| --- |
| license: mit |
| library_name: tait |
| tags: |
| - tait |
| - numpy |
| - on-device |
| - text-generation |
| --- |
| |
| # tinybili |
|
|
| A small MiniGPT model trained with TAIT, a NumPy-only toolkit for training AI on phones and resource-constrained devices. |
|
|
| ## Model details |
|
|
| - **Format:** TAIT `.npz` (NumPy weights + BPE tokenizer, readable by the KAIT runner) |
| - **File:** `tinyGPT0.1.npz` (1072206 bytes, sha256 `1a2b1c51ba9b4f01...`) |
| - **Vocab size:** 1003 |
| - **d_model:** 64 | **d_ff:** 128 | **block_size:** 64 |
| - **TAIT version:** 2.3.0 |
| - **Training data:** mixed-1.jsonl |
| - **Training examples:** 6666 |
| - **Epochs:** 50 |
| - **Final training loss:** 0.4900 |
| - **Validation loss:** 1.0367 |
| |
| ## How to use it |
| |
| ```bash |
| pip install "tait[publish]" # publish support; training itself needs only NumPy |
| # Download the model, then chat with it locally: |
| tait chat --model tinyGPT0.1.npz |
| ``` |
| |
| You can also load the same `.npz` file in KAIT, the Android |
| runner for TAIT models, by pasting this repository link there. |
| |
| ## Limitations |
| |
| This is a tiny model trained from scratch on a small dataset. It may hallucinate, repeat itself, or answer off-topic. Treat its output as an experiment, not as a factual source. |
| |
| ## Source |
| |
| Trained with TAIT: https://github.com/muzammilsoft/tait |
| |
