<00:10,  1.32s/batch]\n",
      " 97%|█████████▋| 249/256 [05:47<00:10,  1.46s/batch]\n",
      " 98%|█████████▊| 250/256 [05:49<00:09,  1.55s/batch]\n",
      " 98%|█████████▊| 251/256 [05:52<00:10,  2.03s/batch]\n",
      " 98%|█████████▊| 252/256 [05:54<00:08,  2.05s/batch]\n",
      " 99%|█████████▉| 253/256 [05:57<00:06,  2.24s/batch]\n",
      " 99%|█████████▉| 254/256 [06:00<00:04,  2.48s/batch]\n",
      "100%|█████████▉| 255/256 [06:01<00:02,  2.27s/batch]\n",
      "100%|██████████| 256/256 [06:03<00:00,  2.18s/batch]\n",
      "100%|██████████| 256/256 [06:03<00:00,  1.42s/batch]\n",
      "\n",
      "Bias-field correction completed successfully.\n",
      "stdout: \n",
      "N4 Bias Correction Parameters:\n",
      "\n",
      "- verbosity: -1\n",
      "- input volume: /Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer/OAS_0001-2/mri/orig.mgz\n",
      "- output volume: do not save\n",
      "- rescaled volume: /Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer/OAS_0001-2/mri/orig_nu.mgz\n",
      "- mask: default (>0)\n",
      "- aseg: /Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer/OAS_0001-2/mri/aparc.DKTatlas+aseg.deep.mgz\n",
      "- shrink factor: 4\n",
      "- number fitting levels: 4\n",
      "- number iterations: 50\n",
      "- convergence threshold: 0.0\n",
      "- threads: 4\n",
      "read MGZ (FreeSurfer) image via nibabel...\n",
      "executing N4 correction ...\n",
      "- mask: ones (default)\n",
      "normalize WM to 105.0 (find WM from aseg)\n",
      "read MGZ (FreeSurfer) image via nibabel...\n",
      "- source white matter intensity: 112.36\n",
      "- m: 0.9257\n",
      "converting rescaled to UCHAR\n",
      "writing PosixPath: /Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer/OAS_0001-2/mri/orig_nu.mgz\n",
      "write MGZ (FreeSurfer) image via nibabel...\n",
      "\n",
      "stderr: \n",
      "FastSurfer pipeline completed successfully.\n",
      "The following expected output files are missing:\n",
      "/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer/OAS_0001-2/scripts/deep-seg.log\n",
      "/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer/OAS_0001-2/scripts/build.log\n",
      "/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer/OAS_0001-2/stats/aseg+DKT.stats\n",
      "/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer/OAS_0001-2/mri/cerebellum.CerebNet.nii.gz\n",
      "/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer/OAS_0001-2/stats/cerebellum.CerebNet.stats\n",
      "/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer/OAS_0001-2/mri/fastsurfer.merged.mgz\n"
     ]
    }
   ],
   "source": [
    "import os\n",
    "import subprocess\n",
    "import sys\n",
    "\n",
    "# # Set FreeSurfer environment variables\n",
    "# os.environ[\"FREESURFER_HOME\"] = \"/Applications/freesurfer/7.4.1\"\n",
    "# freesurfer_setup_script = os.path.join(os.environ[\"FREESURFER_HOME\"], \"SetUpFreeSurfer.sh\")\n",
    "\n",
    "# # Source the FreeSurfer setup script\n",
    "# subprocess.run(f\"source {freesurfer_setup_script}\", shell=True, executable='/usr/local/bin/bash')\n",
    "\n",
    "merged_segfile = os.path.join(sd, \"mri\", \"fastsurfer.merged.mgz\")\n",
    "asegdkt_segfile = os.path.join(sd, \"mri\", \"aparc.DKTatlas+aseg.deep.mgz\")\n",
    "asegdkt_statsfile = os.path.join(sd, \"stats\", \"aseg+DKT.stats\")\n",
    "cereb_segfile = os.path.join(sd, \"mri\", \"cerebellum.CerebNet.nii.gz\")\n",
    "cereb_statsfile = os.path.join(sd, \"stats\", \"cerebellum.CerebNet.stats\")\n",
    "mask_name = os.path.join(sd, \"mri\", \"mask.mgz\")\n",
    "conformed_name = os.path.join(sd, \"mri\", \"orig.mgz\")\n",
    "norm_name = os.path.join(sd, \"mri\", \"orig_nu.mgz\")\n",
    "seg_log = os.path.join(sd, \"scripts\", \"deep-seg.log\")\n",
    "build_log = os.path.join(sd, \"scripts\", \"build.log\")\n",
    "aseg_segfile = os.path.join(sd, \"mri\", \"aseg.auto_noCCseg.mgz\")\n",
    "\n",
    "device = \"auto\"\n",
    "viewagg = \"auto\"\n",
    "batch_size = 1\n",
    "threads = 4\n",
    "python = \"python3.10\"\n",
    "python_args = [\"-s\"]\n",
    "allow_root = []\n",
    "\n",
    "# Print to verify\n",
    "print(\"FREESURFER_HOME:\", os.environ.get(\"FREESURFER_HOME\"))\n",
    "print(\"PYTHONPATH:\", os.environ.get(\"PYTHONPATH\"))\n",
    "print(\"sys.path:\", sys.path)\n",
    "print(\"Data Directory:\", datadir)\n",
    "print(\"FastSurfer Directory:\", fastsurferdir)\n",
    "print(\"FastSurferCNN Directory:\", fastsurfercnndir)\n",
    "print(\"ReconSurf Directory:\", reconsurfdir)\n",
    "\n",
    "def run_segmentation(t1, subject, threads, sd, allow_root=False):\n",
    "    \"\"\"\n",
    "    Run the FastSurfer segmentation process.\n",
    "\n",
    "    Args:\n",
    "        t1 (str): Path to the T1-weighted MRI image file.\n",
    "        subject (str): Subject ID.\n",
    "        threads (int): Number of threads to use for parallel processing.\n",
    "        sd (str): Output subject directory.\n",
    "        allow_root (bool, optional): Allow running the script as root. Defaults to False.\n",
    "\n",
    "    Raises:\n",
    "        subprocess.CalledProcessError: If the segmentation process fails.\n",
    "    \"\"\"\n",
    "    cmd = [\n",
    "        python, *python_args,\n",
    "        os.path.join(fastsurfercnndir, \"run_prediction.py\"),\n",
    "        \"--t1\", t1,\n",
    "        \"--sid\", subject,\n",
    "        \"--threads\", str(threads),\n",
    "        \"--sd\", fastsurferdir\n",
    "    ]\n",
    "\n",
    "    if allow_root:\n",
    "        cmd.append(\"--allow_root\")\n",
    "\n",
    "    try:\n",
    "        result = subprocess.run(cmd, check=True, capture_output=True, text=True)\n",
    "        print(\"Segmentation completed successfully.\")\n",
    "        print(\"stdout:\", result.stdout)\n",
    "        print(\"stderr:\", result.stderr)\n",
    "    except subprocess.CalledProcessError as e:\n",
    "        print(\"Segmentation failed:\")\n",
    "        print(f\"Command: {e.cmd}\")\n",
    "        print(f\"Return code: {e.returncode}\")\n",
    "        print(f\"Output: {e.stdout}\")\n",
    "        print(f\"Error: {e.stderr}\")\n",
    "        raise\n",
    "\n",
    "def run_biasfield_correction(conformed_name, norm_name, asegdkt_segfile, threads):\n",
    "    \"\"\"\n",
    "    Run the N4 bias field correction on the input MRI image.\n",
    "\n",
    "    Args:\n",
    "        conformed_name (str): Path to the conformed MRI image.\n",
    "        norm_name (str): Path to the normalized MRI image.\n",
    "        asegdkt_segfile (str): Path to the DKT atlas segmentation file.\n",
    "        threads (int): Number of threads to use for parallel processing.\n",
    "\n",
    "    Raises:\n",
    "        subprocess.CalledProcessError: If the bias field correction fails.\n",
    "    \"\"\"\n",
    "    cmd = [\n",
    "        python,\n",
    "        os.path.join(reconsurfdir, \"N4_bias_correct.py\"),\n",
    "        \"--in\", conformed_name,\n",
    "        \"--rescale\", norm_name,\n",
    "        \"--aseg\", asegdkt_segfile,\n",
    "        \"--threads\", str(threads)\n",
    "    ]\n",
    "\n",
    "    try:\n",
    "        result = subprocess.run(cmd, check=True, capture_output=True, text=True)\n",
    "        print(\"Bias-field correction completed successfully.\")\n",
    "        print(\"stdout:\", result.stdout)\n",
    "        print(\"stderr:\", result.stderr)\n",
    "    except subprocess.CalledProcessError as e:\n",
    "        print(\"Bias-field correction failed:\")\n",
    "        print(f\"Command: {e.cmd}\")\n",
    "        print(f\"Return code: {e.returncode}\")\n",
    "        print(f\"Output: {e.stdout}\")\n",
    "        print(f\"Error: {e.stderr}\")\n",
    "        raise\n",
    "\n",
    "def run_recon_surf(subject, sd, threads):\n",
    "    \"\"\"\n",
    "    Run the recon-surf process.\n",
    "\n",
    "    Args:\n",
    "        subject (str): Subject ID.\n",
    "        sd (str): Output subject directory.\n",
    "        threads (int): Number of threads to use for parallel processing.\n",
    "\n",
    "    Raises:\n",
    "        subprocess.CalledProcessError: If the recon-surf process fails.\n",
    "    \"\"\"\n",
    "    cmd = [\n",
    "        python,\n",
    "        os.path.join(reconsurfdir, \"recon_surf.py\"),\n",
    "        \"--sid\", subject,\n",
    "        \"--sd\", sd,\n",
    "        \"--threads\", str(threads)\n",
    "    ]\n",
    "\n",
    "    try:\n",
    "        result = subprocess.run(cmd, check=True, capture_output=True, text=True)\n",
    "        print(\"Recon-surf process completed successfully.\")\n",
    "        print(\"stdout:\", result.stdout)\n",
    "        print(\"stderr:\", result.stderr)\n",
    "    except subprocess.CalledProcessError as e:\n",
    "        print(\"Recon-surf process failed:\")\n",
    "        print(f\"Command: {e.cmd}\")\n",
    "        print(f\"Return code: {e.returncode}\")\n",
    "        print(f\"Output: {e.stdout}\")\n",
    "        print(f\"Error: {e.stderr}\")\n",
    "        raise\n",
    "    \n",
    "# Run the entire pipeline\n",
    "try:\n",
    "    run_segmentation(t1, subject, threads, sd, allow_root)\n",
    "    run_biasfield_correction(conformed_name, norm_name, asegdkt_segfile, threads)\n",
    "    run_recon_surf(subject, sd, threads)\n",
    "    print(\"FastSurfer pipeline completed successfully.\")\n",
    "    \n",
    "    # Check if the necessary output files are generated\n",
    "    output_files = [\n",
    "        asegdkt_segfile,\n",
    "        aseg_segfile,\n",
    "        conformed_name,\n",
    "        norm_name,\n",
    "        seg_log,\n",
    "        build_log,\n",
    "        asegdkt_statsfile,\n",
    "        cereb_segfile,\n",
    "        cereb_statsfile,\n",
    "        merged_segfile\n",
    "    ]\n",
    "    \n",
    "    missing_files = [f for f in output_files if not os.path.exists(f)]\n",
    "    if missing_files:\n",
    "        print(\"The following expected output files are missing:\")\n",
    "        for f in missing_files:\n",
    "            print(f)\n",
    "    else:\n",
    "        print(\"Output files generated successfully.\")\n",
    "        \n",
    "     \n",
    "        \n",
    "except Exception as e:\n",
    "    print(\"Pipeline failed:\", e)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import nibabel as nib\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import os\n",
    "from skimage.measure import find_contours\n",
    "\n",
    "def find_foramen_of_monro_coronal_slice(segmentation_path, lv_labels=[4, 43], third_ventricle_label=14):\n",
    "    # Load the segmentation file\n",
    "    seg_img = nib.load(segmentation_path)\n",
    "    seg_data = seg_img.get_fdata()\n",
    "\n",
    "    # Get the shape of the segmentation data\n",
    "    shape = seg_data.shape\n",
    "\n",
    "    # Iterate through each coronal slice\n",
    "    for i in range(shape[1]):\n",
    "        coronal_slice = seg_data[:, i, :]\n",
    "\n",
    "        # Find coordinates of lateral ventricles and third ventricle in this slice\n",
    "        lv_coords = np.argwhere(np.isin(coronal_slice, lv_labels))\n",
    "        third_ventricle_coords = np.argwhere(coronal_slice == third_ventricle_label)\n",
    "\n",
    "        if lv_coords.size > 0 and third_ventricle_coords.size > 0:\n",
    "            # Check for intersection of voxels between lateral ventricles and third ventricle\n",
    "            if np.any(np.isin(lv_coords[:, 1], third_ventricle_coords[:, 1])):\n",
    "                return i\n",
    "\n",
    "    return None\n",
    "\n",
    "def calculate_and_visualize_evans_index(segmentation_path, coronal_slice_idx, lv_labels=[4, 43], brain_labels=[2, 41]):\n",
    "    # Load the segmentation file\n",
    "    seg_img = nib.load(segmentation_path)\n",
    "    seg_data = seg_img.get_fdata()\n",
    "\n",
    "    # Extract the coronal slice\n",
    "    coronal_slice = seg_data[:, coronal_slice_idx, :]\n",
    "\n",
    "    # Find coordinates of lateral ventricles and brain regions in this slice\n",
    "    lv_coords = np.argwhere(np.isin(coronal_slice, lv_labels))\n",
    "    brain_coords = np.argwhere(np.isin(coronal_slice, brain_labels))\n",
    "\n",
    "    if lv_coords.size == 0 or brain_coords.size == 0:\n",
    "        print(\"No suitable data found in the specified coronal slice.\")\n",
    "        return\n",
    "\n",
    "    # Find upper and lower extremities to determine the midpoint\n",
    "    max_y = lv_coords[:, 1].max()\n",
    "    min_y = lv_coords[:, 1].min()\n",
    "    midpoint_y = (max_y + min_y) / 2\n",
    "\n",
    "    # Filter points above the midpoint\n",
    "    upper_lv_coords = lv_coords[lv_coords[:, 1] > midpoint_y]\n",
    "\n",
    "    # Check if there are enough points above the midpoint\n",
    "    if upper_lv_coords.shape[0] < 2:\n",
    "        print(\"Not enough points found in the upper part of the lateral ventricles.\")\n",
    "        return\n",
    "\n",
    "    # Calculate the maximum horizontal distance of the lateral ventricles\n",
    "    y_unique = np.unique(upper_lv_coords[:, 1])\n",
    "    max_lv_distance = 0\n",
    "    best_lv_coords = None\n",
    "\n",
    "    for y in y_unique:\n",
    "        y_points = upper_lv_coords[upper_lv_coords[:, 1] == y]\n",
    "        if y_points.shape[0] < 2:\n",
    "            continue\n",
    "\n",
    "        # Calculate the horizontal distance between points in this row\n",
    "        x_coords = y_points[:, 0]\n",
    "        dist = np.max(x_coords) - np.min(x_coords)\n",
    "\n",
    "        if dist > max_lv_distance:\n",
    "            max_lv_distance = dist\n",
    "            y_value = y  # Y-value for horizontal measurement\n",
    "            best_lv_coords = np.array([[np.min(x_coords), y_value], [np.max(x_coords), y_value]])\n",
    "\n",
    "    # Find the maximum left-right distance of the brain contours\n",
    "    contours = find_contours(coronal_slice.T, level=0.5)\n",
    "    max_brain_distance = 0\n",
    "    best_brain_coords = None\n",
    "\n",
    "    for contour in contours:\n",
    "        max_x = contour[:, 1].max()\n",
    "        min_x = contour[:, 1].min()\n",
    "        brain_dist = max_x - min_x\n",
    "        if brain_dist > max_brain_distance:\n",
    "            max_brain_distance = brain_dist\n",
    "            y_value = contour[contour[:, 1] == min_x, 0][0]  # Y-value for horizontal measurement\n",
    "            best_brain_coords = np.array([[min_x, y_value], [max_x, y_value]])\n",
    "\n",
    "    # Calculate the Evans Index\n",
    "    if max_brain_distance > 0:\n",
    "        evans_index = max_lv_distance / max_brain_distance\n",
    "    else:\n",
    "        evans_index = None\n",
    "\n",
    "    # Plot the coronal slice with measurements\n",
    "    plt.figure(figsize=(8, 8))\n",
    "    plt.imshow(coronal_slice.T, cmap='gray', origin='lower')\n",
    "\n",
    "    # Highlight the lateral ventricles\n",
    "    plt.scatter(upper_lv_coords[:, 0], upper_lv_coords[:, 1], c='red', s=1, label='Lateral Ventricles')\n",
    "\n",
    "    # Highlight the maximum width points for the ventricles\n",
    "    if best_lv_coords is not None:\n",
    "        plt.scatter(best_lv_coords[:, 0], best_lv_coords[:, 1], c='yellow', s=10, label='Max Ventricular Width Points')\n",
    "        plt.plot(best_lv_coords[:, 0], best_lv_coords[:, 1], c='yellow')\n",
    "\n",
    "    # Highlight the maximum width points for the brain\n",
    "    if best_brain_coords is not None:\n",
    "        plt.scatter(best_brain_coords[:, 0], best_brain_coords[:, 1], c='blue', s=10, label='Max Brain Width Points')\n",
    "        plt.plot(best_brain_coords[:, 0], best_brain_coords[:, 1], c='blue')\n",
    "\n",
    "    title = f'Lateral Ventricles in Coronal Slice {coronal_slice_idx}\\nMax Ventricular Width: {max_lv_distance:.2f} mm, Max Brain Width: {max_brain_distance:.2f} mm'\n",
    "    if evans_index is not None:\n",
    "        title += f'\\nEvans Index: {evans_index:.2f}'\n",
    "\n",
    "    plt.title(title)\n",
    "    plt.legend()\n",
    "    plt.show()\n",
    "\n",
    "    print(f'Maximum width of frontal horns in slice {coronal_slice_idx}: {max_lv_distance:.2f} mm')\n",
    "    print(f'Maximum left-to-right distance of the brain in slice {coronal_slice_idx}: {max_brain_distance:.2f} mm')\n",
    "    if evans_index is not None:\n",
    "        print(f'Evans Index: {evans_index:.2f}')\n",
    "\n",
    "# Construct the path dynamically\n",
    "segmentation_path = os.path.join(fastsurferdir, subject, \"mri\", \"aparc.DKTatlas+aseg.deep.mgz\")\n",
    "\n",
    "# Find the coronal slice containing the foramen of Monro\n",
    "coronal_slice_idx = find_foramen_of_monro_coronal_slice(segmentation_path)\n",
    "\n",
    "if coronal_slice_idx is not None:\n",
    "    print(f\"Coronal slice containing the foramen of Monro: {coronal_slice_idx}\")\n",
    "    # Calculate and visualize Evans Index for the identified coronal slice\n",
    "    calculate_and_visualize_evans_index(segmentation_path, coronal_slice_idx)\n",
    "else:\n",
    "    print(\"No coronal slice found containing the foramen of Monro.\")\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "AC-PC identify talairach space"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# import nibabel as nib\n",
    "# import numpy as np\n",
    "# import matplotlib.pyplot as plt\n",
    "# import os\n",
    "\n",
    "# def load_talairach_transform(transform_path):\n",
    "#     with open(transform_path, 'r') as file:\n",
    "#         lines = file.readlines()\n",
    "#         matrix = np.array([list(map(float, line.split())) for line in lines[5:9]])\n",
    "#     return matrix\n",
    "\n",
    "# def apply_transform(segmentation_data, transform_matrix):\n",
    "#     shape = segmentation_data.shape\n",
    "#     transformed_data = np.zeros_like(segmentation_data)\n",
    "#     for x in range(shape[0]):\n",
    "#         for y in range(shape[1]):\n",
    "#             for z in range(shape[2]):\n",
    "#                 coords = np.array([x, y, z, 1])\n",
    "#                 transformed_coords = np.dot(transform_matrix, coords)\n",
    "#                 transformed_coords = transformed_coords[:3] / transformed_coords[3]\n",
    "#                 transformed_coords = np.round(transformed_coords).astype(int)\n",
    "#                 if np.all(transformed_coords >= 0) and np.all(transformed_coords < shape):\n",
    "#                     transformed_data[tuple(transformed_coords)] = segmentation_data[x, y, z]\n",
    "#     return transformed_data\n",
    "\n",
    "# def visualize_talairach_origin(segmentation_data):\n",
    "#     origin = [128, 128, 128]  # Talairach space origin in FreeSurfer is typically at (128, 128, 128)\n",
    "#     axial_slice = segmentation_data[:, origin[1], :]\n",
    "\n",
    "#     # Extract third ventricle data\n",
    "#     third_ventricle_coords = np.column_stack(np.where(axial_slice == 14))\n",
    "\n",
    "#     # Find the bottommost point (maximum y value)\n",
    "#     if third_ventricle_coords.size > 0:\n",
    "#         bottommost_point = third_ventricle_coords[np.argmin(third_ventricle_coords[:, 1])]\n",
    "#         pc_point = (bottommost_point[0], origin[1], bottommost_point[1])\n",
    "#     else:\n",
    "#         bottommost_point = None\n",
    "#         pc_point = None\n",
    "\n",
    "#     plt.figure(figsize=(8, 8))\n",
    "#     plt.imshow(axial_slice.T, cmap='gray', origin='lower')\n",
    "#     plt.scatter(origin[0], origin[2], c='red', s=10, label='Talairach Origin (AC)')\n",
    "#     if third_ventricle_coords.size > 0:\n",
    "#         plt.scatter(third_ventricle_coords[:, 0], third_ventricle_coords[:, 1], c='blue', s=1, label='Third Ventricle')\n",
    "#         plt.scatter(bottommost_point[0], bottommost_point[1], c='green', s=10, label='Bottommost Point')\n",
    "#     plt.title(f'Axial Slice {origin[2]} (Talairach Origin)')\n",
    "#     plt.legend()\n",
    "#     plt.show()\n",
    "\n",
    "#     ac_point = (origin[0], origin[1], origin[2])\n",
    "#     return ac_point, pc_point\n",
    "\n",
    "# # Usage\n",
    "# fastsurferdir = \"/Users/suhanasaigoli/Desktop/Projects/SURF/iNPH/FastSurfer-Analysis\"\n",
    "# subject = \"OAS_0001-1\"\n",
    "# segmentation_path = os.path.join(fastsurferdir, subject, \"mri\", \"aparc.DKTatlas+aseg.deep.mgz\")\n",
    "# transform_path = os.path.join(fastsurferdir, subject, \"mri\", \"transforms\", \"talairach.lta\")\n",
    "\n",
    "# # Load segmentation data\n",
    "# seg_img = nib.load(segmentation_path)\n",
    "# seg_data = seg_img.get_fdata()\n",
    "\n",
    "# # Load Talairach transform\n",
    "# transform_matrix = load_talairach_transform(transform_path)\n",
    "\n",
    "# # Apply transform to segmentation data\n",
    "# transformed_data = apply_transform(seg_data, transform_matrix)\n",
    "\n",
    "# # Visualize the origin of the Talairach space and get AC and PC points\n",
    "# ac_point, pc_point = visualize_talairach_origin(transformed_data)\n",
    "\n",
    "# print(f\"AC Point: {ac_point}\")\n",
    "# print(f\"PC Point: {pc_point}\")\n"
   ]
  }
 ],
 "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": "3.10.14"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}