# Wildnerve-tlm01 Model Checkpoints

This repository contains the trained weights for the Wildnerve-tlm01 hybrid neural-symbolic language model architecture, which combines transformer-based language processing with biologically-inspired Spike-Timing-Dependent Plasticity (STDP).

## Repository Structure

The weights are organized into two main components:

- **`snn/`**: Contains the STDP-trained synaptic weights that enable neuromorphic learning
  - `stdp_model_epoch_30.bin`: Synaptic weights after 30 epochs of STDP training

- **`transformer/`**: Contains the transformer model weights for language processing
  - `model_weights.bin`: Transformer component weights

## Usage

These weights are designed to be used with the Wildnerve-tlm01 model implementation located in the [EvolphTech/Model](https://huggingface.co/EvolphTech/Model) repository.

### Loading the Weights

```python
import torch
import os
from huggingface_hub import hf_hub_download

# Download the components
snn_path = hf_hub_download(
    repo_id="EvolphTech/Checkpoints",
    filename="snn/stdp_model_epoch_30.bin",
    cache_dir="model_cache"
)

transformer_path = hf_hub_download(
    repo_id="EvolphTech/Checkpoints",
    filename="transformer/model_weights.bin",
    cache_dir="model_cache"
)

# Load the weights
snn_weights = torch.load(snn_path, map_location="cpu")
transformer_weights = torch.load(transformer_path, map_location="cpu")
```

For complete integration instructions, see the [model documentation](https://huggingface.co/EvolphTech/Model#usage).

## Training Details

- **SNN Component**: Trained for 30 epochs using STDP learning rule with adaptive synaptic weights
- **Transformer Component**: Trained on multiple programming and general knowledge datasets

## License

Mozilla Public License 2.0 (MPL 2.0)