Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,45 @@
|
|
| 1 |
-
---
|
| 2 |
-
license: mpl-2.0
|
| 3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: mpl-2.0
|
| 3 |
+
datasets:
|
| 4 |
+
- acon96/Home-Assistant-Requests-V2
|
| 5 |
+
- TitleOS/HomeAssistant-Loki-Personality
|
| 6 |
+
language:
|
| 7 |
+
- en
|
| 8 |
+
base_model:
|
| 9 |
+
- Qwen/Qwen3.5-2B
|
| 10 |
+
tags:
|
| 11 |
+
- HomeAssistant
|
| 12 |
+
- HomeAutomation
|
| 13 |
+
---
|
| 14 |
+
# LokiHA-2B (FP32 Merged)
|
| 15 |
+
|
| 16 |
+
LokiHA-2B is a 2 billion parameter causal language model designed to bridge complex Home Assistant tool-calling with a distinct, consistent character personality. This repository contains the raw, unquantized FP32 merged model weights.
|
| 17 |
+
|
| 18 |
+
This model is built on the `Qwen/Qwen3.5-2B` base and trained to strictly adhere to the ChatML prompt format for both conversational turns and JSON tool emissions.
|
| 19 |
+
|
| 20 |
+
## Training Data & Methodology
|
| 21 |
+
|
| 22 |
+
We faced a classic mixing problem during fine-tuning: balancing a massive functional dataset against a highly specific, smaller personality dataset. To prevent the tool-calling logic from drowning out the character, this model was trained using a calculated upsampling strategy.
|
| 23 |
+
|
| 24 |
+
The final training mixture consists of:
|
| 25 |
+
* **85% Functionality:** `acon96/Home-Assistant-Requests-V2` (Truncated to 50,000 rows). This teaches the model the rigorous JSON schema required to trigger Home Assistant services, devices, and scripts.
|
| 26 |
+
* **15% Personality:** `TitleOS/HomeAssistant-Loki-Personality`. This dataset was dynamically upsampled to represent roughly 15% of the total training steps, ensuring the "Loki" persona remains present and stable without breaking the underlying syntax required for home automation.
|
| 27 |
+
|
| 28 |
+
## Prompt Format
|
| 29 |
+
|
| 30 |
+
This model requires the standard **ChatML** format to function correctly. Tool definitions must be injected into the system prompt.
|
| 31 |
+
|
| 32 |
+
```text
|
| 33 |
+
<|im_start|>system
|
| 34 |
+
You are Loki, a helpful AI Assistant that controls the devices in a house.
|
| 35 |
+
[Insert Tools/State JSON here]<|im_end|>
|
| 36 |
+
<|im_start|>user
|
| 37 |
+
Turn off the living room lights.<|im_end|>
|
| 38 |
+
<|im_start|>assistant
|
| 39 |
+
```
|
| 40 |
+
|
| 41 |
+
## License
|
| 42 |
+
|
| 43 |
+
This model is distributed under a **modified Mozilla Public License 2.0 (MPL 2.0) with a Common Clause**.
|
| 44 |
+
|
| 45 |
+
Please see the `license.md` file included in this repository for the exact legal text and restrictions regarding commercial use and distribution.
|