lewiswatson/Frame2KG-SmolVLM-256m-JSON
LoRA-style (fewer layers) vs HuggingFaceTB/SmolVLM2-256M-Video-Instruct. lewiswatson/Frame2KG-SmolVLM-256m-JSON matches part of HuggingFaceTB/SmolVLM2-256M-Video-Instruct (273 of its 471 tensors; the rest of HuggingFaceTB/SmolVLM2-256M-Video-Instruct has no counterpart). On the shared part: lewiswatson/Frame2KG-SmolVLM-256m-JSON is a GGUF (F16) of different weights. 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 30 of 30 layers; norms, biases and embeddings are untouched.
Tensors changed: lm_head 0/1, Embeddings 0/1, Norms 0/61, MLP 90/90, Attention 120/120