allura-org/MoE-Girl-800MA-3BT
Partial fine-tune vs ibm-granite/granite-3.0-3b-a800m-base. Partial change: 6% of the weights changed (attention, other), in 32 of 32 layers.
Tensors changed: other 32/96, Norms 0/65, Attention 128/128
2 derivatives checked
- mradermacher/MoE-Girl-800MA-3BT-GGUF: Same weights, converted
- DevQuasar-8/allura-org.MoE-Girl-800MA-3BT-GGUF: Quantization