mradermacher commited on
Commit
81944a4
·
verified ·
1 Parent(s): fdea0b5

auto-patch README.md

Browse files
Files changed (1) hide show
  1. README.md +84 -0
README.md CHANGED
@@ -1,6 +1,90 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  <!-- ### quantize_version: 2 -->
2
  <!-- ### output_tensor_quantised: 1 -->
3
  <!-- ### convert_type: hf -->
4
  <!-- ### vocab_type: -->
5
  <!-- ### tags: nicoboss -->
6
  weighted/imatrix quants of https://huggingface.co/shadowlilac/Llama-4-Scout-17B-6E-Instruct
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ base_model: shadowlilac/Llama-4-Scout-17B-6E-Instruct
3
+ language:
4
+ - ar
5
+ - de
6
+ - en
7
+ - es
8
+ - fr
9
+ - hi
10
+ - id
11
+ - it
12
+ - pt
13
+ - th
14
+ - tl
15
+ - vi
16
+ library_name: transformers
17
+ quantized_by: mradermacher
18
+ tags:
19
+ - pytorch
20
+ - llama
21
+ - llama-4
22
+ - mixture of experts
23
+ ---
24
+ ## About
25
+
26
  <!-- ### quantize_version: 2 -->
27
  <!-- ### output_tensor_quantised: 1 -->
28
  <!-- ### convert_type: hf -->
29
  <!-- ### vocab_type: -->
30
  <!-- ### tags: nicoboss -->
31
  weighted/imatrix quants of https://huggingface.co/shadowlilac/Llama-4-Scout-17B-6E-Instruct
32
+
33
+ <!-- provided-files -->
34
+ static quants are available at https://huggingface.co/mradermacher/Llama-4-Scout-17B-6E-Instruct-GGUF
35
+ ## Usage
36
+
37
+ If you are unsure how to use GGUF files, refer to one of [TheBloke's
38
+ READMEs](https://huggingface.co/TheBloke/KafkaLM-70B-German-V0.1-GGUF) for
39
+ more details, including on how to concatenate multi-part files.
40
+
41
+ ## Provided Quants
42
+
43
+ (sorted by size, not necessarily quality. IQ-quants are often preferable over similar sized non-IQ quants)
44
+
45
+ | Link | Type | Size/GB | Notes |
46
+ |:-----|:-----|--------:|:------|
47
+ | [GGUF](https://huggingface.co/mradermacher/Llama-4-Scout-17B-6E-Instruct-i1-GGUF/resolve/main/Llama-4-Scout-17B-6E-Instruct.i1-IQ1_S.gguf) | i1-IQ1_S | 10.3 | for the desperate |
48
+ | [GGUF](https://huggingface.co/mradermacher/Llama-4-Scout-17B-6E-Instruct-i1-GGUF/resolve/main/Llama-4-Scout-17B-6E-Instruct.i1-IQ1_M.gguf) | i1-IQ1_M | 11.3 | mostly desperate |
49
+ | [GGUF](https://huggingface.co/mradermacher/Llama-4-Scout-17B-6E-Instruct-i1-GGUF/resolve/main/Llama-4-Scout-17B-6E-Instruct.i1-IQ2_XXS.gguf) | i1-IQ2_XXS | 13.0 | |
50
+ | [GGUF](https://huggingface.co/mradermacher/Llama-4-Scout-17B-6E-Instruct-i1-GGUF/resolve/main/Llama-4-Scout-17B-6E-Instruct.i1-IQ2_XS.gguf) | i1-IQ2_XS | 14.4 | |
51
+ | [GGUF](https://huggingface.co/mradermacher/Llama-4-Scout-17B-6E-Instruct-i1-GGUF/resolve/main/Llama-4-Scout-17B-6E-Instruct.i1-IQ2_S.gguf) | i1-IQ2_S | 14.7 | |
52
+ | [GGUF](https://huggingface.co/mradermacher/Llama-4-Scout-17B-6E-Instruct-i1-GGUF/resolve/main/Llama-4-Scout-17B-6E-Instruct.i1-IQ2_M.gguf) | i1-IQ2_M | 16.1 | |
53
+ | [GGUF](https://huggingface.co/mradermacher/Llama-4-Scout-17B-6E-Instruct-i1-GGUF/resolve/main/Llama-4-Scout-17B-6E-Instruct.i1-Q2_K_S.gguf) | i1-Q2_K_S | 16.6 | very low quality |
54
+ | [GGUF](https://huggingface.co/mradermacher/Llama-4-Scout-17B-6E-Instruct-i1-GGUF/resolve/main/Llama-4-Scout-17B-6E-Instruct.i1-Q2_K.gguf) | i1-Q2_K | 17.8 | IQ3_XXS probably better |
55
+ | [GGUF](https://huggingface.co/mradermacher/Llama-4-Scout-17B-6E-Instruct-i1-GGUF/resolve/main/Llama-4-Scout-17B-6E-Instruct.i1-IQ3_XXS.gguf) | i1-IQ3_XXS | 18.6 | lower quality |
56
+ | [GGUF](https://huggingface.co/mradermacher/Llama-4-Scout-17B-6E-Instruct-i1-GGUF/resolve/main/Llama-4-Scout-17B-6E-Instruct.i1-IQ3_XS.gguf) | i1-IQ3_XS | 19.9 | |
57
+ | [GGUF](https://huggingface.co/mradermacher/Llama-4-Scout-17B-6E-Instruct-i1-GGUF/resolve/main/Llama-4-Scout-17B-6E-Instruct.i1-Q3_K_S.gguf) | i1-Q3_K_S | 20.9 | IQ3_XS probably better |
58
+ | [GGUF](https://huggingface.co/mradermacher/Llama-4-Scout-17B-6E-Instruct-i1-GGUF/resolve/main/Llama-4-Scout-17B-6E-Instruct.i1-IQ3_S.gguf) | i1-IQ3_S | 20.9 | beats Q3_K* |
59
+ | [GGUF](https://huggingface.co/mradermacher/Llama-4-Scout-17B-6E-Instruct-i1-GGUF/resolve/main/Llama-4-Scout-17B-6E-Instruct.i1-IQ3_M.gguf) | i1-IQ3_M | 21.3 | |
60
+ | [GGUF](https://huggingface.co/mradermacher/Llama-4-Scout-17B-6E-Instruct-i1-GGUF/resolve/main/Llama-4-Scout-17B-6E-Instruct.i1-Q3_K_M.gguf) | i1-Q3_K_M | 23.1 | IQ3_S probably better |
61
+ | [GGUF](https://huggingface.co/mradermacher/Llama-4-Scout-17B-6E-Instruct-i1-GGUF/resolve/main/Llama-4-Scout-17B-6E-Instruct.i1-Q3_K_L.gguf) | i1-Q3_K_L | 24.9 | IQ3_M probably better |
62
+ | [GGUF](https://huggingface.co/mradermacher/Llama-4-Scout-17B-6E-Instruct-i1-GGUF/resolve/main/Llama-4-Scout-17B-6E-Instruct.i1-IQ4_XS.gguf) | i1-IQ4_XS | 25.6 | |
63
+ | [GGUF](https://huggingface.co/mradermacher/Llama-4-Scout-17B-6E-Instruct-i1-GGUF/resolve/main/Llama-4-Scout-17B-6E-Instruct.i1-Q4_0.gguf) | i1-Q4_0 | 27.1 | fast, low quality |
64
+ | [GGUF](https://huggingface.co/mradermacher/Llama-4-Scout-17B-6E-Instruct-i1-GGUF/resolve/main/Llama-4-Scout-17B-6E-Instruct.i1-Q4_K_S.gguf) | i1-Q4_K_S | 27.3 | optimal size/speed/quality |
65
+ | [GGUF](https://huggingface.co/mradermacher/Llama-4-Scout-17B-6E-Instruct-i1-GGUF/resolve/main/Llama-4-Scout-17B-6E-Instruct.i1-Q4_K_M.gguf) | i1-Q4_K_M | 28.9 | fast, recommended |
66
+ | [GGUF](https://huggingface.co/mradermacher/Llama-4-Scout-17B-6E-Instruct-i1-GGUF/resolve/main/Llama-4-Scout-17B-6E-Instruct.i1-Q4_1.gguf) | i1-Q4_1 | 29.9 | |
67
+ | [GGUF](https://huggingface.co/mradermacher/Llama-4-Scout-17B-6E-Instruct-i1-GGUF/resolve/main/Llama-4-Scout-17B-6E-Instruct.i1-Q5_K_S.gguf) | i1-Q5_K_S | 32.8 | |
68
+ | [GGUF](https://huggingface.co/mradermacher/Llama-4-Scout-17B-6E-Instruct-i1-GGUF/resolve/main/Llama-4-Scout-17B-6E-Instruct.i1-Q5_K_M.gguf) | i1-Q5_K_M | 33.8 | |
69
+ | [GGUF](https://huggingface.co/mradermacher/Llama-4-Scout-17B-6E-Instruct-i1-GGUF/resolve/main/Llama-4-Scout-17B-6E-Instruct.i1-Q6_K.gguf) | i1-Q6_K | 39.0 | practically like static Q6_K |
70
+
71
+ Here is a handy graph by ikawrakow comparing some lower-quality quant
72
+ types (lower is better):
73
+
74
+ ![image.png](https://www.nethype.de/huggingface_embed/quantpplgraph.png)
75
+
76
+ And here are Artefact2's thoughts on the matter:
77
+ https://gist.github.com/Artefact2/b5f810600771265fc1e39442288e8ec9
78
+
79
+ ## FAQ / Model Request
80
+
81
+ See https://huggingface.co/mradermacher/model_requests for some answers to
82
+ questions you might have and/or if you want some other model quantized.
83
+
84
+ ## Thanks
85
+
86
+ I thank my company, [nethype GmbH](https://www.nethype.de/), for letting
87
+ me use its servers and providing upgrades to my workstation to enable
88
+ this work in my free time. Additional thanks to [@nicoboss](https://huggingface.co/nicoboss) for giving me access to his private supercomputer, enabling me to provide many more imatrix quants, at much higher quality, than I would otherwise be able to.
89
+
90
+ <!-- end -->