localized-ft/Qwen3-32B-german-city-names-kld-20260920-seed1

TEXT GENERATIONPricing:Input $0.408 / Cached $0.0816 / Output $1.972Concurrent Unit Cost:2Model Size:32BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Sep 21, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

The localized-ft/Qwen3-32B-german-city-names-kld-20260920-seed1 is a 32 billion parameter Qwen3-based language model fine-tuned by localized-ft using a LoRA adapter. This model specializes in generating German city names, leveraging its base architecture for robust language understanding. It is designed for applications requiring precise and contextually relevant German geographical entities.

Loading preview...

Model Overview

This model, localized-ft/Qwen3-32B-german-city-names-kld-20260920-seed1, is a fine-tuned version of the Qwen/Qwen3-32B base model. It utilizes a Low-Rank Adaptation (LoRA) adapter to specialize in generating German city names. The repository contains only the LoRA adapter, with redundant merged model shards removed to optimize storage.

Key Capabilities

  • Specialized Generation: Primarily focused on generating accurate and contextually appropriate German city names.
  • LoRA Adapter: Implements a LoRA adapter for efficient fine-tuning, allowing for smaller model sizes and easier deployment compared to full model fine-tuning.
  • Reproducibility: Includes adapter/recovery_manifest.json for retaining training configuration and file checksums, ensuring reproducibility of the fine-tuning process.

Usage and Integration

To use this model, developers need to load the base Qwen/Qwen3-32B model and then apply the provided LoRA adapter. The transformers and peft libraries are used for this process. An example Python snippet is provided in the original README for seamless integration.

Good For

  • Applications requiring the generation or validation of German city names.
  • Geographical data processing and entity recognition tasks specific to Germany.
  • Developers looking for a specialized model with a smaller footprint due to LoRA adaptation.