mradermacher commited on
Commit
0c18580
1 Parent(s): 5d77048

auto-patch README.md

Browse files
Files changed (1) hide show
  1. README.md +67 -0
README.md CHANGED
@@ -1,6 +1,73 @@
 
 
 
 
 
 
 
 
 
 
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/leafspark/Reflection-Llama-3.1-70B-bf16
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ base_model: leafspark/Reflection-Llama-3.1-70B-bf16
3
+ language:
4
+ - en
5
+ library_name: transformers
6
+ license: llama3.1
7
+ quantized_by: mradermacher
8
+ ---
9
+ ## About
10
+
11
  <!-- ### quantize_version: 2 -->
12
  <!-- ### output_tensor_quantised: 1 -->
13
  <!-- ### convert_type: hf -->
14
  <!-- ### vocab_type: -->
15
  <!-- ### tags: nicoboss -->
16
  weighted/imatrix quants of https://huggingface.co/leafspark/Reflection-Llama-3.1-70B-bf16
17
+
18
+ <!-- provided-files -->
19
+ static quants are available at https://huggingface.co/mradermacher/Reflection-Llama-3.1-70B-bf16-GGUF
20
+ ## Usage
21
+
22
+ If you are unsure how to use GGUF files, refer to one of [TheBloke's
23
+ READMEs](https://huggingface.co/TheBloke/KafkaLM-70B-German-V0.1-GGUF) for
24
+ more details, including on how to concatenate multi-part files.
25
+
26
+ ## Provided Quants
27
+
28
+ (sorted by size, not necessarily quality. IQ-quants are often preferable over similar sized non-IQ quants)
29
+
30
+ | Link | Type | Size/GB | Notes |
31
+ |:-----|:-----|--------:|:------|
32
+ | [GGUF](https://huggingface.co/mradermacher/Reflection-Llama-3.1-70B-bf16-i1-GGUF/resolve/main/Reflection-Llama-3.1-70B-bf16.i1-IQ1_S.gguf) | i1-IQ1_S | 15.4 | for the desperate |
33
+ | [GGUF](https://huggingface.co/mradermacher/Reflection-Llama-3.1-70B-bf16-i1-GGUF/resolve/main/Reflection-Llama-3.1-70B-bf16.i1-IQ1_M.gguf) | i1-IQ1_M | 16.9 | mostly desperate |
34
+ | [GGUF](https://huggingface.co/mradermacher/Reflection-Llama-3.1-70B-bf16-i1-GGUF/resolve/main/Reflection-Llama-3.1-70B-bf16.i1-IQ2_XXS.gguf) | i1-IQ2_XXS | 19.2 | |
35
+ | [GGUF](https://huggingface.co/mradermacher/Reflection-Llama-3.1-70B-bf16-i1-GGUF/resolve/main/Reflection-Llama-3.1-70B-bf16.i1-IQ2_XS.gguf) | i1-IQ2_XS | 21.2 | |
36
+ | [GGUF](https://huggingface.co/mradermacher/Reflection-Llama-3.1-70B-bf16-i1-GGUF/resolve/main/Reflection-Llama-3.1-70B-bf16.i1-IQ2_S.gguf) | i1-IQ2_S | 22.3 | |
37
+ | [GGUF](https://huggingface.co/mradermacher/Reflection-Llama-3.1-70B-bf16-i1-GGUF/resolve/main/Reflection-Llama-3.1-70B-bf16.i1-IQ2_M.gguf) | i1-IQ2_M | 24.2 | |
38
+ | [GGUF](https://huggingface.co/mradermacher/Reflection-Llama-3.1-70B-bf16-i1-GGUF/resolve/main/Reflection-Llama-3.1-70B-bf16.i1-Q2_K.gguf) | i1-Q2_K | 26.5 | IQ3_XXS probably better |
39
+ | [GGUF](https://huggingface.co/mradermacher/Reflection-Llama-3.1-70B-bf16-i1-GGUF/resolve/main/Reflection-Llama-3.1-70B-bf16.i1-IQ3_XXS.gguf) | i1-IQ3_XXS | 27.6 | lower quality |
40
+ | [GGUF](https://huggingface.co/mradermacher/Reflection-Llama-3.1-70B-bf16-i1-GGUF/resolve/main/Reflection-Llama-3.1-70B-bf16.i1-IQ3_XS.gguf) | i1-IQ3_XS | 29.4 | |
41
+ | [GGUF](https://huggingface.co/mradermacher/Reflection-Llama-3.1-70B-bf16-i1-GGUF/resolve/main/Reflection-Llama-3.1-70B-bf16.i1-IQ3_S.gguf) | i1-IQ3_S | 31.0 | beats Q3_K* |
42
+ | [GGUF](https://huggingface.co/mradermacher/Reflection-Llama-3.1-70B-bf16-i1-GGUF/resolve/main/Reflection-Llama-3.1-70B-bf16.i1-Q3_K_S.gguf) | i1-Q3_K_S | 31.0 | IQ3_XS probably better |
43
+ | [GGUF](https://huggingface.co/mradermacher/Reflection-Llama-3.1-70B-bf16-i1-GGUF/resolve/main/Reflection-Llama-3.1-70B-bf16.i1-IQ3_M.gguf) | i1-IQ3_M | 32.0 | |
44
+ | [GGUF](https://huggingface.co/mradermacher/Reflection-Llama-3.1-70B-bf16-i1-GGUF/resolve/main/Reflection-Llama-3.1-70B-bf16.i1-Q3_K_M.gguf) | i1-Q3_K_M | 34.4 | IQ3_S probably better |
45
+ | [GGUF](https://huggingface.co/mradermacher/Reflection-Llama-3.1-70B-bf16-i1-GGUF/resolve/main/Reflection-Llama-3.1-70B-bf16.i1-Q3_K_L.gguf) | i1-Q3_K_L | 37.2 | IQ3_M probably better |
46
+ | [GGUF](https://huggingface.co/mradermacher/Reflection-Llama-3.1-70B-bf16-i1-GGUF/resolve/main/Reflection-Llama-3.1-70B-bf16.i1-IQ4_XS.gguf) | i1-IQ4_XS | 38.0 | |
47
+ | [GGUF](https://huggingface.co/mradermacher/Reflection-Llama-3.1-70B-bf16-i1-GGUF/resolve/main/Reflection-Llama-3.1-70B-bf16.i1-Q4_0.gguf) | i1-Q4_0 | 40.2 | fast, low quality |
48
+ | [GGUF](https://huggingface.co/mradermacher/Reflection-Llama-3.1-70B-bf16-i1-GGUF/resolve/main/Reflection-Llama-3.1-70B-bf16.i1-Q4_K_S.gguf) | i1-Q4_K_S | 40.4 | optimal size/speed/quality |
49
+ | [GGUF](https://huggingface.co/mradermacher/Reflection-Llama-3.1-70B-bf16-i1-GGUF/resolve/main/Reflection-Llama-3.1-70B-bf16.i1-Q4_K_M.gguf) | i1-Q4_K_M | 42.6 | fast, recommended |
50
+ | [GGUF](https://huggingface.co/mradermacher/Reflection-Llama-3.1-70B-bf16-i1-GGUF/resolve/main/Reflection-Llama-3.1-70B-bf16.i1-Q5_K_S.gguf) | i1-Q5_K_S | 48.8 | |
51
+ | [GGUF](https://huggingface.co/mradermacher/Reflection-Llama-3.1-70B-bf16-i1-GGUF/resolve/main/Reflection-Llama-3.1-70B-bf16.i1-Q5_K_M.gguf) | i1-Q5_K_M | 50.0 | |
52
+ | [PART 1](https://huggingface.co/mradermacher/Reflection-Llama-3.1-70B-bf16-i1-GGUF/resolve/main/Reflection-Llama-3.1-70B-bf16.i1-Q6_K.gguf.part1of2) [PART 2](https://huggingface.co/mradermacher/Reflection-Llama-3.1-70B-bf16-i1-GGUF/resolve/main/Reflection-Llama-3.1-70B-bf16.i1-Q6_K.gguf.part2of2) | i1-Q6_K | 58.0 | practically like static Q6_K |
53
+
54
+ Here is a handy graph by ikawrakow comparing some lower-quality quant
55
+ types (lower is better):
56
+
57
+ ![image.png](https://www.nethype.de/huggingface_embed/quantpplgraph.png)
58
+
59
+ And here are Artefact2's thoughts on the matter:
60
+ https://gist.github.com/Artefact2/b5f810600771265fc1e39442288e8ec9
61
+
62
+ ## FAQ / Model Request
63
+
64
+ See https://huggingface.co/mradermacher/model_requests for some answers to
65
+ questions you might have and/or if you want some other model quantized.
66
+
67
+ ## Thanks
68
+
69
+ I thank my company, [nethype GmbH](https://www.nethype.de/), for letting
70
+ me use its servers and providing upgrades to my workstation to enable
71
+ 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.
72
+
73
+ <!-- end -->