---
license: cc-by-nc-4.0
language:
- en
- zh
- ja
- de
datasets:
- JosephusCheung/GuanacoDataset
- meta-math/MetaMathQA
- jondurbin/airoboros-3.1
- WizardLM/WizardLM_evol_instruct_V2_196k
- RyokoAI/ShareGPT52K
- RyokoAI/Fandom23K
- milashkaarshif/MoeGirlPedia_wikitext_raw_archive
- wikipedia
- wiki_lingua
- garage-bAInd/Open-Platypus
- LDJnr/Puffin
- BAAI/COIG
- TigerResearch/tigerbot-zhihu-zh-10k
- liwu/MNBVC
- teknium/openhermes
- CausalLM/Refined-Anime-Text
- microsoft/orca-math-word-problems-200k
- m-a-p/CodeFeedback-Filtered-Instruction
---
# Notes
- 3.75bpw test quant of CausalLM/35b-beta-long, which is in itself a finetune of CohereForAI/c4ai-command-r-v01 (hence the corrected licensing).
- Theoretically should fit within 24GB of VRAM for inference.

## TBA

Tokenizer is different from cohere - and chat template is ChatML - fully fine-tuned at 128K+

No loras, no quants, no tricks, 30M+ sft data.

Pressure Testing from: https://github.com/LeonEricsson/llmcontext

![image/png](https://cdn-uploads.huggingface.co/production/uploads/63468a143ea42ee2cb49ddd1/2XbONpyTeMH1qWCtE9ziH.png)