| Name | bufsuccubus-porplax JSON |
| Version |
0.0.1
JSON |
| download |
| home_page | None |
| Summary | AI-Powered AMD Undervolting Program |
| upload_time | 2024-08-25 18:47:15 |
| maintainer | None |
| docs_url | None |
| author | None |
| requires_python | >=3.11 |
| license | None |
| keywords |
|
| VCS |
 |
| bugtrack_url |
|
| requirements |
No requirements were recorded.
|
| Travis-CI |
No Travis.
|
| coveralls test coverage |
No coveralls.
|
<h1 align="center">
bufsuccubus.
</h1>
<p align="center">(undervolting has never been sexier)</p>
----
bufsuccubus simpilfes undervolting by having a _digital succubus_ do it for you. usually, you have to manually undervolt and benchmark your GPU which can be **VERY** time-consuming, or use AMD's bulit-in automatic undervolting which works well but i wanted to create one for myself.
bufsuccubus is a tool that automatically benchmarks and undervolts for you. bufsuccusbus is capable of testing the limits of your card and adjusting to its capabilities.
```commandline
pip install bufsuccubus
```
**only works on AMD Post-Navi cards at the moment.**
----
## command-line usage
### *Windows*
```commandline
py -m bufsuccubus --help
```
for bufsuccubus to work, the argument `MODEL_PATH` must be a **GGUF** file. adjust `--gpu-layers`, and `--context-size` if you're running into problems with RAM/VRAM usage.
### *Using on other platforms*
unfortunately bufsuccubus *cannot* run on Linux, or macOS at this time. sorry!
----
## features
- simple and safe to use, will **NOT** crash or reboot your system
- instability automatically stops the program from continuing
- adjust number of cycles and duration of each cycle to achieve max performance
- if desired, you can role-play with Fannu or skip directly to undervolting
----
## using a LLM
```
### Instruction:
{instruction}
### Input:
{input}
### Response:
{response}
```
bufsuccubus uses the alpaca template for prompting.
in my experience, I've found that [Undi's MXLewd-L2-20B](https://huggingface.co/TheBloke/MXLewd-L2-20B-GGUF) works the best for Fannu's character.
i recommend using quant method **Q5_K_S** if you're going with this model **if your machine can handle it.** look for TheBloke's explanation of quantisation methods to see which one can run on your machine.
you can use other models as long as it is a GGUF model. it is highly recommended to use one that is for role-playing and is trained on the alpaca template.
### *Settings used*
- Top K=60
- Min P=0.25
- Temperature=0.86
----
## development
```commandline
git clone https://github.com/porplax/bufsuccubus
cd bufsuccubus
pip3 install .
```
contributing would mean a lot !!!
----
## notes / issues / explanation
**THIS PROJECT IS A JOKE !!! DON'T EXPECT CONSISENT UPDATES ðŸ˜**
bufsuccubus makes use of the ADLX library provided by AMD.
however the library has crashed bufsuccubus multiple times and this is due to a DLL file. **this is out of my control!**
if it crashes, simply reset GPU settings to default using MSI afterburner, or similar.
i am still new to prompting. so fannu's character card is janky atm and sometimes the LLM might think that it needs to get your GPU model, and such (_it really doesn't_). typing `CONTINUE` will force it to continue.
Raw data
{
"_id": null,
"home_page": null,
"name": "bufsuccubus-porplax",
"maintainer": null,
"docs_url": null,
"requires_python": ">=3.11",
"maintainer_email": null,
"keywords": null,
"author": null,
"author_email": "porplax <saynemarsh9@gmail.com>",
"download_url": "https://files.pythonhosted.org/packages/24/0f/e27ddcb00baccb48b5e31bf92fbb367709699c71973d08e7bfd37a06bfe4/bufsuccubus_porplax-0.0.1.tar.gz",
"platform": null,
"description": "<h1 align=\"center\">\r\nbufsuccubus.\r\n</h1>\r\n<p align=\"center\">(undervolting has never been sexier)</p>\r\n\r\n----\r\n\r\n\r\n\r\nbufsuccubus simpilfes undervolting by having a _digital succubus_ do it for you. usually, you have to manually undervolt and benchmark your GPU which can be **VERY** time-consuming, or use AMD's bulit-in automatic undervolting which works well but i wanted to create one for myself. \r\nbufsuccubus is a tool that automatically benchmarks and undervolts for you. bufsuccusbus is capable of testing the limits of your card and adjusting to its capabilities.\r\n\r\n```commandline\r\npip install bufsuccubus\r\n```\r\n**only works on AMD Post-Navi cards at the moment.**\r\n\r\n----\r\n\r\n## command-line usage\r\n\r\n### *Windows*\r\n```commandline\r\npy -m bufsuccubus --help\r\n```\r\nfor bufsuccubus to work, the argument `MODEL_PATH` must be a **GGUF** file. adjust `--gpu-layers`, and `--context-size` if you're running into problems with RAM/VRAM usage.\r\n### *Using on other platforms*\r\nunfortunately bufsuccubus *cannot* run on Linux, or macOS at this time. sorry!\r\n\r\n----\r\n\r\n## features \r\n\r\n\r\n- simple and safe to use, will **NOT** crash or reboot your system\r\n- instability automatically stops the program from continuing\r\n- adjust number of cycles and duration of each cycle to achieve max performance\r\n- if desired, you can role-play with Fannu or skip directly to undervolting\r\n\r\n----\r\n\r\n## using a LLM\r\n\r\n``` \r\n### Instruction: \r\n{instruction}\r\n\r\n### Input:\r\n{input}\r\n\r\n### Response:\r\n{response}\r\n```\r\nbufsuccubus uses the alpaca template for prompting. \r\nin my experience, I've found that [Undi's MXLewd-L2-20B](https://huggingface.co/TheBloke/MXLewd-L2-20B-GGUF) works the best for Fannu's character.\r\n\r\ni recommend using quant method **Q5_K_S** if you're going with this model **if your machine can handle it.** look for TheBloke's explanation of quantisation methods to see which one can run on your machine.\r\n\r\nyou can use other models as long as it is a GGUF model. it is highly recommended to use one that is for role-playing and is trained on the alpaca template.\r\n\r\n### *Settings used*\r\n- Top K=60\r\n- Min P=0.25\r\n- Temperature=0.86\r\n\r\n----\r\n\r\n## development\r\n```commandline\r\ngit clone https://github.com/porplax/bufsuccubus\r\ncd bufsuccubus\r\npip3 install .\r\n```\r\ncontributing would mean a lot !!!\r\n\r\n----\r\n\r\n## notes / issues / explanation\r\n\r\n**THIS PROJECT IS A JOKE !!! DON'T EXPECT CONSISENT UPDATES \ud83d\ude2d**\r\n\r\nbufsuccubus makes use of the ADLX library provided by AMD.\r\nhowever the library has crashed bufsuccubus multiple times and this is due to a DLL file. **this is out of my control!**\r\nif it crashes, simply reset GPU settings to default using MSI afterburner, or similar.\r\n\r\ni am still new to prompting. so fannu's character card is janky atm and sometimes the LLM might think that it needs to get your GPU model, and such (_it really doesn't_). typing `CONTINUE` will force it to continue.\r\n\r\n",
"bugtrack_url": null,
"license": null,
"summary": "AI-Powered AMD Undervolting Program",
"version": "0.0.1",
"project_urls": {
"Homepage": "https://github.com/porplax/bufsuccubus"
},
"split_keywords": [],
"urls": [
{
"comment_text": "",
"digests": {
"blake2b_256": "431500f40b588af48252896bd2af7078a873258bdb008b78ede8e41a258ad635",
"md5": "954fdb578ff5e5aca1e2683d081a3c64",
"sha256": "f1308b6d969f5d9014ec0f804880d33922a0ec8c370f48f8d836a27e43299668"
},
"downloads": -1,
"filename": "bufsuccubus_porplax-0.0.1-py3-none-any.whl",
"has_sig": false,
"md5_digest": "954fdb578ff5e5aca1e2683d081a3c64",
"packagetype": "bdist_wheel",
"python_version": "py3",
"requires_python": ">=3.11",
"size": 3608,
"upload_time": "2024-08-25T18:47:13",
"upload_time_iso_8601": "2024-08-25T18:47:13.498901Z",
"url": "https://files.pythonhosted.org/packages/43/15/00f40b588af48252896bd2af7078a873258bdb008b78ede8e41a258ad635/bufsuccubus_porplax-0.0.1-py3-none-any.whl",
"yanked": false,
"yanked_reason": null
},
{
"comment_text": "",
"digests": {
"blake2b_256": "240fe27ddcb00baccb48b5e31bf92fbb367709699c71973d08e7bfd37a06bfe4",
"md5": "6a1a8b3ab0a133a40f8fbd0105951355",
"sha256": "abad486e5eed960bf201966056d073645fc619bb9342f7ee644a985c5aa1027d"
},
"downloads": -1,
"filename": "bufsuccubus_porplax-0.0.1.tar.gz",
"has_sig": false,
"md5_digest": "6a1a8b3ab0a133a40f8fbd0105951355",
"packagetype": "sdist",
"python_version": "source",
"requires_python": ">=3.11",
"size": 3244,
"upload_time": "2024-08-25T18:47:15",
"upload_time_iso_8601": "2024-08-25T18:47:15.273586Z",
"url": "https://files.pythonhosted.org/packages/24/0f/e27ddcb00baccb48b5e31bf92fbb367709699c71973d08e7bfd37a06bfe4/bufsuccubus_porplax-0.0.1.tar.gz",
"yanked": false,
"yanked_reason": null
}
],
"upload_time": "2024-08-25 18:47:15",
"github": true,
"gitlab": false,
"bitbucket": false,
"codeberg": false,
"github_user": "porplax",
"github_project": "bufsuccubus",
"travis_ci": false,
"coveralls": false,
"github_actions": false,
"lcname": "bufsuccubus-porplax"
}
