sakusakumura/Qwen2-7b-cleanup
The sakusakumura/Qwen2-7b-cleanup is a 7.6 billion parameter Qwen2-based causal language model developed by sakusakumura. This model was fine-tuned from unsloth/Qwen2-7B, leveraging Unsloth and Huggingface's TRL library for accelerated training. It is specifically designed for text quality assessment, classifying input text as either 'low quality' or 'high quality' based on predefined criteria. Its primary application is evaluating text for repetition, grammatical errors, coherence, and inappropriate content.
Loading preview...
Overview
The sakusakumura/Qwen2-7b-cleanup is a 7.6 billion parameter language model based on the Qwen2 architecture. Developed by sakusakumura, this model is a fine-tuned version of unsloth/Qwen2-7B and was trained with significant speed improvements using the Unsloth library in conjunction with Huggingface's TRL.
Key Capabilities
This model specializes in text quality assessment, providing a binary classification of input text as either "low quality" or "high quality." It evaluates text against specific criteria:
- Repetition: Identifies sentences that are not repetitive.
- Grammar: Detects grammatically incorrect sentences.
- Coherence: Flags incoherent or nonsensical sentences.
- Content Appropriateness: Screens for discriminatory, violent, or obscene content.
Prompt Format
The model expects a structured prompt format to perform its text quality evaluation. Users should provide instructions within <instruction> tags, outlining the assessment conditions, and the target text within <task> tags. The model then outputs its quality judgment within <answer> tags.
Good For
- Automated content moderation for basic quality checks.
- Filtering out undesirable text based on predefined rules.
- Assessing the structural and semantic quality of generated or user-submitted text.