tacodevs/Behemoth-T1-123B
LoRA-style vs TheDrummer/Behemoth-R1-123B-v2. 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 88 of 88 layers; norms, biases and embeddings are untouched.
Tensors changed: lm_head 0/1, Embeddings 0/1, Norms 0/177, MLP 264/264, Attention 352/352
2 derivatives checked
- mradermacher/Behemoth-T1-123B-i1-GGUF: Quantization
- mradermacher/Behemoth-T1-123B-GGUF: Quantization