exdysa/INTELLECT-3-Q4_K_M-layers
Full fine-tune (fewer layers) vs zai-org/GLM-4.5-Air-Base. exdysa/INTELLECT-3-Q4_K_M-layers matches part of zai-org/GLM-4.5-Air-Base (12 of its 18329 tensors; the rest of zai-org/GLM-4.5-Air-Base has no counterpart). On the shared part: exdysa/INTELLECT-3-Q4_K_M-layers is a GGUF (Q4_K/Q8_0) of different weights. Fine-tune: 100% of the weights changed, across 1 of 1 layers.
Tensors changed: Norms 2/2, MLP 3/3, Attention 7/7