bunnycore/Qwen2.5-7B-RRP-1M-Thinker
The bunnycore/Qwen2.5-7B-RRP-1M-Thinker is a 7.6 billion parameter language model, merged using the SCE method with bunnycore/Qwen2.5-7B-RRP-1M as its base. This model integrates capabilities from several Qwen-based models, including OpenR1-Qwen-7B and OpenThinker-7B, to enhance its reasoning and general language understanding. It is designed for applications requiring a robust 7B-class model with a 32K context length, benefiting from a diverse merge of specialized Qwen variants.
Loading preview...
Model Overview
The bunnycore/Qwen2.5-7B-RRP-1M-Thinker is a 7.6 billion parameter language model created through a sophisticated merge process using mergekit. It leverages the SCE (Selective Channel Expansion) merge method, building upon bunnycore/Qwen2.5-7B-RRP-1M as its foundational base model. This approach combines the strengths of multiple specialized Qwen-based models to produce a more versatile and capable language model.
Key Merged Components
This model integrates capabilities from a diverse set of Qwen-based models, including:
open-r1/OpenR1-Qwen-7Bbunnycore/Qwen2.5-7B-MixStock-Sce-V0.3bunnycore/Qwen-2.5-7B-Deep-Stock-v4bunnycore/QandoraExp-7Bopen-thoughts/OpenThinker-7B
Technical Configuration
The merge process utilized specific parameters, including select_topk: 1.5 and int8_mask: true, with the model's dtype set to bfloat16. This configuration aims to optimize the integration of the constituent models, enhancing overall performance and efficiency.
Potential Use Cases
Given its merged architecture and diverse origins, this model is suitable for applications that benefit from a broad range of language understanding and generation capabilities within a 7B parameter constraint. Its 32K context length further supports more complex and extended interactions.