mradermacher commited on
Commit
dda0746
1 Parent(s): 834f0c5

auto-patch README.md

Browse files
Files changed (1) hide show
  1. README.md +73 -0
README.md CHANGED
@@ -1,6 +1,79 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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/Azure99/blossom-v5-32b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ base_model: Azure99/blossom-v5-32b
3
+ datasets:
4
+ - Azure99/blossom-chat-v3
5
+ - Azure99/blossom-math-v4
6
+ - Azure99/blossom-wizard-v3
7
+ - Azure99/blossom-orca-v3
8
+ language:
9
+ - zh
10
+ - en
11
+ library_name: transformers
12
+ license: apache-2.0
13
+ quantized_by: mradermacher
14
+ ---
15
+ ## About
16
+
17
  <!-- ### quantize_version: 2 -->
18
  <!-- ### output_tensor_quantised: 1 -->
19
  <!-- ### convert_type: hf -->
20
  <!-- ### vocab_type: -->
21
  <!-- ### tags: nicoboss -->
22
  weighted/imatrix quants of https://huggingface.co/Azure99/blossom-v5-32b
23
+
24
+ <!-- provided-files -->
25
+ static quants are available at https://huggingface.co/mradermacher/blossom-v5-32b-GGUF
26
+ ## Usage
27
+
28
+ If you are unsure how to use GGUF files, refer to one of [TheBloke's
29
+ READMEs](https://huggingface.co/TheBloke/KafkaLM-70B-German-V0.1-GGUF) for
30
+ more details, including on how to concatenate multi-part files.
31
+
32
+ ## Provided Quants
33
+
34
+ (sorted by size, not necessarily quality. IQ-quants are often preferable over similar sized non-IQ quants)
35
+
36
+ | Link | Type | Size/GB | Notes |
37
+ |:-----|:-----|--------:|:------|
38
+ | [GGUF](https://huggingface.co/mradermacher/blossom-v5-32b-i1-GGUF/resolve/main/blossom-v5-32b.i1-IQ1_S.gguf) | i1-IQ1_S | 7.3 | for the desperate |
39
+ | [GGUF](https://huggingface.co/mradermacher/blossom-v5-32b-i1-GGUF/resolve/main/blossom-v5-32b.i1-IQ1_M.gguf) | i1-IQ1_M | 8.0 | mostly desperate |
40
+ | [GGUF](https://huggingface.co/mradermacher/blossom-v5-32b-i1-GGUF/resolve/main/blossom-v5-32b.i1-IQ2_XXS.gguf) | i1-IQ2_XXS | 9.1 | |
41
+ | [GGUF](https://huggingface.co/mradermacher/blossom-v5-32b-i1-GGUF/resolve/main/blossom-v5-32b.i1-IQ2_XS.gguf) | i1-IQ2_XS | 10.0 | |
42
+ | [GGUF](https://huggingface.co/mradermacher/blossom-v5-32b-i1-GGUF/resolve/main/blossom-v5-32b.i1-IQ2_S.gguf) | i1-IQ2_S | 10.4 | |
43
+ | [GGUF](https://huggingface.co/mradermacher/blossom-v5-32b-i1-GGUF/resolve/main/blossom-v5-32b.i1-IQ2_M.gguf) | i1-IQ2_M | 11.3 | |
44
+ | [GGUF](https://huggingface.co/mradermacher/blossom-v5-32b-i1-GGUF/resolve/main/blossom-v5-32b.i1-Q2_K.gguf) | i1-Q2_K | 12.3 | IQ3_XXS probably better |
45
+ | [GGUF](https://huggingface.co/mradermacher/blossom-v5-32b-i1-GGUF/resolve/main/blossom-v5-32b.i1-IQ3_XXS.gguf) | i1-IQ3_XXS | 12.8 | lower quality |
46
+ | [GGUF](https://huggingface.co/mradermacher/blossom-v5-32b-i1-GGUF/resolve/main/blossom-v5-32b.i1-IQ3_XS.gguf) | i1-IQ3_XS | 13.7 | |
47
+ | [GGUF](https://huggingface.co/mradermacher/blossom-v5-32b-i1-GGUF/resolve/main/blossom-v5-32b.i1-Q3_K_S.gguf) | i1-Q3_K_S | 14.4 | IQ3_XS probably better |
48
+ | [GGUF](https://huggingface.co/mradermacher/blossom-v5-32b-i1-GGUF/resolve/main/blossom-v5-32b.i1-IQ3_S.gguf) | i1-IQ3_S | 14.4 | beats Q3_K* |
49
+ | [GGUF](https://huggingface.co/mradermacher/blossom-v5-32b-i1-GGUF/resolve/main/blossom-v5-32b.i1-IQ3_M.gguf) | i1-IQ3_M | 14.8 | |
50
+ | [GGUF](https://huggingface.co/mradermacher/blossom-v5-32b-i1-GGUF/resolve/main/blossom-v5-32b.i1-Q3_K_M.gguf) | i1-Q3_K_M | 15.9 | IQ3_S probably better |
51
+ | [GGUF](https://huggingface.co/mradermacher/blossom-v5-32b-i1-GGUF/resolve/main/blossom-v5-32b.i1-Q3_K_L.gguf) | i1-Q3_K_L | 17.2 | IQ3_M probably better |
52
+ | [GGUF](https://huggingface.co/mradermacher/blossom-v5-32b-i1-GGUF/resolve/main/blossom-v5-32b.i1-IQ4_XS.gguf) | i1-IQ4_XS | 17.7 | |
53
+ | [GGUF](https://huggingface.co/mradermacher/blossom-v5-32b-i1-GGUF/resolve/main/blossom-v5-32b.i1-Q4_0.gguf) | i1-Q4_0 | 18.7 | fast, low quality |
54
+ | [GGUF](https://huggingface.co/mradermacher/blossom-v5-32b-i1-GGUF/resolve/main/blossom-v5-32b.i1-Q4_K_S.gguf) | i1-Q4_K_S | 18.7 | optimal size/speed/quality |
55
+ | [GGUF](https://huggingface.co/mradermacher/blossom-v5-32b-i1-GGUF/resolve/main/blossom-v5-32b.i1-Q4_K_M.gguf) | i1-Q4_K_M | 19.8 | fast, recommended |
56
+ | [GGUF](https://huggingface.co/mradermacher/blossom-v5-32b-i1-GGUF/resolve/main/blossom-v5-32b.i1-Q5_K_S.gguf) | i1-Q5_K_S | 22.6 | |
57
+ | [GGUF](https://huggingface.co/mradermacher/blossom-v5-32b-i1-GGUF/resolve/main/blossom-v5-32b.i1-Q5_K_M.gguf) | i1-Q5_K_M | 23.2 | |
58
+ | [GGUF](https://huggingface.co/mradermacher/blossom-v5-32b-i1-GGUF/resolve/main/blossom-v5-32b.i1-Q6_K.gguf) | i1-Q6_K | 26.8 | practically like static Q6_K |
59
+
60
+ Here is a handy graph by ikawrakow comparing some lower-quality quant
61
+ types (lower is better):
62
+
63
+ ![image.png](https://www.nethype.de/huggingface_embed/quantpplgraph.png)
64
+
65
+ And here are Artefact2's thoughts on the matter:
66
+ https://gist.github.com/Artefact2/b5f810600771265fc1e39442288e8ec9
67
+
68
+ ## FAQ / Model Request
69
+
70
+ See https://huggingface.co/mradermacher/model_requests for some answers to
71
+ questions you might have and/or if you want some other model quantized.
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. Additional thanks to [@nicoboss](https://huggingface.co/nicoboss) for giving me access to his hardware for calculating the imatrix for these quants.
78
+
79
+ <!-- end -->