prithivMLmods/Qwen2.5-14B-DeepSeek-R1-1M
The prithivMLmods/Qwen2.5-14B-DeepSeek-R1-1M is a 14.8 billion parameter merged language model, built upon the Qwen2.5-14B-Instruct-1M base using the TIES merge method. It combines deepseek-ai/DeepSeek-R1-Distill-Qwen-14B and Qwen/Qwen2.5-14B-Instruct with equal weighting and density. This model is optimized for performance through normalization, int8 masking, and bfloat16 precision, offering a unique blend of capabilities from its constituent models.
Loading preview...
Overview
This model, prithivMLmods/Qwen2.5-14B-DeepSeek-R1-1M, is a 14.8 billion parameter merged language model. It was created using the TIES merge method via MergeKit, building on Qwen/Qwen2.5-14B-Instruct-1M as its base. The merge process integrates two distinct models: deepseek-ai/DeepSeek-R1-Distill-Qwen-14B and Qwen/Qwen2.5-14B-Instruct, each contributing with equal weight and density.
Key Capabilities
- Hybrid Intelligence: Combines the strengths of
DeepSeek-R1-Distill-Qwen-14BandQwen2.5-14B-Instruct, potentially offering enhanced reasoning and instruction-following capabilities. - Optimized Performance: Configured with normalization, int8 masking, and
bfloat16precision for efficient operation. - MergeKit Integration: Developed using a structured and reproducible merge process, allowing for transparent understanding of its composition.
Good for
- Experimentation with Merged Models: Ideal for researchers and developers interested in exploring the performance characteristics of models created via the TIES merge method.
- Leveraging Combined Strengths: Suitable for applications that could benefit from the blended capabilities of the DeepSeek and Qwen 2.5 instruction-tuned models.
- Resource-Efficient Deployment: The
bfloat16precision and int8 masking contribute to potentially more efficient inference compared to full precision models of similar size.