  e88 88e                               d8     
 d888 888b  8888 8888  ,"Y88b 888 8e   d88     
C8888 8888D 8888 8888 "8" 888 888 88b d88888   
 Y888 888P  Y888 888P ,ee 888 888 888  888     
  "88 88"    "88 88"  "88 888 888 888  888     
      b                                        
      8b,                                      
 
  e88'Y88                  d8           888    
 d888  'Y  ,"Y88b 888,8,  d88    ,e e,  888    
C8888     "8" 888 888 "  d88888 d88 88b 888    
 Y888  ,d ,ee 888 888     888   888   , 888    
  "88,d88 "88 888 888     888    "YeeP" 888    
                                               
PROUDLY PRESENTS

Phainesthesia-8x7B-iMat-GGUF

Update 4/28 This specific model was retested once more after llama.cpp PR 6884 was merged. It has been confirmed to pass the recently added --check-tensors test.

Quantized from fp16 with love.

Quantizations made possible using mixtral-8x7b-instruct-v0.1.imatrix file from this repo (special thanks to ikawrakow again)

For a brief rundown of iMatrix quant performance please see this PR

All quants are verified working prior to uploading to repo for your safety and convenience.

Please note importance matrix quantizations are a work in progress, IQ3 and above is recommended for best results.

Tip: Pick a size that can fit in your GPU while still allowing some room for context for best speed from the table below. You may need to pad this further depending on if you are running image gen or TTS as well.

Quant	Size (GB)	Comments
IQ2_S	14.1
IQ2_M	15.5
IQ3_XXS	18.2
IQ3_XS	19.3
IQ3_S	20.4
IQ3_M	21.4
IQ4_XS	25.1	Better quality than Q3_K_L and below
Q4_K_M	28.4
Q5_K_M	33.2
Q6_K	38.4
Q8_0	49.6

Original model card can be found here and below.

Warning this model is highly experimental and unpredictable.

Phainesthesia-8x7B began its life as a copy of Verdict-DADA-8x7B

In a training configuration accident it gained some rather uncanny abilities but was extremely disorderly. Further experimentation was done to decide upon a course of action going forward.

It then received a corrective qlora session to attempt to correct the disorder.

It hasn't been tested on long roleplay scenarios yet. Although testing with simple temperatures with t=1 while requiring frequent rerolls the messages that are good are good, although it is constantly walking the razor's edge with looping on one side and a complete loss of coherence on the other (with the edge of the razor being pure gold)

Its assistant personality on determinstic sampling is rather run of the mill but with simple sampling at t=1 It gives much more interesting replies and is prone to hallucination when asking leading questions. Others might refer to this as an imagination.