bartowski commited on
Commit
6107d40
1 Parent(s): a656910

measurement.json

Browse files
Files changed (2) hide show
  1. README.md +72 -0
  2. measurement.json +0 -0
README.md ADDED
@@ -0,0 +1,72 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ datasets:
4
+ - ajibawa-2023/Code-290k-ShareGPT
5
+ - m-a-p/Code-Feedback
6
+ - microsoft/orca-math-word-problems-200k
7
+ - teknium/openhermes
8
+ language:
9
+ - en
10
+ tags:
11
+ - code
12
+ - mathematics
13
+ quantized_by: bartowski
14
+ pipeline_tag: text-generation
15
+ ---
16
+
17
+ ## Exllama v2 Quantizations of Code-Mistral-7B
18
+
19
+ Using <a href="https://github.com/turboderp/exllamav2/releases/tag/v0.0.16">turboderp's ExLlamaV2 v0.0.16</a> for quantization.
20
+
21
+ <b>The "main" branch only contains the measurement.json, download one of the other branches for the model (see below)</b>
22
+
23
+ Each branch contains an individual bits per weight, with the main one containing only the meaurement.json for further conversions.
24
+
25
+ Original model: https://huggingface.co/ajibawa-2023/Code-Mistral-7B
26
+
27
+ | Branch | Bits | lm_head bits | VRAM (4k) | VRAM (16k) | VRAM (32k) | Description |
28
+ | ----- | ---- | ------- | ------ | ------ | ------ | ------------ |
29
+ | [8_0](https://huggingface.co/bartowski/Code-Mistral-7B-exl2/tree/8_0) | 8.0 | 8.0 | 8.4 GB | 9.8 GB | 11.8 GB | Maximum quality that ExLlamaV2 can produce, near unquantized performance. |
30
+ | [6_5](https://huggingface.co/bartowski/Code-Mistral-7B-exl2/tree/6_5) | 6.5 | 8.0 | 7.2 GB | 8.6 GB | 10.6 GB | Very similar to 8.0, good tradeoff of size vs performance, **recommended**. |
31
+ | [5_0](https://huggingface.co/bartowski/Code-Mistral-7B-exl2/tree/5_0) | 5.0 | 6.0 | 6.0 GB | 7.4 GB | 9.4 GB | Slightly lower quality vs 6.5, but usable on 8GB cards. |
32
+ | [4_25](https://huggingface.co/bartowski/Code-Mistral-7B-exl2/tree/4_25) | 4.25 | 6.0 | 5.3 GB | 6.7 GB | 8.7 GB | GPTQ equivalent bits per weight, slightly higher quality. |
33
+ | [3_5](https://huggingface.co/bartowski/Code-Mistral-7B-exl2/tree/3_5) | 3.5 | 6.0 | 4.7 GB | 6.1 GB | 8.1 GB | Lower quality, only use if you have to. |
34
+
35
+ ## Download instructions
36
+
37
+ With git:
38
+
39
+ ```shell
40
+ git clone --single-branch --branch 6_5 https://huggingface.co/bartowski/Code-Mistral-7B-exl2 Code-Mistral-7B-exl2-6_5
41
+ ```
42
+
43
+ With huggingface hub (credit to TheBloke for instructions):
44
+
45
+ ```shell
46
+ pip3 install huggingface-hub
47
+ ```
48
+
49
+ To download the `main` (only useful if you only care about measurement.json) branch to a folder called `Code-Mistral-7B-exl2`:
50
+
51
+ ```shell
52
+ mkdir Code-Mistral-7B-exl2
53
+ huggingface-cli download bartowski/Code-Mistral-7B-exl2 --local-dir Code-Mistral-7B-exl2 --local-dir-use-symlinks False
54
+ ```
55
+
56
+ To download from a different branch, add the `--revision` parameter:
57
+
58
+ Linux:
59
+
60
+ ```shell
61
+ mkdir Code-Mistral-7B-exl2-6_5
62
+ huggingface-cli download bartowski/Code-Mistral-7B-exl2 --revision 6_5 --local-dir Code-Mistral-7B-exl2-6_5 --local-dir-use-symlinks False
63
+ ```
64
+
65
+ Windows (which apparently doesn't like _ in folders sometimes?):
66
+
67
+ ```shell
68
+ mkdir Code-Mistral-7B-exl2-6.5
69
+ huggingface-cli download bartowski/Code-Mistral-7B-exl2 --revision 6_5 --local-dir Code-Mistral-7B-exl2-6.5 --local-dir-use-symlinks False
70
+ ```
71
+
72
+ Want to support my work? Visit my ko-fi page here: https://ko-fi.com/bartowski
measurement.json ADDED
The diff for this file is too large to render. See raw diff