DavidAU commited on
Commit
5ee9861
1 Parent(s): 998bcaf

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +12 -4
README.md CHANGED
@@ -29,8 +29,6 @@ tags:
29
  pipeline_tag: text-generation
30
  ---
31
 
32
- <I>Quants uploading... , IQs are first, followed by Qs, this will take some time due to file size/local upload speeds. </i>
33
-
34
  <h3>Command-R-01-Ultra-NEO-DARK-HORROR V1 and V2 at 35B in IMATRIX.</h3>
35
 
36
  <img src="command-r-horror.jpg" style="float:right; width:300px; height:300px; padding:10px;">
@@ -80,13 +78,23 @@ I suggest downloading both (of the same quant) and test in your use case(s).
80
  The "flavoring" each brings to Command-R will differ - some slightly, some by a lot - depending on the prompt(s) used / use case(s).
81
  One version is not stronger than the other, they are different and result in different "horror" effects so to speak.
82
 
 
 
 
 
 
 
 
 
 
 
83
  <B>NOTES:</b>
84
 
85
  - This model requires "Command-R" template, responds to standard parameters, and has a max context of 128k (131,000).
86
- - "IQ" quants will be uploaded first, as the Imatrix effect is far stronger in these than "Q" quants. Q quants - all sizes - will upload after the IQ versions.
87
  - IQ4 is the most powerful/balanced in terms of raw power (see examples below), however IQ3/IQ2 may be stronger "horror" wise due to increased IMATRIX effects the lower you go in terms of bit level.
88
  - IQ2s will also be effective due to sheer number of parameters (35 billion) in the model. (see examples below)
89
- - Q4s and Q5s will still be very strong, with Q6 being medium-strong relatively speaking in terms of "horror" changes. This is due to how Imatrix process affects quants of different bit sizes - lower, is stronger, higher is weaker. Again, these are relative.
90
 
91
  <b>Optional Enhancement:</B>
92
 
 
29
  pipeline_tag: text-generation
30
  ---
31
 
 
 
32
  <h3>Command-R-01-Ultra-NEO-DARK-HORROR V1 and V2 at 35B in IMATRIX.</h3>
33
 
34
  <img src="command-r-horror.jpg" style="float:right; width:300px; height:300px; padding:10px;">
 
78
  The "flavoring" each brings to Command-R will differ - some slightly, some by a lot - depending on the prompt(s) used / use case(s).
79
  One version is not stronger than the other, they are different and result in different "horror" effects so to speak.
80
 
81
+ <B>Recommended Quants:</B>
82
+
83
+ - Due to the unique nature of this project, quants IQ1 to IQ4 are recommended for maximum horror effect.
84
+
85
+ - Secondaries are Q2-Q4.
86
+
87
+ - Effects diminish from Q5 and up.
88
+
89
+ - Q8 there is no change (as the Imatrix process does not affect this quant), and therefore not uploaded.
90
+
91
  <B>NOTES:</b>
92
 
93
  - This model requires "Command-R" template, responds to standard parameters, and has a max context of 128k (131,000).
94
+ - "IQ" quants will be uploaded first, as the Imatrix effect is far stronger in these than "Q" quants.
95
  - IQ4 is the most powerful/balanced in terms of raw power (see examples below), however IQ3/IQ2 may be stronger "horror" wise due to increased IMATRIX effects the lower you go in terms of bit level.
96
  - IQ2s will also be effective due to sheer number of parameters (35 billion) in the model. (see examples below)
97
+ - Q4s and Q5s will still be very strong, with Q6 being medium to low relatively speaking in terms of "horror" changes. This is due to how Imatrix process affects quants of different bit sizes - lower, is stronger, higher is weaker. Again, these are relative.
98
 
99
  <b>Optional Enhancement:</B>
100