electronmicroscopy · arthurmccray · Jan 31, 2026 · Jan 31, 2026 · Jan 31, 2026 · Feb 2, 2026
diff --git a/.gitignore b/.gitignore
@@ -1,6 +1,6 @@
 .ipynb_checkpoints
 *.zip
-*.npy
+# *.npy
 .DS_Store
 dvlp_notebooks
-notebooks/runs/vols/
+outputs/
diff --git a/data/phantom.npy b/data/phantom.npy
diff --git a/data/tilt_angles_1_deg_tilt_axis.npy b/data/tilt_angles_1_deg_tilt_axis.npy
diff --git a/data/tilt_series_1_deg_tilt_axis.npy b/data/tilt_series_1_deg_tilt_axis.npy
diff --git a/notebooks/tomography/runs/logs/auxiliary_params.pt b/notebooks/tomography/runs/logs/auxiliary_params.pt
diff --git a/notebooks/tomography/runs/logs/events.out.tfevents.1765006677.mallard.2288855.1 b/notebooks/tomography/runs/logs/events.out.tfevents.1765006677.mallard.2288855.1
diff --git a/notebooks/tomography/tomo_01.ipynb b/notebooks/tomography/tomo_01.ipynb
diff --git a/notebooks/core/config.ipynb → tutorials/core/config.ipynb b/notebooks/core/config.ipynb → tutorials/core/config.ipynb
diff --git a/notebooks/core/dp_augment.ipynb → tutorials/core/dp_augment.ipynb b/notebooks/core/dp_augment.ipynb → tutorials/core/dp_augment.ipynb
diff --git a/...ore/neural_networks_01_architecture.ipynb → ...ore/neural_networks_01_architecture.ipynb b/...ore/neural_networks_01_architecture.ipynb → ...ore/neural_networks_01_architecture.ipynb
diff --git a/notebooks/core/neural_networks_02_unet.ipynb → tutorials/core/neural_networks_02_unet.ipynb b/notebooks/core/neural_networks_02_unet.ipynb → tutorials/core/neural_networks_02_unet.ipynb
diff --git a/...core/neural_networks_03_autoencoder.ipynb → ...core/neural_networks_03_autoencoder.ipynb b/...core/neural_networks_03_autoencoder.ipynb → ...core/neural_networks_03_autoencoder.ipynb
diff --git a/notebooks/diffraction/placeholder.ipynb → tutorials/diffraction/placeholder.ipynb b/notebooks/diffraction/placeholder.ipynb → tutorials/diffraction/placeholder.ipynb
diff --git a/...maging/apoF_dataset_phase_retrieval.ipynb → ...maging/apoF_dataset_phase_retrieval.ipynb b/...maging/apoF_dataset_phase_retrieval.ipynb → ...maging/apoF_dataset_phase_retrieval.ipynb
diff --git a/.../crystalline_sample_phase_retrieval.ipynb → .../crystalline_sample_phase_retrieval.ipynb b/.../crystalline_sample_phase_retrieval.ipynb → .../crystalline_sample_phase_retrieval.ipynb
diff --git a/...aging/ducky_dataset_phase_retrieval.ipynb → ...aging/ducky_dataset_phase_retrieval.ipynb b/...aging/ducky_dataset_phase_retrieval.ipynb → ...aging/ducky_dataset_phase_retrieval.ipynb
diff --git a/notebooks/hyperparameter_optimization.ipynb → ...imaging/hyperparameter_optimization.ipynb b/notebooks/hyperparameter_optimization.ipynb → ...imaging/hyperparameter_optimization.ipynb
diff --git a/...ractive_imaging/ptycho_iter_01_lite.ipynb → ...ractive_imaging/ptycho_iter_01_lite.ipynb b/...ractive_imaging/ptycho_iter_01_lite.ipynb → ...ractive_imaging/ptycho_iter_01_lite.ipynb
diff --git a/...ractive_imaging/ptycho_iter_02_full.ipynb → ...ractive_imaging/ptycho_iter_02_full.ipynb b/...ractive_imaging/ptycho_iter_02_full.ipynb → ...ractive_imaging/ptycho_iter_02_full.ipynb
diff --git a/...active_imaging/ptycho_iter_03_MOSS6.ipynb → ...active_imaging/ptycho_iter_03_MOSS6.ipynb b/...active_imaging/ptycho_iter_03_MOSS6.ipynb → ...active_imaging/ptycho_iter_03_MOSS6.ipynb
diff --git a/...ging/ptycho_iter_04_hyperparameters.ipynb → ...ging/ptycho_iter_04_hyperparameters.ipynb b/...ging/ptycho_iter_04_hyperparameters.ipynb → ...ging/ptycho_iter_04_hyperparameters.ipynb
diff --git a/...ive_imaging/ptycho_iter_05_workflow.ipynb → ...ive_imaging/ptycho_iter_05_workflow.ipynb b/...ive_imaging/ptycho_iter_05_workflow.ipynb → ...ive_imaging/ptycho_iter_05_workflow.ipynb
diff --git a/...active_imaging/ptycho_iter_XX_noise.ipynb → ...active_imaging/ptycho_iter_XX_noise.ipynb b/...active_imaging/ptycho_iter_XX_noise.ipynb → ...active_imaging/ptycho_iter_XX_noise.ipynb
diff --git a/...n_notebooks/apoferritin-pdb-dataset.ipynb → ...n_notebooks/apoferritin-pdb-dataset.ipynb b/...n_notebooks/apoferritin-pdb-dataset.ipynb → ...n_notebooks/apoferritin-pdb-dataset.ipynb
diff --git a/...lation_notebooks/crystalline_sample.ipynb → ...lation_notebooks/crystalline_sample.ipynb b/...lation_notebooks/crystalline_sample.ipynb → ...lation_notebooks/crystalline_sample.ipynb
diff --git a/.../simulation_notebooks/ducky-dataset.ipynb → .../simulation_notebooks/ducky-dataset.ipynb b/.../simulation_notebooks/ducky-dataset.ipynb → .../simulation_notebooks/ducky-dataset.ipynb
diff --git a/...lation_notebooks/white-noise-object.ipynb → ...lation_notebooks/white-noise-object.ipynb b/...lation_notebooks/white-noise-object.ipynb → ...lation_notebooks/white-noise-object.ipynb
diff --git a/.../white_noise_object_phase_retrieval.ipynb → .../white_noise_object_phase_retrieval.ipynb b/.../white_noise_object_phase_retrieval.ipynb → .../white_noise_object_phase_retrieval.ipynb
diff --git a/notebooks/imaging/placeholder.ipynb → tutorials/imaging/placeholder.ipynb b/notebooks/imaging/placeholder.ipynb → tutorials/imaging/placeholder.ipynb
diff --git a/notebooks/spectroscopy/placeholder.ipynb → tutorials/spectroscopy/placeholder.ipynb b/notebooks/spectroscopy/placeholder.ipynb → tutorials/spectroscopy/placeholder.ipynb
diff --git a/tutorials/tomography/hpc/README.md b/tutorials/tomography/hpc/README.md
@@ -0,0 +1,6 @@
+# HPC (NERSC) Tomography Reconstructions
+
+Included in this folder are two files:
+
+- `run.sh`: Initializes all necessary parameters for `torch` to see all gpus across nodes using `torchrun`.
+- `tomography_recon.py`: Contains the full reconstruction script; similar to `Scripts/tomography_02_full.py`
diff --git a/tutorials/tomography/hpc/run.sh b/tutorials/tomography/hpc/run.sh
@@ -0,0 +1,12 @@
+#!/bin/bash
+
+export MASTER_ADDR=$(scontrol show hostnames $SLURM_JOB_NODELIST | head -n 1)
+export MASTER_PORT=29500
+export OMP_NUM_THREADS=8
+ml nccl/2.24.3
+ml conda
+ml nccl
+conda activate /global/common/software/mxxxx/'INSERT USER HERE'/conda/quantem
+
+
+srun -l torchrun --nnodes=$SLURM_JOB_NUM_NODES --nproc-per-node=$SLURM_GPUS_PER_NODE --rdzv-backend=c10d --rdzv-endpoint=$MASTER_ADDR:$MASTER_PORT tomography_recon.py
diff --git a/tutorials/tomography/hpc/run_job.sh b/tutorials/tomography/hpc/run_job.sh
@@ -0,0 +1,21 @@
+#!/bin/bash
+#SBATCH -A mxxxx
+#SBATCH -C gpu
+#SBATCH -q regular
+#SBATCH -t 4:00:00
+#SBATCH -N 4
+#SBATCH --ntasks-per-node=1
+#SBATCH -c 128
+#SBATCH --gpus-per-node=4
+#SBATCH --gpu-bind=none
+#SBATCH --mail-type=begin,end,fail
+#SBATCH --mail-user=xxx@xxx.xxx
+
+export MASTER_ADDR=$(scontrol show hostnames $SLURM_JOB_NODELIST | head -n 1)
+export MASTER_PORT=29500
+export OMP_NUM_THREADS=8
+ml nccl/2.24.3
+ml conda
+ml nccl
+conda activate /global/common/software/mxxxx/'INSERT USER HERE'/conda/quantem
+srun -l torchrun --nnodes=$SLURM_JOB_NUM_NODES --nproc-per-node=$SLURM_GPUS_PER_NODE --rdzv-backend=c10d --rdzv-endpoint=$MASTER_ADDR:$MASTER_PORT tomography_recon.py
diff --git a/tutorials/tomography/hpc/tomography_recon.py b/tutorials/tomography/hpc/tomography_recon.py
@@ -0,0 +1,131 @@
+from quantem.tomography.tomography import TomographyConventional, Tomography
+from quantem.tomography.dataset_models import TomographyPixDataset, TomographyINRDataset, TomographyINRPretrainDataset
+from quantem.tomography.object_models import ObjectINR, ObjectPixelated
+from quantem.tomography.logger_tomography import LoggerTomography
+from quantem.core.ml.inr import HSiren
+import numpy as np
+
+from quantem.tomography.utils import fourier_cropping
+from quantem.core.visualization import show_2d
+import torch
+
+"""
+Example script to run the tomography reconstruction on HPC (NERSC).
+
+- Major difference is not needing to specify the device, as it will be automatically set by the DDP framework.
+
+*Cedric Lim, 2/2/26*
+"""
+
+
+# Load Phantom Dataset
+tilt_series = np.load('../../../data/tilt_series.npy')
+tilt_angles = np.load('../../../data/tilt_angles.npy')
+
+tilt_series = np.array([fourier_cropping(img, (100, 100)) for img in tilt_series]) # Cropped down to 100x100 for speed
+
+dset = TomographyINRDataset(
+    tilt_stack = tilt_series,
+    tilt_angles = tilt_angles,
+)
+
+# Initialize INR Model
+model = HSiren(alpha = 1, winner_initialization = True)
+
+# Initialize INR Object
+obj_inr = ObjectINR(
+    shape = (100, 100, 100),
+    model = model,
+)
+
+# Define a logger
+logger = LoggerTomography(
+    log_dir = "../../../outputs/tomography/tutorial_02_scripts/",
+    run_prefix = "inr_tomography_warmup_cosineanneal_hpc",
+    run_suffix = "",
+    log_images_every = 2,
+)
+
+# Initialize INR-Based Tomography Object
+tomo_inr = Tomography(
+    dset = dset,
+    obj_model = obj_inr,
+    logger = logger,
+)
+
+# Define optimizer and scheduler parameters - only optimizing the object.
+
+optimizer_params = {
+    "object": {
+        "type": "adam",
+        "lr": 5e-4,
+    },
+}
+
+scheduler_params = {
+    "object": {
+        "type": "linear",
+    },
+}
+
+# Define constraints
+
+constraints = {
+    "tv_vol": 5e-7,
+    "positivity": True,
+}
+
+# Warmup Schedule for 10 epochs
+
+num_samples_per_ray = [
+    (0, 20),
+    (1, 20),
+    (2, 40),
+    (3, 40),
+    (4, 60),
+    (4, 60),
+    (6, 80),
+    (7, 80),
+    (8, 100),
+    (9, 100),
+]
+
+tomo_inr.reconstruct(
+    num_iter = 10,
+    optimizer_params = optimizer_params,
+    scheduler_params = scheduler_params,
+    constraints = constraints,
+    num_samples_per_ray = num_samples_per_ray,
+    num_workers = 32,
+)
+
+
+# Initialzie pose optimizer 
+
+optimizer_params = {
+    "pose": {
+        "type": "adam",
+        "lr": 1e-2,
+    }
+}
+
+# Define new schedulers for both optimizers using CosineAnnealing
+scheduler_params = {
+    "object": {
+        "type": "cosine_annealing",
+    },
+    "pose": {
+        "type": "cosine_annealing",
+    }
+}
+
+# Reconstruct
+
+tomo_inr.reconstruct(
+    num_iter = 100,
+    optimizer_params = optimizer_params,
+    scheduler_params = scheduler_params,
+    num_samples_per_ray = 100,
+)
+
+
diff --git a/tutorials/tomography/notebooks/tomography_01_lite.ipynb b/tutorials/tomography/notebooks/tomography_01_lite.ipynb
diff --git a/tutorials/tomography/notebooks/tomography_02_full.ipynb b/tutorials/tomography/notebooks/tomography_02_full.ipynb
diff --git a/tutorials/tomography/notebooks/tomography_03_preprocess.ipynb b/tutorials/tomography/notebooks/tomography_03_preprocess.ipynb
diff --git a/tutorials/tomography/scripts/tomography_01_lite.py b/tutorials/tomography/scripts/tomography_01_lite.py
@@ -0,0 +1,52 @@
+from quantem.tomography.dataset_models import TomographyPixDataset, TomographyINRDataset
+from quantem.tomography.tomography_lite import TomographyLiteConv, TomographyLiteINR
+import numpy as np
+from quantem.core.visualization import show_2d
+
+from quantem.tomography.utils import fourier_cropping
+
+"""
+Example script to run the lite version of the INR-based tomography pipeline.
+
+*Cedric Lim, 2/1/26*
+"""
+
+# Load Phantom Dataset
+tilt_series = np.load('../../../data/tilt_series.npy')
+tilt_angles = np.load('../../../data/tilt_angles.npy')
+
+# Fourier crop
+tilt_series = np.array([fourier_cropping(img, (100, 100)) for img in tilt_series]) # Cropped down to 100x100 for speed
+
+# Initialize dataset
+dset = TomographyINRDataset(
+    tilt_stack = tilt_series,
+    tilt_angles = tilt_angles,
+)
+
+# Initialize tomography object
+tomography_inr = TomographyLiteINR.from_dataset(
+    dset,
+    device = 'cuda:0',
+    log_dir = '../../../outputs/tomography/tutorial_01_scripts/tomo_inr_lite',
+    log_images_every = 2,
+)
+
+
+constraints = {
+    "positivity": True,
+    "tv_vol": 1e-6,
+}
+
+tomography_inr.reconstruct(
+    num_iter = 50,
+    # reset = True,
+    obj_lr = 1e-4,
+    pose_lr = 1e-2,
+    batch_size = 1024,
+    num_workers = 32,
+    learn_pose = True,
+    scheduler_type = "cosine_annealing",
+    constraints = constraints,
+)
+
diff --git a/tutorials/tomography/scripts/tomography_02_full.py b/tutorials/tomography/scripts/tomography_02_full.py
@@ -0,0 +1,132 @@
+from quantem.tomography.tomography import TomographyConventional, Tomography
+from quantem.tomography.dataset_models import TomographyPixDataset, TomographyINRDataset, TomographyINRPretrainDataset
+from quantem.tomography.object_models import ObjectINR, ObjectPixelated
+from quantem.tomography.logger_tomography import LoggerTomography
+from quantem.core.ml.inr import HSiren
+import numpy as np
+
+from quantem.tomography.utils import fourier_cropping
+from quantem.core.visualization import show_2d
+import torch
+
+"""
+Example script to run the full version of the INR-based tomography pipeline. In this case, doing the full warmup procedure to cosine annealing.
+
+*Cedric Lim, 2/1/26*
+"""
+
+device = "cuda:2"
+
+# Load Phantom Dataset
+tilt_series = np.load('../../../data/tilt_series.npy')
+tilt_angles = np.load('../../../data/tilt_angles.npy')
+
+tilt_series = np.array([fourier_cropping(img, (100, 100)) for img in tilt_series]) # Cropped down to 100x100 for speed
+
+dset = TomographyINRDataset(
+    tilt_stack = tilt_series,
+    tilt_angles = tilt_angles,
+)
+
+# Initialize INR Model
+model = HSiren(alpha = 1, winner_initialization = True)
+
+# Initialize INR Object
+obj_inr = ObjectINR(
+    shape = (100, 100, 100),
+    device = device,
+    model = model,
+)
+
+# Define a logger
+logger = LoggerTomography(
+    log_dir = "../../../outputs/tomography/tutorial_02_scripts/",
+    run_prefix = "inr_tomography_warmup_cosineanneal",
+    run_suffix = "",
+    log_images_every = 2,
+)
+
+# Initialize INR-Based Tomography Object
+tomo_inr = Tomography(
+    dset = dset,
+    obj_model = obj_inr,
+    device = device,
+    logger = logger,
+)
+
+# Define optimizer and scheduler parameters - only optimizing the object.
+
+optimizer_params = {
+    "object": {
+        "type": "adam",
+        "lr": 1e-4,
+    },
+}
+
+scheduler_params = {
+    "object": {
+        "type": "linear",
+    },
+}
+
+# Define constraints
+
+constraints = {
+    "tv_vol": 5e-7,
+    "positivity": True,
+}
+
+# Warmup Schedule for 10 epochs
+
+num_samples_per_ray = [
+    (0, 20),
+    (1, 20),
+    (2, 40),
+    (3, 40),
+    (4, 60),
+    (4, 60),
+    (6, 80),
+    (7, 80),
+    (8, 100),
+    (9, 100),
+]
+
+tomo_inr.reconstruct(
+    num_iter = 10,
+    optimizer_params = optimizer_params,
+    scheduler_params = scheduler_params,
+    constraints = constraints,
+    num_samples_per_ray = num_samples_per_ray,
+    num_workers = 32,
+)
+
+
+# Initialzie pose optimizer 
+
+optimizer_params = {
+    "pose": {
+        "type": "adam",
+        "lr": 1e-2,
+    }
+}
+
+# Define new schedulers for both optimizers using CosineAnnealing
+scheduler_params = {
+    "object": {
+        "type": "cosine_annealing",
+    },
+    "pose": {
+        "type": "cosine_annealing",
+    }
+}
+
+# Reconstruct
+
+tomo_inr.reconstruct(
+    num_iter = 20,
+    optimizer_params = optimizer_params,
+    scheduler_params = scheduler_params,
+    num_samples_per_ray = 100,
+)
+
+