CloudGoat/Mephisto-1.0-4B
Full fine-tune (fewer layers) vs BAAI/AREX-Turbo. CloudGoat/Mephisto-1.0-4B matches part of BAAI/AREX-Turbo (426 of its 723 tensors; the rest of BAAI/AREX-Turbo has no counterpart). On the shared part: Fine-tune: 100% of the weights changed, across 32 of 32 layers.
Tensors changed: Embeddings 1/1, Norms 81/105, Attention 193/224, MLP 96/96
2 derivatives checked
- mradermacher/Mephisto-1.0-4B-i1-GGUF: Quantization
- mradermacher/Mephisto-1.0-4B-GGUF: Same weights, converted