metadata
license: llama2
datasets:
- totally-not-an-llm/EverythingLM-data-V3
EverythingLM-13b-V3-16k
Introducing EverythingLM, a llama-2 based, general-purpose 13b model with 16k context thanks to LlongMa. The model is trained on the EverythingLM-V3 dataset, more info can be found on the dataset page.
The model is completely uncensored.
Despite being "uncensored", the base model might be resistant; you might have to prompt-engineer certain prompts.
Quants (Thanks TheBloke!):
https://huggingface.co/TheBloke/EverythingLM-13B-V3-16K-GGUF
https://huggingface.co/TheBloke/EverythingLM-13B-V3-16K-GPTQ
https://huggingface.co/TheBloke/EverythingLM-13B-V3-16K-AWQ
Notable features:
- Automatically triggered CoT reasoning.
- Verbose and detailed replies.
- Creative stories.
- Good prompt understanding.
Differences from V2:
- Much more uncensored.
- Actual roleplaying ability now!
- General all around improvements thanks to the new dataset. Check out the dataset for more info.
Prompt format (Alpaca-chat):
USER: <prompt>
ASSISTANT:
Future plans:
- Highest priority right now is V3.1 with more optimized training and iterative dataset improvements based on testing.
Note:
Through testing V2, I realized some alignment data had leaked in, causing the model to be less cooperative then intended. This model should do much better due to stricter filetering.
Open LLM Leaderboard Evaluation Results
Detailed results can be found here
| Metric | Value |
|---|---|
| Avg. | 44.82 |
| ARC (25-shot) | 58.19 |
| HellaSwag (10-shot) | 80.12 |
| MMLU (5-shot) | 50.48 |
| TruthfulQA (0-shot) | 45.18 |
| Winogrande (5-shot) | 70.72 |
| GSM8K (5-shot) | 1.97 |
| DROP (3-shot) | 7.06 |