---
library_name: matgl
tags:
- matgl
- materials-science
- graph-neural-network
---
# M3GNet-Eform-MP-2019.4.1

## Introduction

Pre-trained formation energy model based on M3GNet. 
Retrained due to changes in M3GNet implementation in MatGL.

## Usage

```python
import matgl

model = matgl.load_model("materialyze/M3GNet-Eform-MP-2019.4.1")
```

## Model Details

- Number of parameters: 394,000

## Metrics

| Split | MAE (eV/atom) |
|---|---:|
| Train | 0.006 |
| Validation | 0.022 |

## Metadata

```json
{
  "dataset": "MP-EForm-2019.4.1",
}
```