File size: 861 Bytes
016eee7
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
---
language: en
license: cc0-1.0
tags:
- mnist
- tiny-model
- early-stopping
---

# Tiny MNIST Classifier

- **Parameters**: 970 (<1000)
- **Test accuracy**: 92.35%
- **Epochs trained**: 45 (early stopping after 5 epochs without improvement)

This model was trained on RX 6600.

## Full results

| Metric                    | Value           |
|---------------------------|-----------------|
| Total parameters          | 970  |
| Best validation loss      | 0.2463 |
| Final test accuracy       | 92.35% |
| Early stopping patience   | 5      |
| Training epochs           | 45       |

## Model architecture

AvgPool(4x4) → Linear(49→16) → ReLU → Dropout(0.2) → Linear(16→10)

## How to use

```python
import torch
from train import TinyMNISTModel

model = TinyMNISTModel()
model.load_state_dict(torch.load("mnist_1k_best.pth"))
model.eval()
```