bartowski commited on
Commit
b9bb1ea
1 Parent(s): 572fb1f

measurement.json

Browse files
Files changed (2) hide show
  1. README.md +77 -0
  2. measurement.json +0 -0
README.md ADDED
@@ -0,0 +1,77 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: gemma
3
+ base_model: HuggingFaceH4/zephyr-7b-gemma-sft-v0.1
4
+ tags:
5
+ - alignment-handbook
6
+ - generated_from_trainer
7
+ datasets:
8
+ - argilla/dpo-mix-7k
9
+ model-index:
10
+ - name: DiscoPOP-zephyr-7b-gemma
11
+ results: []
12
+ quantized_by: bartowski
13
+ pipeline_tag: text-generation
14
+ ---
15
+
16
+ ## Exllama v2 Quantizations of DiscoPOP-zephyr-7b-gemma
17
+
18
+ Using <a href="https://github.com/turboderp/exllamav2/releases/tag/v0.1.5">turboderp's ExLlamaV2 v0.1.5</a> for quantization.
19
+
20
+ <b>The "main" branch only contains the measurement.json, download one of the other branches for the model (see below)</b>
21
+
22
+ Each branch contains an individual bits per weight, with the main one containing only the meaurement.json for further conversions.
23
+
24
+ Original model: https://huggingface.co/SakanaAI/DiscoPOP-zephyr-7b-gemma
25
+
26
+ ## Prompt format
27
+
28
+ ```
29
+ <bos><|im_start|>system
30
+ {system_prompt}<|im_end|>
31
+ <|im_start|>user
32
+ {prompt}<|im_end|>
33
+ <|im_start|>assistant
34
+
35
+ ```
36
+
37
+ ## Available sizes
38
+
39
+ No GQA - VRAM requirements will be higher
40
+
41
+ | Branch | Bits | lm_head bits | Size (4k) | Size (16k) | Description |
42
+ | -------------------------------------------------------------- | ---- | ------------ | --------- | ---------- | ----------- |
43
+ | [8_0](https://huggingface.co/bartowski/DiscoPOP-zephyr-7b-gemma-exl2/tree/8_0) | 8.0 | 8.0 | 14.0 GB | 19.4 GB | Maximum quality that ExLlamaV2 can produce, near unquantized performance. |
44
+ | [6_5](https://huggingface.co/bartowski/DiscoPOP-zephyr-7b-gemma-exl2/tree/6_5) | 6.5 | 8.0 | 12.5 GB | 17.9 GB | Near unquantized performance at vastly reduced size, **recommended**. |
45
+ | [5_0](https://huggingface.co/bartowski/DiscoPOP-zephyr-7b-gemma-exl2/tree/5_0) | 5.0 | 6.0 | 10.9 GB | 16.3 GB | Slightly lower quality vs 6.5, but usable on 8GB cards with 4k context. |
46
+ | [4_25](https://huggingface.co/bartowski/DiscoPOP-zephyr-7b-gemma-exl2/tree/4_25) | 4.25 | 6.0 | 10.2 GB | 15.7 GB | GPTQ equivalent bits per weight. |
47
+ | [3_5](https://huggingface.co/bartowski/DiscoPOP-zephyr-7b-gemma-exl2/tree/3_5) | 3.5 | 6.0 | 9.5 GB | 14.9 GB | Lower quality, not recommended. |
48
+
49
+ ## Download instructions
50
+
51
+ With git:
52
+
53
+ ```shell
54
+ git clone --single-branch --branch 6_5 https://huggingface.co/bartowski/DiscoPOP-zephyr-7b-gemma-exl2 DiscoPOP-zephyr-7b-gemma-exl2-6_5
55
+ ```
56
+
57
+ With huggingface hub (credit to TheBloke for instructions):
58
+
59
+ ```shell
60
+ pip3 install huggingface-hub
61
+ ```
62
+
63
+ To download a specific branch, use the `--revision` parameter. For example, to download the 6.5 bpw branch:
64
+
65
+ Linux:
66
+
67
+ ```shell
68
+ huggingface-cli download bartowski/DiscoPOP-zephyr-7b-gemma-exl2 --revision 6_5 --local-dir DiscoPOP-zephyr-7b-gemma-exl2-6_5
69
+ ```
70
+
71
+ Windows (which apparently doesn't like _ in folders sometimes?):
72
+
73
+ ```shell
74
+ huggingface-cli download bartowski/DiscoPOP-zephyr-7b-gemma-exl2 --revision 6_5 --local-dir DiscoPOP-zephyr-7b-gemma-exl2-6.5
75
+ ```
76
+
77
+ 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