McGill-NLP/AfriqueQwen3.5-4B
Full fine-tune vs Qwen/Qwen3.5-4B-Base. Fine-tune: 93% of the weights changed, across 32 of 32 layers.
Tensors changed: Embeddings 1/4, MLP 96/196, Attention 221/320, Norms 105/203
9 derivatives checked
- mradermacher/AfriqueQwen3.5-4B-i1-GGUF: Quantization
- McGill-NLP/AfriqueQwen3.5-4B-Instruct-v2: Full fine-tune (fewer layers)
- israel/AfriGuard-AfriqueQwen3.5-4B: Full fine-tune
- adzcai/AfriGuardInst-AfriqueQwen3.5-4B-lora-merged: LoRA-style
- adzcai/AfriGuardInst-AfriqueQwen3.5-4B: Full fine-tune
- adzcai/AfriGuardPlain-AfriqueQwen3.5-4B: Full fine-tune
- adzcai/AfriGuardPlain-AfriqueQwen3.5-4B-lora-merged: LoRA-style
- McGill-NLP/AfriqueQwen3.5-4B-Instruct-v1: Full fine-tune (fewer layers)
- mradermacher/AfriqueQwen3.5-4B-GGUF: Same weights, converted