qiuxi337/IntrinSight-4B
Partial fine-tune vs google/gemma-3-4b-it. Partial change: 84% of the weights changed (attention, embeddings, mlp, norms, other), in 34 of 34 layers.
Tensors changed: Embeddings 2/4, Norms 123/316, MLP 197/210, Attention 334/352, other 1/1
2 derivatives checked
- mradermacher/IntrinSight-4B-i1-GGUF: Quantization (fewer layers)
- mradermacher/IntrinSight-4B-GGUF: Same weights, converted (fewer layers)