bartowski commited on
Commit
e137340
·
verified ·
1 Parent(s): 9589d6e

measurement.json

Browse files
Files changed (2) hide show
  1. README.md +65 -0
  2. measurement.json +0 -0
README.md ADDED
@@ -0,0 +1,65 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ quantized_by: bartowski
4
+ pipeline_tag: text-generation
5
+ ---
6
+
7
+ ## Exllama v2 Quantizations of Yi-1.5-9B-Chat
8
+
9
+ Using <a href="https://github.com/turboderp/exllamav2/releases/tag/v0.0.21">turboderp's ExLlamaV2 v0.0.21</a> for quantization.
10
+
11
+ <b>The "main" branch only contains the measurement.json, download one of the other branches for the model (see below)</b>
12
+
13
+ Each branch contains an individual bits per weight, with the main one containing only the meaurement.json for further conversions.
14
+
15
+ Original model: https://huggingface.co/01-ai/Yi-1.5-9B-Chat
16
+
17
+ ## Prompt format
18
+
19
+ ```
20
+ {system_prompt}<|im_start|>user
21
+ {prompt}<|im_end|>
22
+ <|im_start|>assistant
23
+ <|im_end|>
24
+
25
+ ```
26
+
27
+ ## Available sizes
28
+
29
+ | Branch | Bits | lm_head bits | VRAM (4k) | VRAM (16k) | VRAM (32k) | Description |
30
+ | ----- | ---- | ------- | ------ | ------ | ------ | ------------ |
31
+ | [8_0](https://huggingface.co/bartowski/Yi-1.5-9B-Chat-exl2/tree/8_0) | 8.0 | 8.0 | 10.2 GB | 11.4 GB | 12.9 GB | Maximum quality that ExLlamaV2 can produce, near unquantized performance. |
32
+ | [6_5](https://huggingface.co/bartowski/Yi-1.5-9B-Chat-exl2/tree/6_5) | 6.5 | 8.0 | 8.7 GB | 9.9 GB | 11.4 GB | Very similar to 8.0, good tradeoff of size vs performance, **recommended**. |
33
+ | [5_0](https://huggingface.co/bartowski/Yi-1.5-9B-Chat-exl2/tree/5_0) | 5.0 | 6.0 | 7.2 GB | 8.4 GB | 9.9 GB | Slightly lower quality vs 6.5, but usable on 8GB cards. |
34
+ | [4_25](https://huggingface.co/bartowski/Yi-1.5-9B-Chat-exl2/tree/4_25) | 4.25 | 6.0 | 6.5 GB | 7.7 GB | 9.2 GB | GPTQ equivalent bits per weight, slightly higher quality. |
35
+ | [3_5](https://huggingface.co/bartowski/Yi-1.5-9B-Chat-exl2/tree/3_5) | 3.5 | 6.0 | 5.8 GB | 7.0 GB | 8.5 GB | Lower quality, only use if you have to. |
36
+
37
+ ## Download instructions
38
+
39
+ With git:
40
+
41
+ ```shell
42
+ git clone --single-branch --branch 6_5 https://huggingface.co/bartowski/Yi-1.5-9B-Chat-exl2 Yi-1.5-9B-Chat-exl2-6_5
43
+ ```
44
+
45
+ With huggingface hub (credit to TheBloke for instructions):
46
+
47
+ ```shell
48
+ pip3 install huggingface-hub
49
+ ```
50
+
51
+ To download a specific branch, use the `--revision` parameter. For example, to download the 6.5 bpw branch:
52
+
53
+ Linux:
54
+
55
+ ```shell
56
+ huggingface-cli download bartowski/Yi-1.5-9B-Chat-exl2 --revision 6_5 --local-dir Yi-1.5-9B-Chat-exl2-6_5 --local-dir-use-symlinks False
57
+ ```
58
+
59
+ Windows (which apparently doesn't like _ in folders sometimes?):
60
+
61
+ ```shell
62
+ huggingface-cli download bartowski/Yi-1.5-9B-Chat-exl2 --revision 6_5 --local-dir Yi-1.5-9B-Chat-exl2-6.5 --local-dir-use-symlinks False
63
+ ```
64
+
65
+ 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