armand0e/Qwen3.5-9B-ultra-uncensored-heretic
armand0e/Qwen3.5-9B-ultra-uncensored-heretic is a 9 billion parameter language model, a Transformers-compatible repair of the upstream Qwen3.5 checkpoint. This model addresses invalid tensor prefixes found in the original safetensors, specifically for language-model and vision tensors. It preserves the raw tensor data while correcting header names, making it suitable for applications requiring a functional Qwen3.5 variant.
Loading preview...
Overview
This model, armand0e/Qwen3.5-9B-ultra-uncensored-heretic, is a 9 billion parameter language model that serves as a Transformers-compatible repair of the original Qwen3.5 checkpoint. The primary purpose of this release is to correct structural issues found in the upstream safetensors.
Key Characteristics
- Parameter Count: 9 billion parameters.
- Context Length: 32768 tokens.
- Structural Repair: The model specifically fixes invalid tensor prefixes present in the original Qwen3.5 safetensors. This includes 426 language-model tensors under a triple
language_modelnesting and 333 vision tensors undermodel.language_model.visual. - Data Integrity: Crucially, this repair only modifies the safetensors header names; the raw tensor data is preserved byte-for-byte, ensuring the underlying model weights remain unchanged.
Use Cases
This model is ideal for developers and researchers who require a functionally corrected Qwen3.5 checkpoint for their applications. It ensures compatibility with the Transformers library by resolving the structural issues that might have hindered its direct use, particularly for tasks involving both language and vision components as indicated by the original tensor structure.