google/gemma-3-12b-it-qat-int4-unquantized
Full fine-tune vs google/gemma-3-12b-it. Fine-tune: 97% of the weights changed, across 48 of 48 layers.
Tensors changed: Embeddings 1/4, Norms 290/400, MLP 144/252, Attention 192/408, other 1/1
3 derivatives checked
- unsloth/gemma-3-12b-it-qat-int4-GGUF: Same weights, converted (fewer layers)
- DevQuasar-12/google.gemma-3-12b-it-qat-int4-unquantized-GGUF: Same weights, converted (fewer layers)
- megabytes/gemma-3-12b-qat-int4-heretic: Abliterated
1 other repos hold byte-for-byte identical weight files.