anakin87/zephyr-7b-alpha-sharded
TEXT GENERATIONConcurrency Cost:1Model Size:7BQuant:FP8Ctx Length:4kPublished:Oct 14, 2023License:mitArchitecture:Transformer0.0K Open Weights Cold

anakin87/zephyr-7b-alpha-sharded is a 7 billion parameter language model, a sharded version of HuggingFaceH4's Zephyr-7B-Alpha, which is fine-tuned from Mistral-7B-v0.1. This model is specifically designed to be easily loaded and run on platforms like Google Colab, particularly with 8-bit quantization. It functions as a helpful assistant, trained using Direct Preference Optimization (DPO) on synthetic datasets, and is intended for educational and research purposes.

Loading preview...