shibing624/llama-3-8b-instruct-262k-chinese
TEXT GENERATIONConcurrency Cost:1Model Size:8BQuant:FP8Ctx Length:8kPublished:Apr 27, 2024License:llama3Architecture:Transformer0.0K Warm

The shibing624/llama-3-8b-instruct-262k-chinese model is an 8 billion parameter instruction-tuned Llama-3 variant, fine-tuned by shibing624 using the ORPO method on a Chinese-English preference dataset. It features an extended context length of 262k tokens, making it suitable for long-context applications like RAG, and supports both Chinese and English for multi-turn conversations, code encoding, and reasoning tasks.

Loading preview...