sequelbox/Ministral-3-14B-Reasoning-2512-PlumEsper1.1
LoRA-style vs ValiantLabs/Ministral-3-14B-Reasoning-2512-Esper3.1. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of k_proj, o_proj, q_proj, v_proj changed, in 40 of 40 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/1, Norms 0/131, MLP 0/192, Attention 160/256, other 0/4, lm_head 0/1
3 derivatives checked
- mradermacher/Ministral-3-14B-Reasoning-2512-PlumEsper1.1-i1-GGUF: Quantization (fewer layers)
- mradermacher/Ministral-3-14B-Reasoning-2512-PlumEsper1.1-GGUF: Quantization (fewer layers)
- gravermistakes/Ministral-3-14B-Reasoning-2512-PlumEsper1.1-i1-GGUF: Quantization (fewer layers)