Dataset · anon8231489123
ShareGPT dialogues for chat model training
anon8231489123/ShareGPT_Vicuna_unfiltered
English ShareGPT dialogues cleaned for SFT of chat models like Vicuna.
What's inside
This is a dump of ShareGPT dialogues prepared for training Vicuna. The card mentions variants with and without the removal of 'I'm sorry, but' phrases, as well as a cleaned split. The language is English; data is presented in JSON and intended for instructional fine-tuning.
What you can do with it
- SFT of chat models on dialogues
- LoRA fine-tuning for instructions
- Evaluating the quality of dialogue cleaning
- Languages
- English
- Downloads this month
- 326.5K
- Likes
- 950
Models trained on it
- rombodawg/rombos_Replete-Coder-Llama3-8BText generation 9.5K
- abacusai/Llama-3-Smaug-8BText generation 9.4K
- double7/vicuna-68mText generation 3.2K
- mradermacher/rombos_Llama3-8B-Instruct-Replete-Adapted-i1-GGUF 1.8K
- TheBloke/manticore-13b-chat-pyg-GGUFText generation 1.8K
- mradermacher/rombos_Replete-Coder-Qwen2-1.5b-i1-GGUF 1.6K
- LiteLLMs/Llama-3-Smaug-8B-GGUF 1.4K
- mradermacher/Llama3-8B-Instruct-Replete-Adapted-i1-GGUF 1.2K
How to load it
from datasets import load_dataset
ds = load_dataset("anon8231489123/ShareGPT_Vicuna_unfiltered")