shreydan/SmolVLM-256M-Detection
Partial fine-tune vs HuggingFaceTB/SmolVLM-256M-Instruct. Partial change: 21% of the weights changed (attention), in 30 of 30 layers.
Tensors changed: lm_head 0/1, other 0/1, Embeddings 0/4, Norms 0/111, MLP 0/138, Attention 216/216
2 derivatives checked
- mradermacher/SmolVLM-256M-Detection-i1-GGUF: Quantization (fewer layers)
- mradermacher/SmolVLM-256M-Detection-GGUF: Same weights, converted (fewer layers)