### Install GRIDAY using Python

Source: https://github.com/hspark1212/moftransformer/blob/master/docs/source/installation.md

Install the GRIDAY tool programmatically by calling a Python function. This is an alternative to the command-line installation.

```python
from moftransformer.utils import install_griday

install_griday()
```

--------------------------------

### Install GRIDAY for MOFTransformer

Source: https://github.com/hspark1212/moftransformer/blob/master/docs/source/getting_started/prepare_data.md

Use this command to install the GRIDAY tool, which is required for calculating energy-grids from cif files for MOFTransformer.

```bash
moftransformer install-griday
```

--------------------------------

### Example Command-Line Training

Source: https://github.com/hspark1212/moftransformer/blob/master/docs/source/getting_started/training.md

A practical example of using the command-line interface for training. This command specifies the dataset root, downstream task, number of epochs, number of devices, and batch size.

```bash
$ moftransformer run --root_dataset './data' --downstream 'example' --max_epochs 10 --devices 2 --batch_size 216
```

--------------------------------

### Install MOFTransformer using PIP

Source: https://github.com/hspark1212/moftransformer/blob/master/docs/source/installation.md

Use this command to install the MOFTransformer package directly from PyPI. Ensure Python and PyTorch are installed first.

```bash
pip install moftransformer
```

--------------------------------

### Train for Multi-Task Learning

Source: https://github.com/hspark1212/moftransformer/blob/master/docs/source/getting_started/training.md

Example demonstrating multi-task learning setup using `moftransformer.run`. This configuration is suitable when dealing with multiple target variables, requiring lists for `mean`, `std`, and specifying `n_targets`.

```python
import moftransformer
from moftransformer.examples import example_path

# data root and downstream from example
root_dataset = example_path['root_dataset']
downstream = example_path['downstream']
log_dir = './logs/'
# load_path = "pmtransformer" (default)

# kwargs (optional)
max_epochs = 10
batch_size = 8
mean = [0, 1, 2]
std = [1, 2, 3]
n_targets = 3

moftransformer.run(root_dataset, downstream, log_dir=log_dir,                   
                   max_epochs=max_epochs, batch_size=batch_size,
                   mean=mean, std=std, n_targets=n_targets)
```

--------------------------------

### Install CGCNN Prerequisites

Source: https://github.com/hspark1212/moftransformer/blob/master/baseline_model/CGCNN/README.md

Commands to create a conda environment and install necessary dependencies for the CGCNN package.

```bash
conda upgrade conda
conda create -n cgcnn python=3 scikit-learn pytorch torchvision pymatgen -c pytorch -c conda-forge
```

--------------------------------

### Install GRIDAY using Make

Source: https://github.com/hspark1212/moftransformer/blob/master/docs/source/installation.md

Manually install GRIDAY using the Makefile if automatic installation fails. This process may require updating GCC/G++ versions, especially in Anaconda environments.

```bash
$ conda install -c conda-forge gcc=9.5.0
$ conda install -c conda-forge gxx=9.5.0

$ cd [PATH_MOFTransformer]/libs/GRIDAY/  # move to path of griday-file
$ make              # make Makefile
$ cd scripts/
$ make              # make Makefile
```

--------------------------------

### Run MOFTransformer Data Preparation

Source: https://github.com/hspark1212/moftransformer/blob/master/docs/source/getting_started/prepare_data.md

This Python script demonstrates how to use the `prepare_data` function to generate atom-wise graph embeddings and energy-grids from cif files. Ensure GRIDAY is installed and cif files are in the specified directory.

```python
from moftransformer.examples import example_path
from moftransformer.utils import prepare_data

# Get example path
root_cifs = example_path['root_cif']
root_dataset = example_path['root_dataset']
downstream = example_path['downstream']

train_fraction = 0.8  # default value
test_fraction = 0.1   # default value

# Run prepare data
prepare_data(root_cifs, root_dataset, downstream=downstream, 
             train_fraction=train_fraction, test_fraction=test_fraction)
```

--------------------------------

### Editable Install MOFTransformer

Source: https://github.com/hspark1212/moftransformer/blob/master/docs/source/installation.md

Clone the repository and install in editable mode to modify the source code. This requires Git and a Python environment.

```bash
git clone https://github.com/hspark1212/MOFTransformer.git
cd MOFTransformer
pip install -e .
```

--------------------------------

### Train PMTransformer (Single-Task)

Source: https://github.com/hspark1212/moftransformer/blob/master/docs/source/getting_started/training.md

Example of training a PMTransformer for a single downstream task using the `moftransformer.run` function. Ensure the dataset and downstream task are correctly specified. Optional keyword arguments can configure training parameters.

```python
import moftransformer
from moftransformer.examples import example_path

# data root and downstream from example
root_dataset = example_path['root_dataset']
downstream = example_path['downstream']
log_dir = './logs/'
# load_path = "pmtransformer" (default)

# kwargs (optional)
max_epochs = 10
batch_size = 8
mean = 0
std = 1


moftransformer.run(root_dataset, downstream, log_dir=log_dir,                   
                   max_epochs=max_epochs, batch_size=batch_size,
                   mean=mean, std=std)
```

--------------------------------

### Training MOFTransformer using Python `run` function

Source: https://github.com/hspark1212/moftransformer/blob/master/docs/source/getting_started/training.md

This section details how to train MOFTransformers using the `moftransformer.run` function, including parameter explanations and examples for single-task and multi-task learning.

```APIDOC
## MOFTRANSFORMER.RUN

### Description
Trains a MOFTransformer model with specified dataset and configuration.

### Method
`moftransformer.run`

### Parameters
- **root_dataset** (string) - Required - A folder containing graph data, grid data, and json of MOFs.
- **downstream** (string) - Required - Name of user-specific task (e.g. bandgap, gasuptake).
- **log_dir** (string) - Optional - Directory to save log, models, and params. (default: `logs/`)
- **test_only** (boolean) - Optional - If True, only the test process is performed without the learning model. (default: `False`)
- **kwargs** (dict) - Optional - Configuration for MOFTransformer, including:
  - **max_epochs** (int) - Maximum number of training epochs.
  - **batch_size** (int) - Batch size for training.
  - **mean** (float or list of floats) - Mean value(s) for normalization.
  - **std** (float or list of floats) - Standard deviation value(s) for normalization.
  - **n_targets** (int) - Number of targets for multi-task learning.

### Request Example (Single-task)
```python
import moftransformer
from moftransformer.examples import example_path

root_dataset = example_path['root_dataset']
downstream = example_path['downstream']
log_dir = './logs/'

max_epochs = 10
batch_size = 8
mean = 0
std = 1

moftransformer.run(root_dataset, downstream, log_dir=log_dir,                   
                   max_epochs=max_epochs, batch_size=batch_size,
                   mean=mean, std=std)
```

### Request Example (Multi-task)
```python
import moftransformer
from moftransformer.examples import example_path

root_dataset = example_path['root_dataset']
downstream = example_path['downstream']
log_dir = './logs/'

max_epochs = 10
batch_size = 8
mean = [0, 1, 2]
std = [1, 2, 3]
n_targets = 3

moftransformer.run(root_dataset, downstream, log_dir=log_dir,                   
                   max_epochs=max_epochs, batch_size=batch_size,
                   mean=mean, std=std, n_targets=n_targets)
```

### Post-training
After training, the trained model, logs, and hyperparameters will be saved at `log_dir`. You can monitor the results using tensorboard:

```bash
$ tensorboard --logdir=[log_dir] --bind_all
```
```

--------------------------------

### Training MOFTransformer using Command-Line Interface

Source: https://github.com/hspark1212/moftransformer/blob/master/docs/source/getting_started/training.md

This section explains how to initiate MOFTransformer training directly from the command line, mirroring the parameters available in the Python `run` function.

```APIDOC
## MOFTRANSFORMER RUN (Command-Line)

### Description
Initiates MOFTransformer training using the command-line interface.

### Method
`moftransformer run`

### Endpoint
`moftransformer run --root_dataset [root_dataset] --downstream [downstream] --logdir [logdir] ...`

### Parameters
Parameters are similar to the Python `moftransformer.run` function. Use `-h` for a full list.

### Example
```bash
$ moftransformer run --root_dataset './data' --downstream 'example' --max_epochs 10 --devices 2 --batch_size 216
```

### Help
For detailed parameter information, run:

```bash
$ moftransformer run -h
```
```

--------------------------------

### View Training Results with TensorBoard

Source: https://github.com/hspark1212/moftransformer/blob/master/docs/source/getting_started/training.md

Command to launch TensorBoard for visualizing training logs, models, and hyperparameters saved in the specified log directory. Use the `--bind_all` flag to make it accessible externally.

```bash
tensorboard --logdir=[log_dir] --bind_all
```

--------------------------------

### Download MOFTransformer Models and Data via Command-Line

Source: https://github.com/hspark1212/moftransformer/blob/master/docs/source/installation.md

Use the command-line interface to download pre-trained models, fine-tuned models, or datasets for CoREMOF and QMOF. Specify the target and optionally an output directory.

```bash
moftransformer download [target] (--outdir outdir) (--remove_tarfile)
```

```bash
# download pre-trained model (required)
$ moftransformer download pretrain_model

# download graph-data and graph-data for CoREMOF (optional)
$ moftransformer download coremof

# download graph-data and graph-data for QMOF (optional)
$ moftransformer download qmof

# download fine-tuned model (optional)
$ mofransformer download finetuned_model
```

--------------------------------

### Initialize PatchVisualizer

Source: https://github.com/hspark1212/moftransformer/blob/master/docs/source/getting_started/visualization.md

Methods to instantiate the visualizer using either a specific CIF name or an index from a directory.

```python
from moftransformer.visualize import PatchVisualizer
from moftransformer.examples import visualize_example_path

model_path = "examples/finetuned_bandgap.ckpt" # or 'examples/finetuned_h2_uptake.ckpt'
data_path = visualize_example_path
cifname = 'MIBQAR01_FSR'

vis = PatchVisualizer.from_cifname(cifname, model_path, data_path)
vis.draw_graph()
```

```python
from moftransformer.visualize import PatchVisualizer
from moftransformer.examples import visualize_example_path

model_path = "examples/finetuned_bandgap.ckpt" # or 'examples/finetuned_h2_uptake.ckpt'
data_path = visualize_example_path
batch_id = 0

vis = PatchVisualizer.from_index(0, model_path, data_path)
```

--------------------------------

### Train MOFTransformer via Command-Line

Source: https://github.com/hspark1212/moftransformer/blob/master/docs/source/getting_started/training.md

Initiate MOFTransformer training directly from the command line using the `run` subcommand. This method accepts parameters equivalent to the Python `run` function, allowing for flexible configuration.

```bash
$ moftransformer run --root_dataset [root_dataset] --downstream [downstream] --logdir [logdir] ...
```

--------------------------------

### Download Finetuned Models

Source: https://github.com/hspark1212/moftransformer/blob/master/docs/source/getting_started/visualization.md

Command to download pre-trained model checkpoints for analysis.

```bash
$ moftransformer download finetuned_model -o ./examples
```

--------------------------------

### Download Pretrained Model and Database

Source: https://github.com/hspark1212/moftransformer/blob/master/docs/source/tutorial.md

Commands to download the required pretrained model and the hMOF dataset.

```bash
$ moftransformer download pretrain_model
```

```bash
$ moftransformer download hmof -o ./hmof 
```

--------------------------------

### Download Pre-trained Model

Source: https://github.com/hspark1212/moftransformer/blob/master/docs/source/getting_started/training.md

Use this command to download a pre-trained MOFTransformer model for further training or fine-tuning.

```bash
moftransformer download pretrain_model
```

--------------------------------

### Train a CGCNN model with custom dataset

Source: https://github.com/hspark1212/moftransformer/blob/master/baseline_model/CGCNN/README.md

Use this command to train a CGCNN model. Specify the root directory for your dataset. You can control the data split using `--train-size`, `--val-size`, `--test-size` or `--train-ratio`, `--val-ratio`, `--test-ratio` flags. Do not use size and ratio flags simultaneously.

```bash
python main.py root_dir
```

```bash
python main.py --train-size 6 --val-size 2 --test-size 2 data/sample-regression
```

```bash
python main.py --train-ratio 0.6 --val-ratio 0.2 --test-ratio 0.2 data/sample-regression
```

```bash
python main.py --task classification --train-size 5 --val-size 2 --test-size 3 data/sample-classification
```

--------------------------------

### Import PatchVisualizer

Source: https://github.com/hspark1212/moftransformer/blob/master/docs/source/getting_started/visualization.md

Initial import required to access visualization tools.

```python
from visualize import PatchVisualizer
```

--------------------------------

### Test MOFTransformer via Command Line

Source: https://github.com/hspark1212/moftransformer/blob/master/docs/source/getting_started/test.md

Executes model testing directly from the terminal using the moftransformer CLI.

```bash
$ moftransformer test --root_dataset './data' --downstream 'example' --load_path 'model.ckpt' --mean 0 --std 1
```

--------------------------------

### Download Pre-embeddings for CoREMOF, QMOF

Source: https://github.com/hspark1212/moftransformer/blob/master/README.md

Download pre-computed embeddings for CoREMOF and QMOF datasets. These embeddings serve as inputs for the PMTransformer model.

```bash
moftransformer download coremof
```

```bash
moftransformer download qmof
```

--------------------------------

### Download MOFTransformer Models and Data using Python

Source: https://github.com/hspark1212/moftransformer/blob/master/docs/source/installation.md

Download necessary models and datasets programmatically using Python functions. Optional arguments for output directory and tar file removal are available.

```python
from moftransformer.utils.download import (
    download_pretrain_model,
    download_qmof,
    download_coremof,
    download_hmof,
    download_finetuned_model
)

# download pre-trained model
download_pretrain_model()
# download coremof
download_coremof()
# download qmof
download_qmof()
# download finetuned_model
download_finetuned_model()
```

--------------------------------

### Initialize PatchVisualizer with Matplotlib Widget

Source: https://github.com/hspark1212/moftransformer/blob/master/README.md

Initializes the PatchVisualizer using a matplotlib widget backend, suitable for interactive environments. Ensure the model and data paths are correctly specified.

```python
%matplotlib widget
from visualize import PatchVisualizer

model_path = "examples/finetuned_bandgap.ckpt" # or 'examples/finetuned_h2_uptake.ckpt'
data_path = 'examples/visualize/dataset/'
cifname = 'MIBQAR01_FSR'

vis = PatchVisualizer.from_cifname(cifname, model_path, data_path)
vis.draw_graph()
```

--------------------------------

### Predict material properties with a pre-trained CGCNN model

Source: https://github.com/hspark1212/moftransformer/blob/master/baseline_model/CGCNN/README.md

Use this command to predict properties using a pre-trained model. Provide the path to the pre-trained model file and the root directory of the dataset for which you want to predict properties. For classification tasks, the output probability is between 0 and 1.

```bash
python predict.py pre-trained.pth.tar root_dir
```

```bash
python predict.py pre-trained/formation-energy-per-atom.pth.tar data/sample-regression
```

```bash
python predict.py pre-trained/semi-metal-classification.pth.tar data/sample-classification
```

--------------------------------

### Visualize Feature Importance with PatchVisualizer (Grid)

Source: https://github.com/hspark1212/moftransformer/blob/master/README.md

Visualize the feature importance analysis of energy-grid embeddings for a fine-tuned model. Requires a trained model checkpoint and dataset.

```python
vis = PatchVisualizer.from_cifname(cifname, model_path, data_path)
vis.draw_grid()
```

--------------------------------

### Test MOFTransformer

Source: https://github.com/hspark1212/moftransformer/blob/master/docs/source/tutorial.md

Execution of the testing phase by loading a saved checkpoint file.

```python
import moftransformer

root_dataset = './hmof/downstream_release'
downstream = 'raspa_100bar'
load_path = './logs/pretrained_mof_seed0_from_pretrained_model/version_0/checkpoints/last.ckpt' 
             # cpkt : ./logs/[target_model_path]/[version]/checkpoints/[model].ckpt
max_epochs = 20
mean = 487.841
std = 63.088
batch_size = 32

moftransformer.run(root_dataset, downstream, test_only=True, load_path=load_path, 
                   max_epochs=max_epochs, mean=mean, std=std, batch_size=batch_size)
```

--------------------------------

### Test MOFTransformer with TensorBoard logging

Source: https://github.com/hspark1212/moftransformer/blob/master/docs/source/getting_started/test.md

Uses the run function with test_only set to True to enable TensorBoard logging during evaluation.

```python
from pathlib import Path
import moftransformer
from moftransformer.examples import example_path

root_dataset = example_path['root_dataset']
downstream = example_path['downstream']

# Get ckpt file
seed = 0               # default seeds
version = 0            # version for model. It increases with the number of trains

# For version > 2.1.1, best.ckpt exists.
checkpoint = 'best'    # Epochs where the model is stored. 

# optional keyword
mean = 0
std = 1

load_path = Path(log_dir) / f'pretrained_mof_seed{seed}_from_pmtransformer/version_{version}/checkpoints/{checkpoint}.ckpt'

if not load_path.exists():
    raise ValueError(f'load_path does not exists. check path for .ckpt file : {load_path}')

moftransformer.run(root_dataset, downstream=downstream, log_dir='logs/',
                   test_only=True, load_path=load_path, 
                   mean=mean, std=std)
```

--------------------------------

### Predicting via Command-Line Interface

Source: https://github.com/hspark1212/moftransformer/blob/master/docs/source/getting_started/predict.md

Execute predictions directly from the terminal using the moftransformer CLI.

```bash
$ moftransformer predict --root_dataset [root_dataset] --load_path [load_path]--downstream [downstream] --split [split] --save_dir [save_dir] ...
```

```bash
$ moftransformer predict --root_dataset './data' --load_path 'path/of/model' --downstream 'exmaple' --mean 0. --std 1.
```

--------------------------------

### Train MOFTransformer

Source: https://github.com/hspark1212/moftransformer/blob/master/docs/source/tutorial.md

Configuration and execution of the training process using the downloaded hMOF database.

```python
import moftransformer

root_dataset = './hmof/downstream_release'
downstream = 'raspa_100bar'
log_dir = './logs'
max_epochs = 20
mean = 487.841
std = 63.088
batch_size = 32
load_path = "pmtransformer" # default : "pmtransformer", you can also choose "moftransformer" or None or path of other training models.

moftransformer.run(root_dataset, downstream, max_epochs=max_epochs, mean=mean, std=std, 
                   batch_size=batch_size, log_dir=log_dir, load_path="pmtransformer")
```

--------------------------------

### Visualize Energy-grid Embeddings

Source: https://github.com/hspark1212/moftransformer/blob/master/docs/source/getting_started/visualization.md

Methods for rendering energy-grid embeddings, supporting filtering by attention scores, specific patch positions, and ranking.

```python
%matplotlib widget
vis.draw_grid()
```

```python
%matplotlib widget
minatt, maxatt = 0.005, 0.008
view_init=(0, -55)
cmap = 'rocket_r'
alpha = 0.3
grid_scale_factor = 1.8
atomic_scale_factor = 2

vis.draw_grid(minatt=minatt, maxatt=maxatt, view_init=view_init, alpha=alpha, cmap=cmap,
              grid_scale_factor=grid_scale_factor, 
              atomic_scale_factor=atomic_scale_factor, 
              remove_under_minatt=True)
```

```python
%matplotlib widget
view_init=(27, -92)
alpha = 1
grid_scale_factor = 1.8
atomic_scale_factor = 2
patch_list = [[0,0,0], [1,1,1], [2,2,2], [3,3,3], [4,4,4], [5,5,5]] # make list to patch 

vis.draw_grid(view_init=view_init, alpha=alpha,
              grid_scale_factor=grid_scale_factor, 
              patch_list=patch_list,
              atomic_scale_factor=atomic_scale_factor)
```

```python
%matplotlib widget
view_init = (0, 125)
grid_scale_factor=1.8
atomic_scale_factor=1.5
rank = range(20)

vis.draw_grid_with_attention_rank(rank, view_init=view_init, 
                                  grid_scale_factor=grid_scale_factor,
                                  atomic_scale_factor=atomic_scale_factor)
```

--------------------------------

### Test MOFTransformer using Python

Source: https://github.com/hspark1212/moftransformer/blob/master/docs/source/getting_started/test.md

Evaluates a model using the test function. Results are saved to the specified directory or the model's default location.

```python
from pathlib import Path
import moftransformer
from moftransformer.examples import example_path

root_dataset = example_path['root_dataset']
downstream = example_path['downstream']

# Get ckpt file
seed = 0               # default seeds
version = 0            # version for model. It increases with the number of trains

# For version > 2.1.1, best.ckpt exists
checkpoint = 'best'    # Epochs where the model is stored. 
save_dir = 'result/'

# optional keyword
mean = 0
std = 1

load_path = Path(log_dir) / f'pretrained_mof_seed{seed}_from_pmtransformer/version_{version}/checkpoints/{checkpoint}.ckpt'

if not load_path.exists():
    raise ValueError(f'load_path does not exists. check path for .ckpt file : {load_path}')

moftransformer.test(root_dataset, load_path, downstream=downstream,
                   save_dir=save_dir, mean=mean, std=std)
```

--------------------------------

### Predict Material Properties with MOFTransformer

Source: https://github.com/hspark1212/moftransformer/blob/master/README.md

Use this snippet to predict material properties using a fine-tuned MOFTransformer model. Ensure the model checkpoint exists at the specified load_path.

```python
from pathlib import Path
import moftransformer
from moftransformer.examples import example_path

root_dataset = example_path['root_dataset']
downstream = example_path['downstream']

# Get ckpt file
log_dir = './logs/'    # same directory make from training
seed = 0               # default seeds
version = 0            # version for model. It increases with the number of trains
checkpoint = 'best'    # Epochs where the model is stored. 
mean = 0
std = 1

load_path = Path(log_dir) / f'pretrained_mof_seed{seed}_from_pmtransformer/version_{version}/checkpoints/{checkpoint}.ckpt'

if not load_path.exists():
    raise ValueError(f'load_path does not exists. check path for .ckpt file : {load_path}')
    
moftransformer.predict(
    root_dataset, load_path=load_path, downstream=downstream, split='all', mean=mean, std=std
)
```

--------------------------------

### Manage CGCNN Environment

Source: https://github.com/hspark1212/moftransformer/blob/master/baseline_model/CGCNN/README.md

Commands to activate, test, and deactivate the CGCNN conda environment.

```bash
source activate cgcnn
```

```bash
python main.py -h
python predict.py -h
```

```bash
source deactivate
```

--------------------------------

### Visualize Specific Patches

Source: https://github.com/hspark1212/moftransformer/blob/master/docs/source/getting_started/visualization.md

Methods to visualize individual patches based on position or attention rank.

```python
%matplotlib widget
patch_position = [2, 2, 2]
view_init = (0, -55)

vis.draw_specific_patch(patch_position, alpha=0.5, view_init=view_init)
```

```python
view_init = (15, -58)

vis.draw_specific_patch_with_attention_rank(1, alpha=0.5, ep=0.8, 
                                            view_init = view_init,
                                           grid_scale_factor=2)
```

--------------------------------

### Command-line Prediction

Source: https://github.com/hspark1212/moftransformer/blob/master/docs/source/getting_started/predict.md

Run predictions directly from the terminal using the moftransformer CLI.

```APIDOC
## CLI: moftransformer predict

### Description
Perform model prediction via the command line using the same parameters as the Python function.

### Usage
```bash
$ moftransformer predict --root_dataset [root_dataset] --load_path [load_path] --downstream [downstream] --split [split] --save_dir [save_dir]
```

### Example
```bash
$ moftransformer predict --root_dataset './data' --load_path 'path/of/model' --downstream 'example' --mean 0. --std 1.
```
```

--------------------------------

### Prepare Data for Single Task

Source: https://github.com/hspark1212/moftransformer/blob/master/docs/source/dataset.md

Use this function to generate atom-wise graph embeddings and energy-grid embeddings for a single downstream task. Ensure CIF files and a corresponding `raw_{downstream}.json` file are present in the `root_cifs` directory.

```python
from moftransformer.utils import prepare_data

# single task
prepare_data(root_cifs, root_dataset, downstream="example")
```

--------------------------------

### Prepare Data for Multiple Tasks

Source: https://github.com/hspark1212/moftransformer/blob/master/docs/source/dataset.md

This function call is used to generate embeddings for multiple downstream tasks simultaneously. Provide a list of task names to the `downstream` argument.

```python
prepare_data(root_cifs, root_dataset, downstream=["example1", "example2", ...])
```

--------------------------------

### Visualize Feature Importance with PatchVisualizer (Graph)

Source: https://github.com/hspark1212/moftransformer/blob/master/README.md

Visualize the feature importance analysis of atom-based graph embeddings for a fine-tuned model. Requires a trained model checkpoint and dataset.

```python
from moftransformer.visualize import PatchVisualizer
from moftransformer.examples import visualize_example_path

model_path = "examples/finetuned_bandgap.ckpt" # or 'examples/finetuned_h2_uptake.ckpt'
data_path = visualize_example_path
cifname = 'MIBQAR01_FSR'

vis = PatchVisualizer.from_cifname(cifname, model_path, data_path)
vis.draw_graph()
```

--------------------------------

### Test Fine-tuned MOFTransformer Model

Source: https://github.com/hspark1212/moftransformer/blob/master/README.md

Test the fine-tuned MOFTransformer model. Specify the dataset root, downstream task, and checkpoint details to load the model for evaluation.

```python
from pathlib import Path
import moftransformer
from moftransformer.examples import example_path

root_dataset = example_path['root_dataset']
downstream = example_path['downstream']

# Get ckpt file
seed = 0               # default seeds
version = 0            # version for model. It increases with the number of trains

# For version > 2.1.1, best.ckpt exists
checkpoint = 'best'    # Epochs where the model is stored. 
save_dir = 'result/'
```

--------------------------------

### Define Custom Dataset Structure

Source: https://github.com/hspark1212/moftransformer/blob/master/baseline_model/CGCNN/README.md

The required directory structure for inputting custom crystal data into the CGCNN model.

```text
root_dir
├── id_prop.csv
├── atom_init.json
├── id0.cif
├── id1.cif
├── ...
```

--------------------------------

### Cite CGCNN in BibTeX

Source: https://github.com/hspark1212/moftransformer/blob/master/baseline_model/CGCNN/README.md

Use this BibTeX entry to cite the CGCNN framework in academic publications.

```bibtex
@article{PhysRevLett.120.145301,
  title = {Crystal Graph Convolutional Neural Networks for an Accurate and Interpretable Prediction of Material Properties},
  author = {Xie, Tian and Grossman, Jeffrey C.},
  journal = {Phys. Rev. Lett.},
  volume = {120},
  issue = {14},
  pages = {145301},
  numpages = {6},
  year = {2018},
  month = {Apr},
  publisher = {American Physical Society},
  doi = {10.1103/PhysRevLett.120.145301},
  url = {https://link.aps.org/doi/10.1103/PhysRevLett.120.145301}
}
```

--------------------------------

### Predicting with Python API

Source: https://github.com/hspark1212/moftransformer/blob/master/docs/source/getting_started/predict.md

Use the moftransformer.predict function to run inference on a dataset using a pre-trained checkpoint.

```python
from pathlib import Path
import moftransformer
from moftransformer.examples import example_path

root_dataset = example_path['root_dataset']
downstream = example_path['downstream']

# Get ckpt file
seed = 0               # default seeds
version = 0            # version for model. It increases with the number of trains
checkpoint = 'best'    # Epochs where the model is stored. 
mean = 0
std = 1

load_path = Path(log_dir) / f'pretrained_mof_seed{seed}_from_pmtransformer/version_{version}/checkpoints/{checkpoint}.ckpt'

if not load_path.exists():
    raise ValueError(f'load_path does not exists. check path for .ckpt file : {load_path}')
    
moftransformer.predict(
    root_dataset, load_path=load_path, downstream=downstream, split='all', mean=mean, std=std
)
```

--------------------------------

### Visualize Atom-based Graph Embeddings

Source: https://github.com/hspark1212/moftransformer/blob/master/docs/source/getting_started/visualization.md

Methods for rendering graph embeddings with attention scores, including parameter customization.

```python
%matplotlib widget
vis.draw_graph()
```

```python
%matplotlib widget
minatt, maxatt = 0.002, 0.006
view_init=(0, -55)
cmap = 'rocket_r'
alpha = 0.5
grid_scale_factor = 1.8
atomic_scale_factor = 1.5
att_scale_factor= 3
vis.draw_graph(minatt=minatt, maxatt=maxatt, view_init=view_init, alpha=alpha, cmap=cmap,
               grid_scale_factor=grid_scale_factor, 
               atomic_scale_factor=atomic_scale_factor,
               att_scale_factor=att_scale_factor)
```

--------------------------------

### moftransformer.predict

Source: https://github.com/hspark1212/moftransformer/blob/master/docs/source/getting_started/predict.md

The predict function generates model outputs for a given dataset and saves the results as a CSV file.

```APIDOC
## moftransformer.predict

### Description
Executes model prediction on a specified dataset and saves the results to a CSV file.

### Parameters
- **root_dataset** (str) - Required - A folder containing graph data, grid data, and JSON of MOFs.
- **load_path** (str) - Required - Path to the model checkpoint file (*.ckpt).
- **downstream** (str) - Optional - Name of the user-specific task (e.g., bandgap, gasuptake).
- **split** (str) - Optional - The dataset split to predict ('all', 'train', 'test', or 'val').
- **save_dir** (str) - Optional - Directory path to save the resulting CSV file.
- **kwargs** (dict) - Optional - Additional configuration for MOFTransformer.

### Request Example
```python
moftransformer.predict(
    root_dataset='./data',
    load_path='path/to/model.ckpt',
    downstream='bandgap',
    split='all'
)
```
```

=== COMPLETE CONTENT === This response contains all available snippets from this library. No additional content exists. Do not make further requests.