mradermacher commited on
Commit
69c161e
·
verified ·
1 Parent(s): 1d4847c

auto-patch README.md

Browse files
Files changed (1) hide show
  1. README.md +5 -0
README.md CHANGED
@@ -40,9 +40,14 @@ more details, including on how to concatenate multi-part files.
40
  | [GGUF](https://huggingface.co/mradermacher/NeuralTrix-bf16-GGUF/resolve/main/NeuralTrix-bf16.Q3_K_S.gguf) | Q3_K_S | 3.3 | |
41
  | [GGUF](https://huggingface.co/mradermacher/NeuralTrix-bf16-GGUF/resolve/main/NeuralTrix-bf16.Q3_K_M.gguf) | Q3_K_M | 3.6 | lower quality |
42
  | [GGUF](https://huggingface.co/mradermacher/NeuralTrix-bf16-GGUF/resolve/main/NeuralTrix-bf16.Q3_K_L.gguf) | Q3_K_L | 3.9 | |
 
43
  | [GGUF](https://huggingface.co/mradermacher/NeuralTrix-bf16-GGUF/resolve/main/NeuralTrix-bf16.Q4_K_S.gguf) | Q4_K_S | 4.2 | fast, recommended |
 
 
 
44
  | [GGUF](https://huggingface.co/mradermacher/NeuralTrix-bf16-GGUF/resolve/main/NeuralTrix-bf16.Q6_K.gguf) | Q6_K | 6.0 | very good quality |
45
  | [GGUF](https://huggingface.co/mradermacher/NeuralTrix-bf16-GGUF/resolve/main/NeuralTrix-bf16.Q8_0.gguf) | Q8_0 | 7.8 | fast, best quality |
 
46
 
47
  Here is a handy graph by ikawrakow comparing some lower-quality quant
48
  types (lower is better):
 
40
  | [GGUF](https://huggingface.co/mradermacher/NeuralTrix-bf16-GGUF/resolve/main/NeuralTrix-bf16.Q3_K_S.gguf) | Q3_K_S | 3.3 | |
41
  | [GGUF](https://huggingface.co/mradermacher/NeuralTrix-bf16-GGUF/resolve/main/NeuralTrix-bf16.Q3_K_M.gguf) | Q3_K_M | 3.6 | lower quality |
42
  | [GGUF](https://huggingface.co/mradermacher/NeuralTrix-bf16-GGUF/resolve/main/NeuralTrix-bf16.Q3_K_L.gguf) | Q3_K_L | 3.9 | |
43
+ | [GGUF](https://huggingface.co/mradermacher/NeuralTrix-bf16-GGUF/resolve/main/NeuralTrix-bf16.IQ4_XS.gguf) | IQ4_XS | 4.0 | |
44
  | [GGUF](https://huggingface.co/mradermacher/NeuralTrix-bf16-GGUF/resolve/main/NeuralTrix-bf16.Q4_K_S.gguf) | Q4_K_S | 4.2 | fast, recommended |
45
+ | [GGUF](https://huggingface.co/mradermacher/NeuralTrix-bf16-GGUF/resolve/main/NeuralTrix-bf16.Q4_K_M.gguf) | Q4_K_M | 4.5 | fast, recommended |
46
+ | [GGUF](https://huggingface.co/mradermacher/NeuralTrix-bf16-GGUF/resolve/main/NeuralTrix-bf16.Q5_K_S.gguf) | Q5_K_S | 5.1 | |
47
+ | [GGUF](https://huggingface.co/mradermacher/NeuralTrix-bf16-GGUF/resolve/main/NeuralTrix-bf16.Q5_K_M.gguf) | Q5_K_M | 5.2 | |
48
  | [GGUF](https://huggingface.co/mradermacher/NeuralTrix-bf16-GGUF/resolve/main/NeuralTrix-bf16.Q6_K.gguf) | Q6_K | 6.0 | very good quality |
49
  | [GGUF](https://huggingface.co/mradermacher/NeuralTrix-bf16-GGUF/resolve/main/NeuralTrix-bf16.Q8_0.gguf) | Q8_0 | 7.8 | fast, best quality |
50
+ | [GGUF](https://huggingface.co/mradermacher/NeuralTrix-bf16-GGUF/resolve/main/NeuralTrix-bf16.f16.gguf) | f16 | 14.6 | 16 bpw, overkill |
51
 
52
  Here is a handy graph by ikawrakow comparing some lower-quality quant
53
  types (lower is better):