StellarX 4B V0 is an open-source language model by Dampish. Features: 4b LLM, VRAM: 8.1GB, Context: 2K, License: cc-by-nc-sa-4.0, LLM Explorer Score: 0.11, Arc: 37, HellaSwag: 61.9, MMLU: 26.9.
Not designed for direct deployment without supervision, Potential biases and limitations inherent in the model
Considerations:
Users should conduct risk and bias assessments before using it for downstream tasks.
Additional Notes
StellarX is optimized for carbon efficiency and can run on local devices. It is primarily intended for research and personal projects under the CC-BY-NC-SA-4.0 license.
Supported Languages
English (Proficient)
Training Details
Data Sources:
redpajama dataset by togethercumputer
Data Volume:
810 billion tokens
Model Architecture:
GPT-NeoX, inspired by GPT-3 and GPT-J-6B
Input Output
Input Format:
Text
Accepted Modalities:
Text
Output Format:
Text
Release Notes
Version:
V0
Notes:
Initial version trained on 300B tokens, with a goal of 810B tokens.