measurement.json
Browse files- README.md +116 -0
- measurement.json +0 -0
README.md
ADDED
@@ -0,0 +1,116 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: other
|
3 |
+
license_name: llama3
|
4 |
+
license_link: https://huggingface.co/meta-llama/Meta-Llama-3-8B/blob/main/LICENSE
|
5 |
+
base_model: meta-llama/Meta-Llama-3-8B
|
6 |
+
tags:
|
7 |
+
- llama-3
|
8 |
+
- bagel
|
9 |
+
datasets:
|
10 |
+
- ai2_arc
|
11 |
+
- allenai/ultrafeedback_binarized_cleaned
|
12 |
+
- argilla/distilabel-intel-orca-dpo-pairs
|
13 |
+
- jondurbin/airoboros-3.2
|
14 |
+
- codeparrot/apps
|
15 |
+
- facebook/belebele
|
16 |
+
- bluemoon-fandom-1-1-rp-cleaned
|
17 |
+
- boolq
|
18 |
+
- camel-ai/biology
|
19 |
+
- camel-ai/chemistry
|
20 |
+
- camel-ai/math
|
21 |
+
- camel-ai/physics
|
22 |
+
- jondurbin/contextual-dpo-v0.1
|
23 |
+
- jondurbin/gutenberg-dpo-v0.1
|
24 |
+
- jondurbin/py-dpo-v0.1
|
25 |
+
- jondurbin/truthy-dpo-v0.1
|
26 |
+
- LDJnr/Capybara
|
27 |
+
- jondurbin/cinematika-v0.1
|
28 |
+
- WizardLM/WizardLM_evol_instruct_70k
|
29 |
+
- glaiveai/glaive-function-calling-v2
|
30 |
+
- jondurbin/gutenberg-dpo-v0.1
|
31 |
+
- grimulkan/LimaRP-augmented
|
32 |
+
- lmsys/lmsys-chat-1m
|
33 |
+
- ParisNeo/lollms_aware_dataset
|
34 |
+
- TIGER-Lab/MathInstruct
|
35 |
+
- Muennighoff/natural-instructions
|
36 |
+
- openbookqa
|
37 |
+
- kingbri/PIPPA-shareGPT
|
38 |
+
- piqa
|
39 |
+
- Vezora/Tested-22k-Python-Alpaca
|
40 |
+
- ropes
|
41 |
+
- cakiki/rosetta-code
|
42 |
+
- Open-Orca/SlimOrca
|
43 |
+
- b-mc2/sql-create-context
|
44 |
+
- squad_v2
|
45 |
+
- mattpscott/airoboros-summarization
|
46 |
+
- migtissera/Synthia-v1.3
|
47 |
+
- unalignment/toxic-dpo-v0.2
|
48 |
+
- WhiteRabbitNeo/WRN-Chapter-1
|
49 |
+
- WhiteRabbitNeo/WRN-Chapter-2
|
50 |
+
- winogrande
|
51 |
+
quantized_by: bartowski
|
52 |
+
pipeline_tag: text-generation
|
53 |
+
---
|
54 |
+
|
55 |
+
## Exllama v2 Quantizations of bagel-8b-v1.0
|
56 |
+
|
57 |
+
Using <a href="https://github.com/turboderp/exllamav2/releases/tag/v0.0.21">turboderp's ExLlamaV2 v0.0.21</a> for quantization.
|
58 |
+
|
59 |
+
<b>The "main" branch only contains the measurement.json, download one of the other branches for the model (see below)</b>
|
60 |
+
|
61 |
+
Each branch contains an individual bits per weight, with the main one containing only the meaurement.json for further conversions.
|
62 |
+
|
63 |
+
Original model: https://huggingface.co/jondurbin/bagel-8b-v1.0
|
64 |
+
|
65 |
+
## Prompt format
|
66 |
+
|
67 |
+
```
|
68 |
+
<|begin_of_text|><|start_header_id|>system<|end_header_id|>
|
69 |
+
|
70 |
+
{system_prompt}<|eot_id|><|start_header_id|>user<|end_header_id|>
|
71 |
+
|
72 |
+
{prompt}<|eot_id|><|start_header_id|>assistant<|end_header_id|>
|
73 |
+
|
74 |
+
|
75 |
+
```
|
76 |
+
|
77 |
+
## Available sizes
|
78 |
+
|
79 |
+
|
80 |
+
| Branch | Bits | lm_head bits | VRAM (4k) | VRAM (8K) | VRAM (16k) | VRAM (32k) | Description |
|
81 |
+
| ----- | ---- | ------- | ------ | ------ | ------ | ------ | ------------ |
|
82 |
+
| [8_0](https://huggingface.co/bartowski/bagel-8b-v1.0-exl2/tree/8_0) | 8.0 | 8.0 | 10.1 GB | 10.5 GB | 11.5 GB | 13.6 GB | Maximum quality that ExLlamaV2 can produce, near unquantized performance. |
|
83 |
+
| [6_5](https://huggingface.co/bartowski/bagel-8b-v1.0-exl2/tree/6_5) | 6.5 | 8.0 | 8.9 GB | 9.3 GB | 10.3 GB | 12.4 GB | Very similar to 8.0, good tradeoff of size vs performance, **recommended**. |
|
84 |
+
| [5_0](https://huggingface.co/bartowski/bagel-8b-v1.0-exl2/tree/5_0) | 5.0 | 6.0 | 7.7 GB | 8.1 GB | 9.1 GB | 11.2 GB | Slightly lower quality vs 6.5, but usable on 8GB cards. |
|
85 |
+
| [4_25](https://huggingface.co/bartowski/bagel-8b-v1.0-exl2/tree/4_25) | 4.25 | 6.0 | 7.0 GB | 7.4 GB | 8.4 GB | 10.5 GB | GPTQ equivalent bits per weight, slightly higher quality. |
|
86 |
+
| [3_5](https://huggingface.co/bartowski/bagel-8b-v1.0-exl2/tree/3_5) | 3.5 | 6.0 | 6.4 GB | 6.8 GB | 7.8 GB | 9.9 GB | Lower quality, only use if you have to. |
|
87 |
+
|
88 |
+
## Download instructions
|
89 |
+
|
90 |
+
With git:
|
91 |
+
|
92 |
+
```shell
|
93 |
+
git clone --single-branch --branch 6_5 https://huggingface.co/bartowski/bagel-8b-v1.0-exl2 bagel-8b-v1.0-exl2-6_5
|
94 |
+
```
|
95 |
+
|
96 |
+
With huggingface hub (credit to TheBloke for instructions):
|
97 |
+
|
98 |
+
```shell
|
99 |
+
pip3 install huggingface-hub
|
100 |
+
```
|
101 |
+
|
102 |
+
To download a specific branch, use the `--revision` parameter. For example, to download the 6.5 bpw branch:
|
103 |
+
|
104 |
+
Linux:
|
105 |
+
|
106 |
+
```shell
|
107 |
+
huggingface-cli download bartowski/bagel-8b-v1.0-exl2 --revision 6_5 --local-dir bagel-8b-v1.0-exl2-6_5 --local-dir-use-symlinks False
|
108 |
+
```
|
109 |
+
|
110 |
+
Windows (which apparently doesn't like _ in folders sometimes?):
|
111 |
+
|
112 |
+
```shell
|
113 |
+
huggingface-cli download bartowski/bagel-8b-v1.0-exl2 --revision 6_5 --local-dir bagel-8b-v1.0-exl2-6.5 --local-dir-use-symlinks False
|
114 |
+
```
|
115 |
+
|
116 |
+
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
|
|