mradermacher commited on
Commit
b5cfebd
1 Parent(s): 98fbc85

auto-patch README.md

Browse files
Files changed (1) hide show
  1. README.md +62 -0
README.md CHANGED
@@ -1,5 +1,67 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  <!-- ### quantize_version: 1 -->
2
  <!-- ### output_tensor_quantised: 1 -->
3
  <!-- ### convert_type: -->
4
  <!-- ### vocab_type: -->
5
  static quants of https://huggingface.co/ThomasComics/Phi-3-mini-128k-instruct-LLaMAfied
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ exported_from: ThomasComics/Phi-3-mini-128k-instruct-LLaMAfied
3
+ language:
4
+ - en
5
+ library_name: transformers
6
+ license: mit
7
+ license_link: https://huggingface.co/microsoft/Phi-3-mini-4k-instruct/resolve/main/LICENSE
8
+ quantized_by: mradermacher
9
+ tags:
10
+ - nlp
11
+ - code
12
+ ---
13
+ ## About
14
+
15
  <!-- ### quantize_version: 1 -->
16
  <!-- ### output_tensor_quantised: 1 -->
17
  <!-- ### convert_type: -->
18
  <!-- ### vocab_type: -->
19
  static quants of https://huggingface.co/ThomasComics/Phi-3-mini-128k-instruct-LLaMAfied
20
+
21
+
22
+ <!-- provided-files -->
23
+ weighted/imatrix quants seem not to be available (by me) at this time. If they do not show up a week or so after the static ones, I have probably not planned for them. Feel free to request them by opening a Community Discussion.
24
+ ## Usage
25
+
26
+ If you are unsure how to use GGUF files, refer to one of [TheBloke's
27
+ READMEs](https://huggingface.co/TheBloke/KafkaLM-70B-German-V0.1-GGUF) for
28
+ more details, including on how to concatenate multi-part files.
29
+
30
+ ## Provided Quants
31
+
32
+ (sorted by size, not necessarily quality. IQ-quants are often preferable over similar sized non-IQ quants)
33
+
34
+ | Link | Type | Size/GB | Notes |
35
+ |:-----|:-----|--------:|:------|
36
+ | [GGUF](https://huggingface.co/mradermacher/Phi-3-mini-128k-instruct-LLaMAfied-GGUF/resolve/main/Phi-3-mini-128k-instruct-LLaMAfied.Q2_K.gguf) | Q2_K | 1.5 | |
37
+ | [GGUF](https://huggingface.co/mradermacher/Phi-3-mini-128k-instruct-LLaMAfied-GGUF/resolve/main/Phi-3-mini-128k-instruct-LLaMAfied.IQ3_XS.gguf) | IQ3_XS | 1.7 | |
38
+ | [GGUF](https://huggingface.co/mradermacher/Phi-3-mini-128k-instruct-LLaMAfied-GGUF/resolve/main/Phi-3-mini-128k-instruct-LLaMAfied.IQ3_S.gguf) | IQ3_S | 1.8 | beats Q3_K* |
39
+ | [GGUF](https://huggingface.co/mradermacher/Phi-3-mini-128k-instruct-LLaMAfied-GGUF/resolve/main/Phi-3-mini-128k-instruct-LLaMAfied.Q3_K_S.gguf) | Q3_K_S | 1.8 | |
40
+ | [GGUF](https://huggingface.co/mradermacher/Phi-3-mini-128k-instruct-LLaMAfied-GGUF/resolve/main/Phi-3-mini-128k-instruct-LLaMAfied.IQ3_M.gguf) | IQ3_M | 1.9 | |
41
+ | [GGUF](https://huggingface.co/mradermacher/Phi-3-mini-128k-instruct-LLaMAfied-GGUF/resolve/main/Phi-3-mini-128k-instruct-LLaMAfied.Q3_K_M.gguf) | Q3_K_M | 2.0 | lower quality |
42
+ | [GGUF](https://huggingface.co/mradermacher/Phi-3-mini-128k-instruct-LLaMAfied-GGUF/resolve/main/Phi-3-mini-128k-instruct-LLaMAfied.Q3_K_L.gguf) | Q3_K_L | 2.1 | |
43
+ | [GGUF](https://huggingface.co/mradermacher/Phi-3-mini-128k-instruct-LLaMAfied-GGUF/resolve/main/Phi-3-mini-128k-instruct-LLaMAfied.IQ4_XS.gguf) | IQ4_XS | 2.2 | |
44
+ | [GGUF](https://huggingface.co/mradermacher/Phi-3-mini-128k-instruct-LLaMAfied-GGUF/resolve/main/Phi-3-mini-128k-instruct-LLaMAfied.Q4_K_S.gguf) | Q4_K_S | 2.3 | fast, recommended |
45
+ | [GGUF](https://huggingface.co/mradermacher/Phi-3-mini-128k-instruct-LLaMAfied-GGUF/resolve/main/Phi-3-mini-128k-instruct-LLaMAfied.Q4_K_M.gguf) | Q4_K_M | 2.4 | fast, recommended |
46
+ | [GGUF](https://huggingface.co/mradermacher/Phi-3-mini-128k-instruct-LLaMAfied-GGUF/resolve/main/Phi-3-mini-128k-instruct-LLaMAfied.Q5_K_S.gguf) | Q5_K_S | 2.7 | |
47
+ | [GGUF](https://huggingface.co/mradermacher/Phi-3-mini-128k-instruct-LLaMAfied-GGUF/resolve/main/Phi-3-mini-128k-instruct-LLaMAfied.Q5_K_M.gguf) | Q5_K_M | 2.8 | |
48
+ | [GGUF](https://huggingface.co/mradermacher/Phi-3-mini-128k-instruct-LLaMAfied-GGUF/resolve/main/Phi-3-mini-128k-instruct-LLaMAfied.Q6_K.gguf) | Q6_K | 3.2 | very good quality |
49
+ | [GGUF](https://huggingface.co/mradermacher/Phi-3-mini-128k-instruct-LLaMAfied-GGUF/resolve/main/Phi-3-mini-128k-instruct-LLaMAfied.Q8_0.gguf) | Q8_0 | 4.2 | fast, best quality |
50
+ | [GGUF](https://huggingface.co/mradermacher/Phi-3-mini-128k-instruct-LLaMAfied-GGUF/resolve/main/Phi-3-mini-128k-instruct-LLaMAfied.f16.gguf) | f16 | 7.7 | 16 bpw, overkill |
51
+
52
+
53
+ Here is a handy graph by ikawrakow comparing some lower-quality quant
54
+ types (lower is better):
55
+
56
+ ![image.png](https://www.nethype.de/huggingface_embed/quantpplgraph.png)
57
+
58
+ And here are Artefact2's thoughts on the matter:
59
+ https://gist.github.com/Artefact2/b5f810600771265fc1e39442288e8ec9
60
+
61
+ ## Thanks
62
+
63
+ I thank my company, [nethype GmbH](https://www.nethype.de/), for letting
64
+ me use its servers and providing upgrades to my workstation to enable
65
+ this work in my free time.
66
+
67
+ <!-- end -->