Lumimaid V0.2 8B

🚀 Lumimaid 0.2

Lumimaid 0.2 is an enhanced model based on Meta-Llama-3.1-8B-Instruct, with significant improvements in the dataset compared to version 0.1.

🚀 Quick Start

Model Basis

This model is based on: Meta-Llama-3.1-8B-Instruct. Wandb link: https://wandb.ai/undis95/Lumi-Llama-3-1-8B?nw=nwuserundis95

Dataset Improvement

The upgrade from Lumimaid 0.1 to 0.2 represents a huge step forward in terms of the dataset. Some users had feedback that our previous models were sloppy. So, Ikari decided to clean up most of the messy chats. Our dataset has remained the core since the beginning. Over time, we've added data, cleaned it, and repeated the process. After a period of not releasing new models due to dissatisfaction, we believe it's time to make a comeback!

✨ Features

Prompt Template

The model uses the Llama - 3 - Instruct prompt template:

<|begin_of_text|><|start_header_id|>system<|end_header_id|>

{system_prompt}<|eot_id|><|start_header_id|>user<|end_header_id|>

{input}<|eot_id|><|start_header_id|>assistant<|end_header_id|>

{output}<|eot_id|>

📚 Documentation

Credits

• Undi
• IkariDev

Training Data

We used the following datasets to create our dataset:

• Epiculous/Gnosis
• ChaoticNeutrals/Luminous_Opus
• ChaoticNeutrals/Synthetic-Dark-RP
• ChaoticNeutrals/Synthetic-RP
• Gryphe/Sonnet3.5-SlimOrcaDedupCleaned
• Gryphe/Opus-WritingPrompts
• meseca/writing-opus-6k
• meseca/opus-instruct-9k
• PJMixers/grimulkan_theory-of-mind-ShareGPT
• NobodyExistsOnTheInternet/ToxicQAFinal
• Undi95/toxic-dpo-v0.1-sharegpt
• cgato/SlimOrcaDedupCleaned
• kalomaze/Opus_Instruct_25k
• Doctor-Shotgun/no-robots-sharegpt
• Norquinal/claude_multiround_chat_30k
• nothingiisreal/Claude-3-Opus-Instruct-15K
• All the Aesirs dataset, cleaned, unslopped
• All le luminae dataset, cleaned, unslopped
• Small part of Airoboros reduced

We're sad to say that we couldn't find the sources of the following datasets. If you recognize your dataset, please DM us!

• Opus_Instruct-v2-6.5K-Filtered-v2-sharegpt
• claude_sharegpt_trimmed
• CapybaraPure_Decontaminated-ShareGPT_reduced

Datasets Credits

• Epiculous
• ChaoticNeutrals
• Gryphe
• meseca
• PJMixers
• NobodyExistsOnTheInternet
• cgato
• kalomaze
• Doctor-Shotgun
• Norquinal
• nothingiisreal

📄 License

The model is licensed under cc - by - nc - 4.0.

Others

• Undi: If you want to support us, you can here.
• IkariDev: Visit my retro/neocities style website please kek