kwaikeg/kagentlms_qwen_7b_mat
The kwaikeg/kagentlms_qwen_7b_mat is a 7 billion parameter large language model from KwaiKEG, part of the KAgentLMs series, specifically fine-tuned for advanced Agent capabilities. This model excels in planning, reflection, and tool-use, acquired through Meta-agent tuning. It is designed to power generalized information-seeking agent systems, leveraging a 32K context length for complex tasks.
Loading preview...
KwaiAgents: Agent-Capable Language Models by KwaiKEG
The kwaikeg/kagentlms_qwen_7b_mat is a 7 billion parameter model from the KwaiAgents initiative by KwaiKEG (Kuaishou Technology). This model is part of the KAgentLMs series, which are large language models specifically developed with advanced Agent capabilities.
Key Capabilities & Features
- Agentic Reasoning: Fine-tuned for core Agent functionalities including planning, reflection, and sophisticated tool-use.
- Meta-agent Tuning: Capabilities are acquired through a novel Meta-agent tuning process, enhancing its ability to act as an intelligent agent.
- Context Length: Supports a 32,768 token context window, enabling processing of extensive information for complex agent tasks.
- Integration: Designed to work within the KwaiAgents ecosystem, which includes KAgentSys-Lite (an experimental Agent Loop), KAgentInstruct (fine-tuning data), and KAgentBench (evaluation data).
Use Cases & Deployment
This model is ideal for developing and deploying intelligent agent systems that require advanced reasoning and interaction with external tools. It can be served using vLLM for GPU inference or llama.cpp for CPU deployment, with a converted GGUF version available for llama.cpp.