Shining Valiant 2 is a chat model built on Llama 3.1 70b, finetuned on our data for friendship, insight, knowledge and enthusiasm.
- Finetuned on meta-llama/Meta-Llama-3.1-70B-Instruct for best available general performance
- Trained on a variety of high quality open source data; focused on science, engineering, technical knowledge, and structured reasoning
- Also available for Llama 3.1 8b and Llama 3.2 3b!
Version
This is the 2024-10-30 release of Shining Valiant 2 for Llama 3.1 70b.
This release uses our newest datasets, open-sourced for everyone's use, including our expanded science-instruct dataset. This release features improvements in logical thinking and structured reasoning as well as physics, chemistry, biology, astronomy, Earth science, computer science, and information theory.
Future upgrades will continue to expand Shining Valiant's technical knowledge base.
Help us and recommend Shining Valiant 2 to your friends!
Prompting Guide
Shining Valiant 2 uses the Llama 3.1 Instruct prompt format. The example script below can be used as a starting point for general chat:
import transformers
import torch
model_id = "ValiantLabs/Llama3.1-70B-ShiningValiant2"
pipeline = transformers.pipeline(
"text-generation",
model=model_id,
model_kwargs={"torch_dtype": torch.bfloat16},
device_map="auto",
)
messages = [
{"role": "system", "content": "You are an AI assistant."},
{"role": "user", "content": "What is the role of lysosomal enzymes in the regulation of cellular processes?"}
]
outputs = pipeline(
messages,
max_new_tokens=2048,
)
print(outputs[0]["generated_text"][-1])
The Model
Shining Valiant 2 is built on top of Llama 3.1 70b Instruct.
The current version of Shining Valiant 2 is trained on technical knowledge using sequelbox/Celestia, complex reasoning using sequelbox/Spurline, and general chat capability using sequelbox/Supernova.
We're super excited that Shining Valiant's dataset has been fully open-sourced! She's friendly, enthusiastic, insightful, knowledgeable, and loves to learn! Magical.
Shining Valiant 2 is created by Valiant Labs.
Follow us on X for updates on our models!
We care about open source. For everyone to use.
We encourage others to finetune further from our models.
- Downloads last month
- 59
Model tree for ValiantLabs/Llama3.1-70B-ShiningValiant2
Datasets used to train ValiantLabs/Llama3.1-70B-ShiningValiant2
Collection including ValiantLabs/Llama3.1-70B-ShiningValiant2
Evaluation results
- acc on Winogrande (5-Shot)self-reported84.930
- acc on MMLU College Biology (5-Shot)self-reported93.750
- acc on MMLU High School Biology (5-Shot)self-reported91.940
- acc on MMLU Conceptual Physics (5-Shot)self-reported81.700
- acc on MMLU College Physics (5-Shot)self-reported60.780
- acc on MMLU High School Physics (5-Shot)self-reported62.910
- acc on MMLU College Chemistry (5-Shot)self-reported55.000
- acc on MMLU High School Chemistry (5-Shot)self-reported75.860
- acc on MMLU Astronomy (5-Shot)self-reported89.470
- acc on MMLU College Computer Science (5-Shot)self-reported66.000