6.1 KiB
license, model-index, library_name
| license | model-index | library_name | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| llama3 |
|
transformers |
Llama-3.1-8B-ArliAI-RPMax-v1.1
=====================================
RPMax Series Overview
| 2B | 3.8B | 8B | 9B | 12B | 20B | 22B | 70B |
RPMax is a series of models that are trained on a diverse set of curated creative writing and RP datasets with a focus on variety and deduplication. This model is designed to be highly creative and non-repetitive by making sure no two entries in the dataset have repeated characters or situations, which makes sure the model does not latch on to a certain personality and be capable of understanding and acting appropriately to any characters or situations.
Early tests by users mentioned that these models does not feel like any other RP models, having a different style and generally doesn't feel in-bred.
You can access the model at https://arliai.com and ask questions at https://www.reddit.com/r/ArliAI/
We also have a models ranking page at https://www.arliai.com/models-ranking
Ask questions in our new Discord Server! https://discord.com/invite/t75KbPgwhk
Model Description
Llama-3.1-8B-ArliAI-RPMax-v1.1 is a variant of the Meta-Llama-3.1-8B model.
v1.1 is just a small fix to not train and save the embeddings layer and small changes to the dataset, since v1.0 had the lm_head unnecessarily trained on accident.
Training Details
- Sequence Length: 8192
- Training Duration: Approximately 1 day on 2x3090Ti
- Epochs: 1 epoch training for minimized repetition sickness
- LORA: 64-rank 128-alpha, resulting in ~2% trainable weights
- Learning Rate: 0.00001
- Gradient accumulation: Very low 32 for better learning.
Quantization
The model is available in quantized formats:
We recommend using full weights or GPTQ
- FP16: https://huggingface.co/ArliAI/Llama-3.1-8B-ArliAI-RPMax-v1.1
- GPTQ_Q4: https://huggingface.co/ArliAI/Llama-3.1-8B-ArliAI-RPMax-v1.1-GPTQ_Q4
- GPTQ_Q8: https://huggingface.co/ArliAI/Llama-3.1-8B-ArliAI-RPMax-v1.1-GPTQ_Q8
- GGUF: https://huggingface.co/ArliAI/Llama-3.1-8B-ArliAI-RPMax-v1.1-GGUF
Suggested Prompt Format
Llama 3 Instruct Format
Example:
<|begin_of_text|><|start_header_id|>system<|end_header_id|>
You are [character]. You have a personality of [personality description]. [Describe scenario]<|eot_id|><|start_header_id|>user<|end_header_id|>
{{ user_message_1 }}<|eot_id|><|start_header_id|>assistant<|end_header_id|>
{{ model_answer_1 }}<|eot_id|><|start_header_id|>user<|end_header_id|>
{{ user_message_2 }}<|eot_id|><|start_header_id|>assistant<|end_header_id|>
Open LLM Leaderboard Evaluation Results
Detailed results can be found here
| Metric | Value |
|---|---|
| Avg. | 23.64 |
| IFEval (0-Shot) | 63.59 |
| BBH (3-Shot) | 28.79 |
| MATH Lvl 5 (4-Shot) | 11.33 |
| GPQA (0-shot) | 4.47 |
| MuSR (0-shot) | 5.31 |
| MMLU-PRO (5-shot) | 28.35 |