mradermacher commited on
Commit
ab98113
1 Parent(s): 074c842

auto-patch README.md

Browse files
Files changed (1) hide show
  1. README.md +11 -0
README.md CHANGED
@@ -24,16 +24,27 @@ more details, including on how to concatenate multi-part files.
24
  | Link | Type | Size/GB | Notes |
25
  |:-----|:-----|--------:|:------|
26
  | [GGUF](https://huggingface.co/mradermacher/Xwin-LM-70B-V0.1-i1-GGUF/resolve/main/Xwin-LM-70B-V0.1.i1-IQ1_S.gguf) | i1-IQ1_S | 15.5 | for the desperate |
 
27
  | [GGUF](https://huggingface.co/mradermacher/Xwin-LM-70B-V0.1-i1-GGUF/resolve/main/Xwin-LM-70B-V0.1.i1-IQ2_XXS.gguf) | i1-IQ2_XXS | 19.3 | |
28
  | [GGUF](https://huggingface.co/mradermacher/Xwin-LM-70B-V0.1-i1-GGUF/resolve/main/Xwin-LM-70B-V0.1.i1-IQ2_XS.gguf) | i1-IQ2_XS | 21.3 | |
 
 
29
  | [GGUF](https://huggingface.co/mradermacher/Xwin-LM-70B-V0.1-i1-GGUF/resolve/main/Xwin-LM-70B-V0.1.i1-Q2_K.gguf) | i1-Q2_K | 26.4 | IQ3_XXS probably better |
30
  | [GGUF](https://huggingface.co/mradermacher/Xwin-LM-70B-V0.1-i1-GGUF/resolve/main/Xwin-LM-70B-V0.1.i1-IQ3_XXS.gguf) | i1-IQ3_XXS | 28.0 | lower quality |
 
31
  | [GGUF](https://huggingface.co/mradermacher/Xwin-LM-70B-V0.1-i1-GGUF/resolve/main/Xwin-LM-70B-V0.1.i1-Q3_K_XS.gguf) | i1-Q3_K_XS | 29.2 | |
 
32
  | [GGUF](https://huggingface.co/mradermacher/Xwin-LM-70B-V0.1-i1-GGUF/resolve/main/Xwin-LM-70B-V0.1.i1-Q3_K_S.gguf) | i1-Q3_K_S | 30.9 | IQ3_XS probably better |
 
33
  | [GGUF](https://huggingface.co/mradermacher/Xwin-LM-70B-V0.1-i1-GGUF/resolve/main/Xwin-LM-70B-V0.1.i1-Q3_K_M.gguf) | i1-Q3_K_M | 34.2 | IQ3_S probably better |
 
34
  | [GGUF](https://huggingface.co/mradermacher/Xwin-LM-70B-V0.1-i1-GGUF/resolve/main/Xwin-LM-70B-V0.1.i1-Q3_K_L.gguf) | i1-Q3_K_L | 37.1 | IQ3_M probably better |
 
35
  | [GGUF](https://huggingface.co/mradermacher/Xwin-LM-70B-V0.1-i1-GGUF/resolve/main/Xwin-LM-70B-V0.1.i1-Q4_K_S.gguf) | i1-Q4_K_S | 40.2 | optimal size/speed/quality |
36
  | [GGUF](https://huggingface.co/mradermacher/Xwin-LM-70B-V0.1-i1-GGUF/resolve/main/Xwin-LM-70B-V0.1.i1-Q4_K_M.gguf) | i1-Q4_K_M | 42.4 | fast, recommended |
 
 
 
37
 
38
  Here is a handy graph by ikawrakow comparing some lower-quality quant
39
  types (lower is better):
 
24
  | Link | Type | Size/GB | Notes |
25
  |:-----|:-----|--------:|:------|
26
  | [GGUF](https://huggingface.co/mradermacher/Xwin-LM-70B-V0.1-i1-GGUF/resolve/main/Xwin-LM-70B-V0.1.i1-IQ1_S.gguf) | i1-IQ1_S | 15.5 | for the desperate |
27
+ | [GGUF](https://huggingface.co/mradermacher/Xwin-LM-70B-V0.1-i1-GGUF/resolve/main/Xwin-LM-70B-V0.1.i1-IQ1_M.gguf) | i1-IQ1_M | 16.0 | mostly desperate |
28
  | [GGUF](https://huggingface.co/mradermacher/Xwin-LM-70B-V0.1-i1-GGUF/resolve/main/Xwin-LM-70B-V0.1.i1-IQ2_XXS.gguf) | i1-IQ2_XXS | 19.3 | |
29
  | [GGUF](https://huggingface.co/mradermacher/Xwin-LM-70B-V0.1-i1-GGUF/resolve/main/Xwin-LM-70B-V0.1.i1-IQ2_XS.gguf) | i1-IQ2_XS | 21.3 | |
30
+ | [GGUF](https://huggingface.co/mradermacher/Xwin-LM-70B-V0.1-i1-GGUF/resolve/main/Xwin-LM-70B-V0.1.i1-IQ2_S.gguf) | i1-IQ2_S | 21.5 | |
31
+ | [GGUF](https://huggingface.co/mradermacher/Xwin-LM-70B-V0.1-i1-GGUF/resolve/main/Xwin-LM-70B-V0.1.i1-IQ2_M.gguf) | i1-IQ2_M | 23.3 | |
32
  | [GGUF](https://huggingface.co/mradermacher/Xwin-LM-70B-V0.1-i1-GGUF/resolve/main/Xwin-LM-70B-V0.1.i1-Q2_K.gguf) | i1-Q2_K | 26.4 | IQ3_XXS probably better |
33
  | [GGUF](https://huggingface.co/mradermacher/Xwin-LM-70B-V0.1-i1-GGUF/resolve/main/Xwin-LM-70B-V0.1.i1-IQ3_XXS.gguf) | i1-IQ3_XXS | 28.0 | lower quality |
34
+ | [GGUF](https://huggingface.co/mradermacher/Xwin-LM-70B-V0.1-i1-GGUF/resolve/main/Xwin-LM-70B-V0.1.i1-IQ3_XS.gguf) | i1-IQ3_XS | 28.4 | |
35
  | [GGUF](https://huggingface.co/mradermacher/Xwin-LM-70B-V0.1-i1-GGUF/resolve/main/Xwin-LM-70B-V0.1.i1-Q3_K_XS.gguf) | i1-Q3_K_XS | 29.2 | |
36
+ | [GGUF](https://huggingface.co/mradermacher/Xwin-LM-70B-V0.1-i1-GGUF/resolve/main/Xwin-LM-70B-V0.1.i1-IQ3_S.gguf) | i1-IQ3_S | 30.0 | beats Q3_K* |
37
  | [GGUF](https://huggingface.co/mradermacher/Xwin-LM-70B-V0.1-i1-GGUF/resolve/main/Xwin-LM-70B-V0.1.i1-Q3_K_S.gguf) | i1-Q3_K_S | 30.9 | IQ3_XS probably better |
38
+ | [GGUF](https://huggingface.co/mradermacher/Xwin-LM-70B-V0.1-i1-GGUF/resolve/main/Xwin-LM-70B-V0.1.i1-IQ3_M.gguf) | i1-IQ3_M | 31.0 | |
39
  | [GGUF](https://huggingface.co/mradermacher/Xwin-LM-70B-V0.1-i1-GGUF/resolve/main/Xwin-LM-70B-V0.1.i1-Q3_K_M.gguf) | i1-Q3_K_M | 34.2 | IQ3_S probably better |
40
+ | [GGUF](https://huggingface.co/mradermacher/Xwin-LM-70B-V0.1-i1-GGUF/resolve/main/Xwin-LM-70B-V0.1.i1-IQ4_XS.gguf) | i1-IQ4_XS | 36.9 | |
41
  | [GGUF](https://huggingface.co/mradermacher/Xwin-LM-70B-V0.1-i1-GGUF/resolve/main/Xwin-LM-70B-V0.1.i1-Q3_K_L.gguf) | i1-Q3_K_L | 37.1 | IQ3_M probably better |
42
+ | [GGUF](https://huggingface.co/mradermacher/Xwin-LM-70B-V0.1-i1-GGUF/resolve/main/Xwin-LM-70B-V0.1.i1-Q4_0.gguf) | i1-Q4_0 | 39.1 | fast, low quality |
43
  | [GGUF](https://huggingface.co/mradermacher/Xwin-LM-70B-V0.1-i1-GGUF/resolve/main/Xwin-LM-70B-V0.1.i1-Q4_K_S.gguf) | i1-Q4_K_S | 40.2 | optimal size/speed/quality |
44
  | [GGUF](https://huggingface.co/mradermacher/Xwin-LM-70B-V0.1-i1-GGUF/resolve/main/Xwin-LM-70B-V0.1.i1-Q4_K_M.gguf) | i1-Q4_K_M | 42.4 | fast, recommended |
45
+ | [GGUF](https://huggingface.co/mradermacher/Xwin-LM-70B-V0.1-i1-GGUF/resolve/main/Xwin-LM-70B-V0.1.i1-Q5_K_S.gguf) | i1-Q5_K_S | 47.6 | |
46
+ | [GGUF](https://huggingface.co/mradermacher/Xwin-LM-70B-V0.1-i1-GGUF/resolve/main/Xwin-LM-70B-V0.1.i1-Q5_K_M.gguf) | i1-Q5_K_M | 48.9 | |
47
+ | [PART 1](https://huggingface.co/mradermacher/Xwin-LM-70B-V0.1-i1-GGUF/resolve/main/Xwin-LM-70B-V0.1.i1-Q6_K.gguf.part1of2) [PART 2](https://huggingface.co/mradermacher/Xwin-LM-70B-V0.1-i1-GGUF/resolve/main/Xwin-LM-70B-V0.1.i1-Q6_K.gguf.part2of2) | i1-Q6_K | 56.7 | practically like static Q6_K |
48
 
49
  Here is a handy graph by ikawrakow comparing some lower-quality quant
50
  types (lower is better):