Main branch
Browse files- README.md +77 -0
- measurement.json +0 -0
README.md
ADDED
@@ -0,0 +1,77 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
base_model: qblocks/zephyr_7b_norobots
|
3 |
+
datasets:
|
4 |
+
- HuggingFaceH4/no_robots
|
5 |
+
inference: false
|
6 |
+
library_name: peft
|
7 |
+
license: apache-2.0
|
8 |
+
model_creator: MonsterAPI
|
9 |
+
model_name: Zephyr 7B Norobots
|
10 |
+
model_type: mistral
|
11 |
+
prompt_template: '<|im_start|>system
|
12 |
+
|
13 |
+
{system_message}<|im_end|>
|
14 |
+
|
15 |
+
<|im_start|>user
|
16 |
+
|
17 |
+
{prompt}<|im_end|>
|
18 |
+
|
19 |
+
<|im_start|>assistant
|
20 |
+
|
21 |
+
'
|
22 |
+
tags:
|
23 |
+
- code
|
24 |
+
- instruct
|
25 |
+
- zephyr
|
26 |
+
---
|
27 |
+
---
|
28 |
+
#### Prompt Structure
|
29 |
+
```
|
30 |
+
<|system|> <|endoftext|> <|user|> [USER PROMPT]<|endoftext|> <|assistant|> [ASSISTANT ANSWER] <|endoftext|>
|
31 |
+
```
|
32 |
+
#### Train loss :
|
33 |
+
|
34 |
+
![training loss](https://cdn-uploads.huggingface.co/production/uploads/63ba46aa0a9866b28cb19a14/EYUTlcPFz-2nXzNj5_gsW.png)
|
35 |
+
quantized_by: bartowski
|
36 |
+
pipeline_tag: text-generation
|
37 |
+
---
|
38 |
+
|
39 |
+
## Exllama v2 Quantizations of zephyr_7b_norobots
|
40 |
+
|
41 |
+
Using <a href="https://github.com/turboderp/exllamav2/releases/tag/v0.0.8">turboderp's ExLlamaV2 v0.0.8</a> for quantization.
|
42 |
+
|
43 |
+
Each branch contains an individual bits per weight, with the main one containing only the meaurement.json for further conversions.
|
44 |
+
|
45 |
+
Conversion was done using wikitext-103-raw-v1-test.parquet as calibration dataset.
|
46 |
+
|
47 |
+
Default arguments used except when the bits per weight is above 6.0, at that point the lm_head layer is quantized at 8 bits per weight instead of the default 6.
|
48 |
+
|
49 |
+
Original model: https://huggingface.co/TheBloke/zephyr_7b_norobots-fp16
|
50 |
+
|
51 |
+
## Download instructions
|
52 |
+
|
53 |
+
With git:
|
54 |
+
|
55 |
+
```shell
|
56 |
+
git clone --single-branch --branch 4_0 https://huggingface.co/bartowski/zephyr_7b_norobots-exl2
|
57 |
+
```
|
58 |
+
|
59 |
+
With huggingface hub (credit to TheBloke for instructions):
|
60 |
+
|
61 |
+
```shell
|
62 |
+
pip3 install huggingface-hub
|
63 |
+
```
|
64 |
+
|
65 |
+
To download the `main` (only useful if you only care about measurement.json) branch to a folder called `zephyr_7b_norobots-exl2`:
|
66 |
+
|
67 |
+
```shell
|
68 |
+
mkdir zephyr_7b_norobots-exl2
|
69 |
+
huggingface-cli download bartowski/zephyr_7b_norobots-exl2 --local-dir zephyr_7b_norobots-exl2 --local-dir-use-symlinks False
|
70 |
+
```
|
71 |
+
|
72 |
+
To download from a different branch, add the `--revision` parameter:
|
73 |
+
|
74 |
+
```shell
|
75 |
+
mkdir zephyr_7b_norobots-exl2
|
76 |
+
huggingface-cli download bartowski/zephyr_7b_norobots-exl2 --revision 4_0 --local-dir zephyr_7b_norobots-exl2 --local-dir-use-symlinks False
|
77 |
+
```
|
measurement.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|