|
--- |
|
license: llama2 |
|
library_name: transformers |
|
pipeline_tag: text-generation |
|
tags: |
|
- llama |
|
--- |
|
Link to GGUF version: [GGUF](https://huggingface.co/androlike/TerraMix_L2_13B_16K_GGUF) |
|
|
|
Thanks to everyone, who finenuted base Llama2 model, made (Q)LoRas, created scripts for merge: [ties-merge](https://github.com/cg123/ties-merge), [BlockMerge_Gradient](https://github.com/Gryphe/BlockMerge_Gradient), [zaraki-tools](https://github.com/zarakiquemparte/zaraki-tools) |
|
#### Model details: |
|
Experiment about merging model with large context length. |
|
Use these rope scaling settings: |
|
``` |
|
--rope-freq-base 10000 --rope-freq-scale 0.25 -c 16384 (llama.cpp) |
|
``` |
|
``` |
|
--ropeconfig 0.25 10000 --contextsize 16384 (koboldcpp) |
|
``` |
|
You can use various instruct formats: |
|
### Alpaca instruct format (Recommended): |
|
``` |
|
### Instruction: |
|
(your instruct prompt is here) |
|
### Response: |
|
``` |
|
### Vicuna 1.1 instruct format: |
|
``` |
|
You are a helpful AI assistant. |
|
USER: <prompt> |
|
ASSISTANT: |
|
``` |
|
### Metharme instruct format: |
|
``` |
|
<|system|> (your instruct prompt) |
|
<|user|> (user's reply)<|model|> (for model's output) |
|
``` |
|
|
|
#### Models used for the merge: |
|
|
|
### Part1: |
|
[Airoboros L2 13B 2.1](https://huggingface.co/jondurbin/airoboros-l2-13b-2.1) + [LLAMA2 13B - Holodeck](https://huggingface.co/KoboldAI/LLAMA2-13B-Holodeck-1) merged with Creative and Reasoning [Airoboros LMoE 13B 2.1](https://huggingface.co/jondurbin/airoboros-lmoe-13b-2.1) |
|
### Part2: |
|
[Chronos 13B V2](https://huggingface.co/elinas/chronos-13b-v2) merged with [Kimiko-v2-13B](https://huggingface.co/nRuaif/Kimiko-v2-13B) + [Nous-Hermes-Llama2-13b](https://huggingface.co/NousResearch/Nous-Hermes-Llama2-13b) merged with [limarp-llama2](https://huggingface.co/lemonilia/limarp-llama2) and [limarp-llama2-v2](https://huggingface.co/lemonilia/limarp-llama2-v2) + [Synthia-13B](https://huggingface.co/migtissera/Synthia-13B) merged with [BluemoonRP-L2-13B](https://huggingface.co/nRuaif/BluemoonRP-L2-13B-This-time-will-be-better) and [LLama-2-13b-chat-erp-lora-mk2](https://huggingface.co/PocketDoc/DansHalfbakedAdapters) + [WizardLM-1.0-Uncensored-Llama2-13b](https://huggingface.co/ehartford/WizardLM-1.0-Uncensored-Llama2-13b) merged with [Llama-2-13B-Storywriter-LORA](https://huggingface.co/Blackroot/Llama-2-13B-Storywriter-LORA) |
|
### Part3: |
|
[Speechless Llama2 13B](https://huggingface.co/uukuguy/speechless-llama2-13b) + [Redmond Puffin 13B](https://huggingface.co/NousResearch/Redmond-Puffin-13B) |
|
### Part4: |
|
[Tsukasa 13B 16K](https://huggingface.co/ludis/tsukasa-13b-16k) (repo is deleted) + [EverythingLM-13b-V2-16k](https://huggingface.co/totally-not-an-llm/EverythingLM-13b-V2-16k) |
|
### Part5: |
|
TerraMix_L2_13B (base) was merged with [PIPPA ShareGPT Subset QLoRa 13B](https://huggingface.co/zarakiquemparte/PIPPA-ShareGPT-Subset-QLora-13b) |
|
### Part6: |
|
Three parts merged in one, then TsuryLM-L2-16K was merged with TerraMix_L2_13B (base). |
|
|
|
Model is intended for creativity purposes (roleplay). It can regularly break formatting or sometimes have poor understanding about small details in occuring situations. |
|
|
|
But yet, this model is almost absent from alignment, can generate direct output, moderately in prose, good in internet RP style. |
|
|
|
### Limitations and risks |
|
Llama2 and its derivatives (finetunes) is licensed under LLama 2 Community License, various finetunes or (Q)LoRAs has appropriate licenses depending on used datasets in finetuning or training (Quantized) Low-Rank Adaptations. This mix can generate heavily biased output, which aren't suitable for minors or common audience. |
