mradermacher commited on
Commit
cdeb6ab
·
verified ·
1 Parent(s): 880d5fd

auto-patch README.md

Browse files
Files changed (1) hide show
  1. README.md +20 -0
README.md CHANGED
@@ -34,9 +34,29 @@ more details, including on how to concatenate multi-part files.
34
 
35
  | Link | Type | Size/GB | Notes |
36
  |:-----|:-----|--------:|:------|
 
 
 
 
 
 
 
37
  | [GGUF](https://huggingface.co/mradermacher/Ultiima-78B-v2-i1-GGUF/resolve/main/Ultiima-78B-v2.i1-Q2_K.gguf) | i1-Q2_K | 31.9 | IQ3_XXS probably better |
 
 
 
 
38
  | [GGUF](https://huggingface.co/mradermacher/Ultiima-78B-v2-i1-GGUF/resolve/main/Ultiima-78B-v2.i1-IQ3_M.gguf) | i1-IQ3_M | 38.0 | |
 
 
 
 
39
  | [GGUF](https://huggingface.co/mradermacher/Ultiima-78B-v2-i1-GGUF/resolve/main/Ultiima-78B-v2.i1-Q4_K_S.gguf) | i1-Q4_K_S | 47.0 | optimal size/speed/quality |
 
 
 
 
 
40
 
41
  Here is a handy graph by ikawrakow comparing some lower-quality quant
42
  types (lower is better):
 
34
 
35
  | Link | Type | Size/GB | Notes |
36
  |:-----|:-----|--------:|:------|
37
+ | [GGUF](https://huggingface.co/mradermacher/Ultiima-78B-v2-i1-GGUF/resolve/main/Ultiima-78B-v2.i1-IQ1_S.gguf) | i1-IQ1_S | 24.4 | for the desperate |
38
+ | [GGUF](https://huggingface.co/mradermacher/Ultiima-78B-v2-i1-GGUF/resolve/main/Ultiima-78B-v2.i1-IQ1_M.gguf) | i1-IQ1_M | 25.5 | mostly desperate |
39
+ | [GGUF](https://huggingface.co/mradermacher/Ultiima-78B-v2-i1-GGUF/resolve/main/Ultiima-78B-v2.i1-IQ2_XXS.gguf) | i1-IQ2_XXS | 27.4 | |
40
+ | [GGUF](https://huggingface.co/mradermacher/Ultiima-78B-v2-i1-GGUF/resolve/main/Ultiima-78B-v2.i1-IQ2_XS.gguf) | i1-IQ2_XS | 29.1 | |
41
+ | [GGUF](https://huggingface.co/mradermacher/Ultiima-78B-v2-i1-GGUF/resolve/main/Ultiima-78B-v2.i1-IQ2_S.gguf) | i1-IQ2_S | 30.0 | |
42
+ | [GGUF](https://huggingface.co/mradermacher/Ultiima-78B-v2-i1-GGUF/resolve/main/Ultiima-78B-v2.i1-IQ2_M.gguf) | i1-IQ2_M | 31.5 | |
43
+ | [GGUF](https://huggingface.co/mradermacher/Ultiima-78B-v2-i1-GGUF/resolve/main/Ultiima-78B-v2.i1-Q2_K_S.gguf) | i1-Q2_K_S | 31.7 | very low quality |
44
  | [GGUF](https://huggingface.co/mradermacher/Ultiima-78B-v2-i1-GGUF/resolve/main/Ultiima-78B-v2.i1-Q2_K.gguf) | i1-Q2_K | 31.9 | IQ3_XXS probably better |
45
+ | [GGUF](https://huggingface.co/mradermacher/Ultiima-78B-v2-i1-GGUF/resolve/main/Ultiima-78B-v2.i1-IQ3_XXS.gguf) | i1-IQ3_XXS | 34.1 | lower quality |
46
+ | [GGUF](https://huggingface.co/mradermacher/Ultiima-78B-v2-i1-GGUF/resolve/main/Ultiima-78B-v2.i1-IQ3_XS.gguf) | i1-IQ3_XS | 35.2 | |
47
+ | [GGUF](https://huggingface.co/mradermacher/Ultiima-78B-v2-i1-GGUF/resolve/main/Ultiima-78B-v2.i1-Q3_K_S.gguf) | i1-Q3_K_S | 36.9 | IQ3_XS probably better |
48
+ | [GGUF](https://huggingface.co/mradermacher/Ultiima-78B-v2-i1-GGUF/resolve/main/Ultiima-78B-v2.i1-IQ3_S.gguf) | i1-IQ3_S | 37.0 | beats Q3_K* |
49
  | [GGUF](https://huggingface.co/mradermacher/Ultiima-78B-v2-i1-GGUF/resolve/main/Ultiima-78B-v2.i1-IQ3_M.gguf) | i1-IQ3_M | 38.0 | |
50
+ | [GGUF](https://huggingface.co/mradermacher/Ultiima-78B-v2-i1-GGUF/resolve/main/Ultiima-78B-v2.i1-Q3_K_M.gguf) | i1-Q3_K_M | 40.4 | IQ3_S probably better |
51
+ | [GGUF](https://huggingface.co/mradermacher/Ultiima-78B-v2-i1-GGUF/resolve/main/Ultiima-78B-v2.i1-Q3_K_L.gguf) | i1-Q3_K_L | 42.4 | IQ3_M probably better |
52
+ | [GGUF](https://huggingface.co/mradermacher/Ultiima-78B-v2-i1-GGUF/resolve/main/Ultiima-78B-v2.i1-IQ4_XS.gguf) | i1-IQ4_XS | 42.7 | |
53
+ | [GGUF](https://huggingface.co/mradermacher/Ultiima-78B-v2-i1-GGUF/resolve/main/Ultiima-78B-v2.i1-Q4_0.gguf) | i1-Q4_0 | 44.4 | fast, low quality |
54
  | [GGUF](https://huggingface.co/mradermacher/Ultiima-78B-v2-i1-GGUF/resolve/main/Ultiima-78B-v2.i1-Q4_K_S.gguf) | i1-Q4_K_S | 47.0 | optimal size/speed/quality |
55
+ | [GGUF](https://huggingface.co/mradermacher/Ultiima-78B-v2-i1-GGUF/resolve/main/Ultiima-78B-v2.i1-Q4_1.gguf) | i1-Q4_1 | 49.1 | |
56
+ | [PART 1](https://huggingface.co/mradermacher/Ultiima-78B-v2-i1-GGUF/resolve/main/Ultiima-78B-v2.i1-Q4_K_M.gguf.part1of2) [PART 2](https://huggingface.co/mradermacher/Ultiima-78B-v2-i1-GGUF/resolve/main/Ultiima-78B-v2.i1-Q4_K_M.gguf.part2of2) | i1-Q4_K_M | 50.8 | fast, recommended |
57
+ | [PART 1](https://huggingface.co/mradermacher/Ultiima-78B-v2-i1-GGUF/resolve/main/Ultiima-78B-v2.i1-Q5_K_S.gguf.part1of2) [PART 2](https://huggingface.co/mradermacher/Ultiima-78B-v2-i1-GGUF/resolve/main/Ultiima-78B-v2.i1-Q5_K_S.gguf.part2of2) | i1-Q5_K_S | 55.2 | |
58
+ | [PART 1](https://huggingface.co/mradermacher/Ultiima-78B-v2-i1-GGUF/resolve/main/Ultiima-78B-v2.i1-Q5_K_M.gguf.part1of2) [PART 2](https://huggingface.co/mradermacher/Ultiima-78B-v2-i1-GGUF/resolve/main/Ultiima-78B-v2.i1-Q5_K_M.gguf.part2of2) | i1-Q5_K_M | 58.4 | |
59
+ | [PART 1](https://huggingface.co/mradermacher/Ultiima-78B-v2-i1-GGUF/resolve/main/Ultiima-78B-v2.i1-Q6_K.gguf.part1of2) [PART 2](https://huggingface.co/mradermacher/Ultiima-78B-v2-i1-GGUF/resolve/main/Ultiima-78B-v2.i1-Q6_K.gguf.part2of2) | i1-Q6_K | 69.1 | practically like static Q6_K |
60
 
61
  Here is a handy graph by ikawrakow comparing some lower-quality quant
62
  types (lower is better):