mradermacher commited on
Commit
b7ddc99
1 Parent(s): cd3e3ab

auto-patch README.md

Browse files
Files changed (1) hide show
  1. README.md +69 -0
README.md CHANGED
@@ -1,6 +1,75 @@
 
 
 
 
 
 
 
 
 
 
 
 
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/codellama/CodeLlama-34b-Python-hf
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ base_model: codellama/CodeLlama-34b-Python-hf
3
+ language:
4
+ - code
5
+ library_name: transformers
6
+ license: llama2
7
+ quantized_by: mradermacher
8
+ tags:
9
+ - llama-2
10
+ ---
11
+ ## About
12
+
13
  <!-- ### quantize_version: 2 -->
14
  <!-- ### output_tensor_quantised: 1 -->
15
  <!-- ### convert_type: hf -->
16
  <!-- ### vocab_type: -->
17
  <!-- ### tags: nicoboss -->
18
  weighted/imatrix quants of https://huggingface.co/codellama/CodeLlama-34b-Python-hf
19
+
20
+ <!-- provided-files -->
21
+ static quants are available at https://huggingface.co/mradermacher/CodeLlama-34b-Python-hf-GGUF
22
+ ## Usage
23
+
24
+ If you are unsure how to use GGUF files, refer to one of [TheBloke's
25
+ READMEs](https://huggingface.co/TheBloke/KafkaLM-70B-German-V0.1-GGUF) for
26
+ more details, including on how to concatenate multi-part files.
27
+
28
+ ## Provided Quants
29
+
30
+ (sorted by size, not necessarily quality. IQ-quants are often preferable over similar sized non-IQ quants)
31
+
32
+ | Link | Type | Size/GB | Notes |
33
+ |:-----|:-----|--------:|:------|
34
+ | [GGUF](https://huggingface.co/mradermacher/CodeLlama-34b-Python-hf-i1-GGUF/resolve/main/CodeLlama-34b-Python-hf.i1-IQ1_S.gguf) | i1-IQ1_S | 7.4 | for the desperate |
35
+ | [GGUF](https://huggingface.co/mradermacher/CodeLlama-34b-Python-hf-i1-GGUF/resolve/main/CodeLlama-34b-Python-hf.i1-IQ1_M.gguf) | i1-IQ1_M | 8.0 | mostly desperate |
36
+ | [GGUF](https://huggingface.co/mradermacher/CodeLlama-34b-Python-hf-i1-GGUF/resolve/main/CodeLlama-34b-Python-hf.i1-IQ2_XXS.gguf) | i1-IQ2_XXS | 9.1 | |
37
+ | [GGUF](https://huggingface.co/mradermacher/CodeLlama-34b-Python-hf-i1-GGUF/resolve/main/CodeLlama-34b-Python-hf.i1-IQ2_XS.gguf) | i1-IQ2_XS | 10.1 | |
38
+ | [GGUF](https://huggingface.co/mradermacher/CodeLlama-34b-Python-hf-i1-GGUF/resolve/main/CodeLlama-34b-Python-hf.i1-IQ2_S.gguf) | i1-IQ2_S | 10.7 | |
39
+ | [GGUF](https://huggingface.co/mradermacher/CodeLlama-34b-Python-hf-i1-GGUF/resolve/main/CodeLlama-34b-Python-hf.i1-IQ2_M.gguf) | i1-IQ2_M | 11.6 | |
40
+ | [GGUF](https://huggingface.co/mradermacher/CodeLlama-34b-Python-hf-i1-GGUF/resolve/main/CodeLlama-34b-Python-hf.i1-Q2_K.gguf) | i1-Q2_K | 12.6 | IQ3_XXS probably better |
41
+ | [GGUF](https://huggingface.co/mradermacher/CodeLlama-34b-Python-hf-i1-GGUF/resolve/main/CodeLlama-34b-Python-hf.i1-IQ3_XXS.gguf) | i1-IQ3_XXS | 13.1 | lower quality |
42
+ | [GGUF](https://huggingface.co/mradermacher/CodeLlama-34b-Python-hf-i1-GGUF/resolve/main/CodeLlama-34b-Python-hf.i1-IQ3_XS.gguf) | i1-IQ3_XS | 14.0 | |
43
+ | [GGUF](https://huggingface.co/mradermacher/CodeLlama-34b-Python-hf-i1-GGUF/resolve/main/CodeLlama-34b-Python-hf.i1-Q3_K_S.gguf) | i1-Q3_K_S | 14.7 | IQ3_XS probably better |
44
+ | [GGUF](https://huggingface.co/mradermacher/CodeLlama-34b-Python-hf-i1-GGUF/resolve/main/CodeLlama-34b-Python-hf.i1-IQ3_S.gguf) | i1-IQ3_S | 14.8 | beats Q3_K* |
45
+ | [GGUF](https://huggingface.co/mradermacher/CodeLlama-34b-Python-hf-i1-GGUF/resolve/main/CodeLlama-34b-Python-hf.i1-IQ3_M.gguf) | i1-IQ3_M | 15.3 | |
46
+ | [GGUF](https://huggingface.co/mradermacher/CodeLlama-34b-Python-hf-i1-GGUF/resolve/main/CodeLlama-34b-Python-hf.i1-Q3_K_M.gguf) | i1-Q3_K_M | 16.4 | IQ3_S probably better |
47
+ | [GGUF](https://huggingface.co/mradermacher/CodeLlama-34b-Python-hf-i1-GGUF/resolve/main/CodeLlama-34b-Python-hf.i1-Q3_K_L.gguf) | i1-Q3_K_L | 17.9 | IQ3_M probably better |
48
+ | [GGUF](https://huggingface.co/mradermacher/CodeLlama-34b-Python-hf-i1-GGUF/resolve/main/CodeLlama-34b-Python-hf.i1-IQ4_XS.gguf) | i1-IQ4_XS | 18.2 | |
49
+ | [GGUF](https://huggingface.co/mradermacher/CodeLlama-34b-Python-hf-i1-GGUF/resolve/main/CodeLlama-34b-Python-hf.i1-Q4_0.gguf) | i1-Q4_0 | 19.2 | fast, low quality |
50
+ | [GGUF](https://huggingface.co/mradermacher/CodeLlama-34b-Python-hf-i1-GGUF/resolve/main/CodeLlama-34b-Python-hf.i1-Q4_K_S.gguf) | i1-Q4_K_S | 19.3 | optimal size/speed/quality |
51
+ | [GGUF](https://huggingface.co/mradermacher/CodeLlama-34b-Python-hf-i1-GGUF/resolve/main/CodeLlama-34b-Python-hf.i1-Q4_K_M.gguf) | i1-Q4_K_M | 20.3 | fast, recommended |
52
+ | [GGUF](https://huggingface.co/mradermacher/CodeLlama-34b-Python-hf-i1-GGUF/resolve/main/CodeLlama-34b-Python-hf.i1-Q5_K_S.gguf) | i1-Q5_K_S | 23.3 | |
53
+ | [GGUF](https://huggingface.co/mradermacher/CodeLlama-34b-Python-hf-i1-GGUF/resolve/main/CodeLlama-34b-Python-hf.i1-Q5_K_M.gguf) | i1-Q5_K_M | 23.9 | |
54
+ | [GGUF](https://huggingface.co/mradermacher/CodeLlama-34b-Python-hf-i1-GGUF/resolve/main/CodeLlama-34b-Python-hf.i1-Q6_K.gguf) | i1-Q6_K | 27.8 | practically like static Q6_K |
55
+
56
+ Here is a handy graph by ikawrakow comparing some lower-quality quant
57
+ types (lower is better):
58
+
59
+ ![image.png](https://www.nethype.de/huggingface_embed/quantpplgraph.png)
60
+
61
+ And here are Artefact2's thoughts on the matter:
62
+ https://gist.github.com/Artefact2/b5f810600771265fc1e39442288e8ec9
63
+
64
+ ## FAQ / Model Request
65
+
66
+ See https://huggingface.co/mradermacher/model_requests for some answers to
67
+ questions you might have and/or if you want some other model quantized.
68
+
69
+ ## Thanks
70
+
71
+ I thank my company, [nethype GmbH](https://www.nethype.de/), for letting
72
+ me use its servers and providing upgrades to my workstation to enable
73
+ 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.
74
+
75
+ <!-- end -->