trinhkhng/ties_Merged_Qwen2-0.5B_0.3

TEXT GENERATIONPricing:Input $0.04 / Cached $0.008 / Output $0.08Concurrent Unit Cost:1Model Size:0.5BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Aug 6, 2026Architecture:Transformer Featherless Exclusive Cold

trinhkhng/ties_Merged_Qwen2-0.5B_0.3 is a 0.5 billion parameter language model based on the Qwen2 architecture, created by trinhkhng. This model was produced using the TIES merge method, combining a Qwen2-0.5B base with a debiased Qwen2-0.5B variant. Its primary differentiation lies in its merge technique, which aims to integrate specific characteristics from the debiased model into the base Qwen2 architecture.

Loading preview...

Model Overview

trinhkhng/ties_Merged_Qwen2-0.5B_0.3 is a 0.5 billion parameter language model built upon the Qwen2 architecture. This model is a product of a merge operation, specifically utilizing the TIES (Trimmed, Iterative, and Selective) merge method. The merging process combined a base /kaggle/working/Qwen2-0.5B model with a debiased variant, /kaggle/working/debias_Qwen2-0.5B, to create a consolidated model.

Merge Details

  • Merge Method: TIES (Trimmed, Iterative, and Selective), as described in the TIES paper.
  • Base Model: /kaggle/working/Qwen2-0.5B.
  • Merged Component: /kaggle/working/debias_Qwen2-0.5B, integrated with a density of 0.5 and a weight of 1.0.
  • Configuration: The merge was performed with int8_mask: true and normalize: true parameters, with a lambda value of 0.3.

Potential Use Cases

This model is suitable for applications requiring a compact Qwen2-based model that incorporates characteristics from a debiased source. Its 0.5 billion parameters and 32768 token context length make it efficient for tasks where resource constraints are a consideration, while the TIES merge method suggests an intentional combination of model properties.

Finetunes
1 models