emilis
/
izzilis
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Improved readme

This commit is contained in:
emilis 2021-09-01 19:49:57 +02:00
parent b842c7bd94
commit 0dbbde5835
1 changed files with 44 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

51
README.md
View File

@ -4,15 +4,52 @@ Meant to be used with a finetuned GPT-2 model
## Usage
just run it, it'll make a `bot_config.json` file in the running path
To run the bot, you need a valid `bot_config.json` file at the path where you're running the bot.
If you do not have one, izzilis will generate a default one for you to fill out.
fill it out
This bot currently *requires* a Telegram bot token to work, as there is no option to disable curation.
To create a bot, please use the [Botfather](https://t.me/botfather). Once created, and running, set a channel for the bot to post curation options in via the `/setmain` bot command. Usually this requires sending `/setmain@bot_username` to the chat with the bot (can be group chats).
run again
This default uses the `Misskey` publisher. If you want to publish to Mastodon, Pleroma, or any other mastodon-compatible API, please replace `Misskey` in the `publisher` object with `Mastodon`.
## Docker Usage
## Config values
The dockerfile makes a few assumptions which you must fulfill:
| Name | Value |
|--------------------|------------------------------------------------------------------------------------------------|
| `python_path` | The path to the system's python3 interpreter |
| `model_name` | The name of the GPT-2 model to use (see gpt-2 docs) |
| `temperature` | The `temperature` value to call gpt-2 with (see gpt-2 docs) |
| `top_k` | The `top_k` value to call gpt-2 with (see gpt-2 docs) |
| `gpt_code_path` | The path to where the gpt-2 source & models are located |
| `interval_seconds` | See [interval_seconds](#interval_seconds) |
| `bot_token` | Telegram Bot API token |
| `chat_ref` | The chat reference ID for the telegram bot, leave at 0, will be filled once `/setmain` is sent |
| `post_buffer` | How many curated samples the bot will hold at maximum at a time |
| `publisher` | See [publisher](#publisher) |
* the gpt python files & model are in the gpt/ directory (this directory doesn't exist, please make it (it must contain both `generate_unconditional_samples.py` and the `models` directory containing your finetuned model))
* you have ran the bot already. Once to generate a `bot_config.json` and a second time, once the aforementioned file is generated and filled out by the user, `fediverse.toml` from authenticating with the instance.
### interval_seconds
| Name | Value |
|-------|-------------------------------------------------|
| `min` | Minimum amount of seconds to wait between posts |
| `max` | Maximum amount of seconds to wait between posts |
### publisher
The publisher can currently hold one of two JSON objects, named either `Mastodon` or `Misskey`, which determines which posting API it will use. Whether the object is `Misskey` or `Mastodon`, it has the following members:
| Name | Value |
|------------|------------------------------------------------------------------------------------------------------------|
| `base_url` | The base URL of the instance |
| `token` | The auth token for the account, leave empty for `Mastodon` as you will be prompted to log in and authorize |
An example `Misskey` publisher entry looks like this:
```json
"publisher": {
"Misskey": {
"base_url": "",
"token": ""
}
}
```