Nickyang/GLM-4.7-Flash-RAZOR-17B-A3B-E32of64
LoRA-style (fewer layers) vs zai-org/GLM-4.7-Flash. Nickyang/GLM-4.7-Flash-RAZOR-17B-A3B-E32of64 matches part of zai-org/GLM-4.7-Flash (5097 of its 9703 tensors; the rest of zai-org/GLM-4.7-Flash has no counterpart). On the shared part: Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of down_proj, gate_proj, up_proj changed, in 47 of 48 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/2, Norms 0/196, MLP 1091/1275, Attention 0/240, lm_head 0/1, other 0/2