nightmedia/gemma-4-E2B-Deckard-ShiningValiant3-BF16

Partial fine-tune vs DavidAU/gemma-4-E2B-it-The-DECKARD-HERETIC-UNCENSORED-Thinking. Partial change: 50% of the weights changed (attention, embeddings, mlp, norms, other), in 16 of 16 layers.

Tensors changed: MLP 105/480, Norms 184/206, other 22/122, Attention 148/585, Embeddings 4/4

2 derivatives checked

• mradermacher/gemma-4-E2B-Deckard-ShiningValiant3-BF16-i1-GGUF: Quantization (fewer layers)
• mradermacher/gemma-4-E2B-Deckard-ShiningValiant3-BF16-GGUF: Same weights, converted (fewer layers)

On the Hub · Downloads on Model Pulse