mradermacher commited on
Commit
ee7ed47
1 Parent(s): ba8ead0

auto-patch README.md

Browse files
Files changed (1) hide show
  1. README.md +76 -0
README.md CHANGED
@@ -1,3 +1,79 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  <!-- ### convert_type: -->
2
  <!-- ### vocab_type: -->
3
  weighted/imatrix quants of https://huggingface.co/jondurbin/airoboros-34b-3.3
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ base_model: 01-ai/yi-34b-200k
3
+ datasets:
4
+ - jondurbin/airoboros-3.2
5
+ - bluemoon-fandom-1-1-rp-cleaned
6
+ - boolq
7
+ - jondurbin/gutenberg-dpo-v0.1
8
+ - LDJnr/Capybara
9
+ - jondurbin/cinematika-v0.1
10
+ - glaiveai/glaive-function-calling-v2
11
+ - grimulkan/LimaRP-augmented
12
+ - piqa
13
+ - Vezora/Tested-22k-Python-Alpaca
14
+ - mattpscott/airoboros-summarization
15
+ - unalignment/toxic-dpo-v0.2
16
+ exported_from: jondurbin/airoboros-34b-3.3
17
+ language:
18
+ - en
19
+ library_name: transformers
20
+ license: other
21
+ license_link: https://huggingface.co/01-ai/Yi-34B-200K/blob/main/LICENSE
22
+ license_name: yi-license
23
+ quantized_by: mradermacher
24
+ ---
25
+ ## About
26
+
27
  <!-- ### convert_type: -->
28
  <!-- ### vocab_type: -->
29
  weighted/imatrix quants of https://huggingface.co/jondurbin/airoboros-34b-3.3
30
+
31
+
32
+ <!-- provided-files -->
33
+ static quants are available at https://huggingface.co/mradermacher/airoboros-34b-3.3-GGUF
34
+ ## Usage
35
+
36
+ If you are unsure how to use GGUF files, refer to one of [TheBloke's
37
+ READMEs](https://huggingface.co/TheBloke/KafkaLM-70B-German-V0.1-GGUF) for
38
+ more details, including on how to concatenate multi-part files.
39
+
40
+ ## Provided Quants
41
+
42
+ (sorted by size, not necessarily quality. IQ-quants are often preferable over similar sized non-IQ quants)
43
+
44
+ | Link | Type | Size/GB | Notes |
45
+ |:-----|:-----|--------:|:------|
46
+ | [GGUF](https://huggingface.co/mradermacher/airoboros-34b-3.3-i1-GGUF/resolve/main/airoboros-34b-3.3.i1-IQ2_XXS.gguf) | i1-IQ2_XXS | 10.0 | |
47
+ | [GGUF](https://huggingface.co/mradermacher/airoboros-34b-3.3-i1-GGUF/resolve/main/airoboros-34b-3.3.i1-IQ2_M.gguf) | i1-IQ2_M | 12.5 | |
48
+ | [GGUF](https://huggingface.co/mradermacher/airoboros-34b-3.3-i1-GGUF/resolve/main/airoboros-34b-3.3.i1-Q2_K.gguf) | i1-Q2_K | 13.5 | IQ3_XXS probably better |
49
+ | [GGUF](https://huggingface.co/mradermacher/airoboros-34b-3.3-i1-GGUF/resolve/main/airoboros-34b-3.3.i1-IQ3_XXS.gguf) | i1-IQ3_XXS | 14.0 | lower quality |
50
+ | [GGUF](https://huggingface.co/mradermacher/airoboros-34b-3.3-i1-GGUF/resolve/main/airoboros-34b-3.3.i1-IQ3_XS.gguf) | i1-IQ3_XS | 14.9 | |
51
+ | [GGUF](https://huggingface.co/mradermacher/airoboros-34b-3.3-i1-GGUF/resolve/main/airoboros-34b-3.3.i1-Q3_K_S.gguf) | i1-Q3_K_S | 15.6 | IQ3_XS probably better |
52
+ | [GGUF](https://huggingface.co/mradermacher/airoboros-34b-3.3-i1-GGUF/resolve/main/airoboros-34b-3.3.i1-IQ3_S.gguf) | i1-IQ3_S | 15.7 | beats Q3_K* |
53
+ | [GGUF](https://huggingface.co/mradermacher/airoboros-34b-3.3-i1-GGUF/resolve/main/airoboros-34b-3.3.i1-IQ3_M.gguf) | i1-IQ3_M | 16.2 | |
54
+ | [GGUF](https://huggingface.co/mradermacher/airoboros-34b-3.3-i1-GGUF/resolve/main/airoboros-34b-3.3.i1-Q3_K_M.gguf) | i1-Q3_K_M | 17.3 | IQ3_S probably better |
55
+ | [GGUF](https://huggingface.co/mradermacher/airoboros-34b-3.3-i1-GGUF/resolve/main/airoboros-34b-3.3.i1-Q3_K_L.gguf) | i1-Q3_K_L | 18.8 | IQ3_M probably better |
56
+ | [GGUF](https://huggingface.co/mradermacher/airoboros-34b-3.3-i1-GGUF/resolve/main/airoboros-34b-3.3.i1-IQ4_XS.gguf) | i1-IQ4_XS | 19.1 | |
57
+ | [GGUF](https://huggingface.co/mradermacher/airoboros-34b-3.3-i1-GGUF/resolve/main/airoboros-34b-3.3.i1-Q4_0.gguf) | i1-Q4_0 | 20.2 | fast, low quality |
58
+ | [GGUF](https://huggingface.co/mradermacher/airoboros-34b-3.3-i1-GGUF/resolve/main/airoboros-34b-3.3.i1-Q4_K_S.gguf) | i1-Q4_K_S | 20.2 | optimal size/speed/quality |
59
+ | [GGUF](https://huggingface.co/mradermacher/airoboros-34b-3.3-i1-GGUF/resolve/main/airoboros-34b-3.3.i1-Q4_K_M.gguf) | i1-Q4_K_M | 21.3 | fast, recommended |
60
+ | [GGUF](https://huggingface.co/mradermacher/airoboros-34b-3.3-i1-GGUF/resolve/main/airoboros-34b-3.3.i1-Q5_K_S.gguf) | i1-Q5_K_S | 24.3 | |
61
+ | [GGUF](https://huggingface.co/mradermacher/airoboros-34b-3.3-i1-GGUF/resolve/main/airoboros-34b-3.3.i1-Q5_K_M.gguf) | i1-Q5_K_M | 25.0 | |
62
+ | [GGUF](https://huggingface.co/mradermacher/airoboros-34b-3.3-i1-GGUF/resolve/main/airoboros-34b-3.3.i1-Q6_K.gguf) | i1-Q6_K | 28.9 | practically like static Q6_K |
63
+
64
+
65
+ Here is a handy graph by ikawrakow comparing some lower-quality quant
66
+ types (lower is better):
67
+
68
+ ![image.png](https://www.nethype.de/huggingface_embed/quantpplgraph.png)
69
+
70
+ And here are Artefact2's thoughts on the matter:
71
+ https://gist.github.com/Artefact2/b5f810600771265fc1e39442288e8ec9
72
+
73
+ ## Thanks
74
+
75
+ I thank my company, [nethype GmbH](https://www.nethype.de/), for letting
76
+ me use its servers and providing upgrades to my workstation to enable
77
+ this work in my free time.
78
+
79
+ <!-- end -->