ConicCat/Apriel-R1P
LoRA-style vs ServiceNow-AI/Apriel-Nemotron-15b-Thinker. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of down_proj, gate_proj, k_proj, o_proj, q_proj, up_proj, v_proj changed, in 50 of 50 layers; norms, biases and embeddings are untouched.
Tensors changed: lm_head 0/1, Embeddings 0/1, Norms 0/101, MLP 150/150, Attention 200/200
2 derivatives checked
- mradermacher/Apriel-R1P-i1-GGUF: Quantization
- mradermacher/Apriel-R1P-GGUF: Quantization