Encoder-Decoder
Encoder-Decoder
Overview
From scratch implementation of Encoder-Decoder
Technical Details
- Framework: PyTorch
- Dataset: Custom
- Category: Sequential Models
Implementation Details
Trained on the on German (de) to English (en) dataset
Sequence to Sequence Learning with Neural Networks
ModelArgs Hyperparameters
| Parameter | Value | Description |
|---|---|---|
batch_size |
32 | The number of samples processed before the model is updated. |
max_lr |
1e-4 | Maximum learning rate. |
dropout |
0.2 | Dropout. |
epochs |
10 | Epochs |
block_size |
32 | Seq Len |
num_layers |
4 | Layers for deep lstms |
No of neurons |
128 | No of neurons in an GRU per layer |
Frameworks:
Pytorch
Epochs/Steps
Epochs (train) = 10
Val iterations = every epoch
Losses
Train loss - 1.38
Val loss - 1.39
Loss Curves
Source Code
📁 GitHub Repository: Encoder-Decoder
View the complete implementation, training scripts, and documentation on GitHub.