Transformers
GGUF
imatrix
mradermacher commited on
Commit
9c1c86c
·
verified ·
1 Parent(s): e6a31d7

auto-patch README.md

Browse files
Files changed (1) hide show
  1. README.md +23 -0
README.md CHANGED
@@ -76,7 +76,30 @@ more details, including on how to concatenate multi-part files.
76
 
77
  | Link | Type | Size/GB | Notes |
78
  |:-----|:-----|--------:|:------|
 
 
 
 
 
 
 
79
  | [GGUF](https://huggingface.co/mradermacher/bloomz-7b1-i1-GGUF/resolve/main/bloomz-7b1.i1-Q2_K.gguf) | i1-Q2_K | 3.5 | IQ3_XXS probably better |
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
80
 
81
  Here is a handy graph by ikawrakow comparing some lower-quality quant
82
  types (lower is better):
 
76
 
77
  | Link | Type | Size/GB | Notes |
78
  |:-----|:-----|--------:|:------|
79
+ | [GGUF](https://huggingface.co/mradermacher/bloomz-7b1-i1-GGUF/resolve/main/bloomz-7b1.i1-IQ1_S.gguf) | i1-IQ1_S | 2.4 | for the desperate |
80
+ | [GGUF](https://huggingface.co/mradermacher/bloomz-7b1-i1-GGUF/resolve/main/bloomz-7b1.i1-IQ1_M.gguf) | i1-IQ1_M | 2.5 | mostly desperate |
81
+ | [GGUF](https://huggingface.co/mradermacher/bloomz-7b1-i1-GGUF/resolve/main/bloomz-7b1.i1-IQ2_XXS.gguf) | i1-IQ2_XXS | 2.7 | |
82
+ | [GGUF](https://huggingface.co/mradermacher/bloomz-7b1-i1-GGUF/resolve/main/bloomz-7b1.i1-IQ2_XS.gguf) | i1-IQ2_XS | 2.9 | |
83
+ | [GGUF](https://huggingface.co/mradermacher/bloomz-7b1-i1-GGUF/resolve/main/bloomz-7b1.i1-IQ2_S.gguf) | i1-IQ2_S | 3.1 | |
84
+ | [GGUF](https://huggingface.co/mradermacher/bloomz-7b1-i1-GGUF/resolve/main/bloomz-7b1.i1-IQ2_M.gguf) | i1-IQ2_M | 3.3 | |
85
+ | [GGUF](https://huggingface.co/mradermacher/bloomz-7b1-i1-GGUF/resolve/main/bloomz-7b1.i1-Q2_K_S.gguf) | i1-Q2_K_S | 3.3 | very low quality |
86
  | [GGUF](https://huggingface.co/mradermacher/bloomz-7b1-i1-GGUF/resolve/main/bloomz-7b1.i1-Q2_K.gguf) | i1-Q2_K | 3.5 | IQ3_XXS probably better |
87
+ | [GGUF](https://huggingface.co/mradermacher/bloomz-7b1-i1-GGUF/resolve/main/bloomz-7b1.i1-IQ3_XXS.gguf) | i1-IQ3_XXS | 3.6 | lower quality |
88
+ | [GGUF](https://huggingface.co/mradermacher/bloomz-7b1-i1-GGUF/resolve/main/bloomz-7b1.i1-IQ3_XS.gguf) | i1-IQ3_XS | 3.9 | |
89
+ | [GGUF](https://huggingface.co/mradermacher/bloomz-7b1-i1-GGUF/resolve/main/bloomz-7b1.i1-IQ3_S.gguf) | i1-IQ3_S | 4.0 | beats Q3_K* |
90
+ | [GGUF](https://huggingface.co/mradermacher/bloomz-7b1-i1-GGUF/resolve/main/bloomz-7b1.i1-Q3_K_S.gguf) | i1-Q3_K_S | 4.0 | IQ3_XS probably better |
91
+ | [GGUF](https://huggingface.co/mradermacher/bloomz-7b1-i1-GGUF/resolve/main/bloomz-7b1.i1-IQ3_M.gguf) | i1-IQ3_M | 4.3 | |
92
+ | [GGUF](https://huggingface.co/mradermacher/bloomz-7b1-i1-GGUF/resolve/main/bloomz-7b1.i1-Q3_K_M.gguf) | i1-Q3_K_M | 4.5 | IQ3_S probably better |
93
+ | [GGUF](https://huggingface.co/mradermacher/bloomz-7b1-i1-GGUF/resolve/main/bloomz-7b1.i1-IQ4_XS.gguf) | i1-IQ4_XS | 4.7 | |
94
+ | [GGUF](https://huggingface.co/mradermacher/bloomz-7b1-i1-GGUF/resolve/main/bloomz-7b1.i1-Q3_K_L.gguf) | i1-Q3_K_L | 4.8 | IQ3_M probably better |
95
+ | [GGUF](https://huggingface.co/mradermacher/bloomz-7b1-i1-GGUF/resolve/main/bloomz-7b1.i1-IQ4_NL.gguf) | i1-IQ4_NL | 4.9 | prefer IQ4_XS |
96
+ | [GGUF](https://huggingface.co/mradermacher/bloomz-7b1-i1-GGUF/resolve/main/bloomz-7b1.i1-Q4_0.gguf) | i1-Q4_0 | 5.0 | fast, low quality |
97
+ | [GGUF](https://huggingface.co/mradermacher/bloomz-7b1-i1-GGUF/resolve/main/bloomz-7b1.i1-Q4_K_S.gguf) | i1-Q4_K_S | 5.0 | optimal size/speed/quality |
98
+ | [GGUF](https://huggingface.co/mradermacher/bloomz-7b1-i1-GGUF/resolve/main/bloomz-7b1.i1-Q4_K_M.gguf) | i1-Q4_K_M | 5.4 | fast, recommended |
99
+ | [GGUF](https://huggingface.co/mradermacher/bloomz-7b1-i1-GGUF/resolve/main/bloomz-7b1.i1-Q4_1.gguf) | i1-Q4_1 | 5.4 | |
100
+ | [GGUF](https://huggingface.co/mradermacher/bloomz-7b1-i1-GGUF/resolve/main/bloomz-7b1.i1-Q5_K_S.gguf) | i1-Q5_K_S | 5.8 | |
101
+ | [GGUF](https://huggingface.co/mradermacher/bloomz-7b1-i1-GGUF/resolve/main/bloomz-7b1.i1-Q5_K_M.gguf) | i1-Q5_K_M | 6.1 | |
102
+ | [GGUF](https://huggingface.co/mradermacher/bloomz-7b1-i1-GGUF/resolve/main/bloomz-7b1.i1-Q6_K.gguf) | i1-Q6_K | 6.8 | practically like static Q6_K |
103
 
104
  Here is a handy graph by ikawrakow comparing some lower-quality quant
105
  types (lower is better):