Weeb-v1.0-Kaizoku-7B-Alpha

Weeb-v1.0-Kaizoku-7B-Alpha : GGUF

An (experimental) AI model fine-tuned to emulate an anime otaku, with extensive knowledge on anime and manga. The AI is still a huge WIP, but I hope you find it useful.

Can be used for Casual conversation, anime trivia, and dere-roleplay for fans of anime/manga.

• October 2023 knowledge cutoff
• It currently has a large gap when it comes to roleplaying a specific character (like Usopp). I am working on fixing this in the next version

[!Note] This model is in ALPHA. It's knowledge is only as good as the training data; expect gaps and occasional hallucinated "facts," particularly for niche or newer series

Available Model files:

• Q4_K_M gguf

Model Details

• Base model: Qwen2.5-7B-Instruct (4-bit, via Unsloth)
• Method: LoRA (rank 32, alpha 64, all modules), trained with Unsloth Studio
• Training data: ~18,500 conversations assembled from public Hugging Face datasets covering character profiles, anime trivia/lore Q&A, iconic quotes, personality-driven dialogue, plot synopses/recommendations, and filtered roleplay conversations. Ill link them when available
• Versioning: X bump for major changes like revamps, Y bump for minor changes and patche
• Naming convention: Following a seperately tiered scheme inspired by One Piece ranks — Kaizoku (7B) → Shichibukai → Yonko — analogous to Haiku/Sonnet/Opus. Determined by the size of the base model

Community

If you would like to discuss about Weeb or have any questions, please join the community Discord server

Usage

CLI usage:

• For text only LLMs: llama-cli -hf ciabi/Weeb-v1.0-Kaizoku-7B-Alpha --jinja
• For multimodal models: llama-mtmd-cli -hf ciabi/Weeb-v1.0-Kaizoku-7B-Alpha --jinja

Ollama

An Ollama Modelfile is included for easy deployment.

---

This was trained 2x faster and finetuned with Unsloth <img src="https://raw.githubusercontent.com/unslothai/unsloth/main/images/unsloth%20made%20with%20love.png" width="200"/>