|
|
--- |
|
|
license: apache-2.0 |
|
|
datasets: |
|
|
- norygano/TRACHI |
|
|
language: |
|
|
- en |
|
|
--- |
|
|
|
|
|
# dolphin-mistral-TRACHI-7b |
|
|
|
|
|
<!-- Provide a quick summary of what the model is/does. --> |
|
|
|
|
|
Quantized versions of the fine-tuned [ |
|
|
dolphin-2.2.1-mistral-7b ](https://huggingface.co/cognitivecomputations/dolphin-2.2.1-mistral-7b) on TRACHI datasets. |
|
|
|
|
|
## Model Details |
|
|
This model aims to facilitate conversations with individual TRACHI characters. Prototyping [Discord Bots](https://discord.gg/mcRke7vB7X) running on Ollama. |
|
|
|
|
|
### Model Description |
|
|
Characters are prompted by including their name in the system role within the prompt: |
|
|
|
|
|
```plaintext |
|
|
<|im_start|>system |
|
|
{character}<|im_end|> |
|
|
<|im_start|>user |
|
|
{prompt}<|im_end|> |
|
|
<|im_start|>assistant |
|
|
``` |
|
|
|
|
|
## Bias, Risks, and Limitations |
|
|
|
|
|
<!-- This section is meant to convey both technical and sociotechnical limitations. --> |
|
|
Like the base model ([ |
|
|
dolphin-2.2.1-mistral-7b ](https://huggingface.co/cognitivecomputations/dolphin-2.2.1-mistral-7b)), this fine-tune is unbiased and uncensored. |
