GRU
GRU
Overview
From scratch implementation of GRU
Technical Details
- Framework: PyTorch
- Dataset: Custom
- Category: Sequential Models
Implementation Details
Trained a GRU model coded from scratch in Pytorch
ModelArgs Hyperparameters
| Parameter | Value | Description |
|---|---|---|
batch_size |
16 | The number of samples processed before the model is updated. |
max_lr |
1e-4 | Maximum learning rate. |
dropout |
0.2 | Dropout. |
epochs |
50 | Epochs |
block_size |
16 | Seq Len |
No of neurons |
16 | No of neurons in an GRU per layer |
Frameworks:
Pytorch
Epochs/Steps
Epochs (train) = 50
Val iterations = every epoch
Losses
Train loss - 0.51
Val loss - 0.48
Loss Curves
Source Code
📁 GitHub Repository: GRU
View the complete implementation, training scripts, and documentation on GitHub.