howey/LLaDA-8B-Instruct-DLPO
Partial fine-tune vs GSAI-ML/LLaDA-8B-Instruct. Partial change: 87% of the weights changed (attention, mlp, norms, other), in 32 of 32 layers.
Tensors changed: Attention 128/128, Norms 18/65, Embeddings 0/1, other 64/65, MLP 32/32
2 derivatives checked
- mradermacher/LLaDA-8B-Instruct-DLPO-i1-GGUF: Quantization
- mradermacher/LLaDA-8B-Instruct-DLPO-GGUF: Same weights, converted