Envoid/Dendrite-22B-checkpoint2-ggml

Warning: This model is rather unpredictable.

Dendrite-22B is the 'rehabilitated' version of MindFlay-22B. It is still a work in progress but results have been rather interseting. It's not as good at following role-play type instructions anymore but it excells at creative tasks now. It's ability to respond to lengthy context has been almost completely restored, although,, when addressing certain subjects it becomes somewhat incoherent so further training is planned to complete the 'rehabiliation' before the FP16 version will be released. This repo will contain the q4_0 and q8_0 ggml quantizations of the model at this stage.