zjunlp/OceanGPT-basic-14B-v0.1

TEXT GENERATIONConcurrent Unit Cost:1Model Size:14.2BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Apr 25, 2024License:mitArchitecture:Transformer0.0K Open Weights Featherless Exclusive Cold

OceanGPT-basic-14B-v0.1 by zjunlp is a 14.2 billion parameter causal language model based on Qwen1.5-14B, specifically fine-tuned for ocean science tasks. It was trained on a bilingual dataset (Chinese and English) focused on the ocean domain, making it specialized for scientific inquiries in this field. With a context length of 32768 tokens, this model is designed to assist with ocean-related research and information retrieval.

Loading preview...

OceanGPT-basic-14B-v0.1: Specialized for Ocean Science

OceanGPT-basic-14B-v0.1 is a 14.2 billion parameter large language model developed by zjunlp, built upon the Qwen1.5-14B architecture. This model is uniquely specialized for tasks within the ocean science domain, distinguishing it from general-purpose LLMs.

Key Capabilities & Features

  • Domain-Specific Training: Fine-tuned on a comprehensive bilingual dataset (Chinese and English) specifically curated for ocean-related knowledge.
  • Ocean Science Focus: Designed to understand and generate content relevant to various ocean science tasks, making it a valuable tool for researchers and professionals in this field.
  • Base Model: Leverages the robust capabilities of the Qwen1.5-14B model as its foundation.
  • Academic Project: Developed as an academic exploration, with ongoing research into its capabilities and limitations.

Good For

  • Oceanographic Research: Answering questions, summarizing information, or generating text related to marine biology, oceanography, marine geology, and other ocean sciences.
  • Bilingual Ocean Content: Handling queries and generating responses in both Chinese and English within the ocean domain.
  • Specialized Information Retrieval: Assisting users in navigating complex scientific data and concepts specific to the ocean.

Limitations

As an academic project, the model may exhibit limitations such as hallucinations and its output can be influenced by prompt tokens, potentially leading to inconsistent results. It is not optimized for identity and may generate responses similar to its base models (Qwen/MiniCPM/LLaMA/GPT series).

Finetunes
1 models