Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/michaelhly/farglot
A Transformer-based SocialNLP toolkit for Farcaster
https://github.com/michaelhly/farglot
farcaster nlp transformers
Last synced: about 1 month ago
JSON representation
A Transformer-based SocialNLP toolkit for Farcaster
- Host: GitHub
- URL: https://github.com/michaelhly/farglot
- Owner: michaelhly
- License: mit
- Created: 2023-08-11T10:44:42.000Z (over 1 year ago)
- Default Branch: master
- Last Pushed: 2023-08-28T15:40:25.000Z (about 1 year ago)
- Last Synced: 2024-10-06T17:36:31.523Z (about 2 months ago)
- Topics: farcaster, nlp, transformers
- Language: Python
- Homepage:
- Size: 321 KB
- Stars: 7
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# FarGlot
A Transformer-based SocialNLP toolkit for [Farcaster](https://www.farcaster.xyz/).
## Installation
```
pip install farglot
```
## Examples
```python
from farglot import CastAnalyzer
sentiment_analyzer=CastAnalyzer.sequence_analzyer_from_model_name(
hub_address="nemes.farcaster.xyz:2283",
model_name="pysentimiento/robertuito-sentiment-analysis"
)
sentiment_analyzer.predict_cast(fid=2, hash_hex="0bcdcbf006ec22b79f37f2cf2a09c33413883937")
# {'NEG': 0.051998768001794815, 'NEU': 0.22470703721046448, 'POS': 0.7232941389083862}
sentiment_analyzer.predict_casts_by_fid(fid=2)
# {'NEG': 0.03734538331627846, 'NEU': 0.505352795124054, 'POS': 0.4573018550872803}
```
## Generate a Training Corpus from a [Hub](https://github.com/farcasterxyz/hub-monorepo/tree/main/apps/hubble)
### Install the FarGlot CLI
```
pip install "farglot[cli]"
```
### Define Training Set Classifier(s)
```json
{
"name": "labels",
"default_value": 1 // optional
}
```
For multi-label classfication:
```json
[
{
"name": "class_one",
"default_value": 1 // optional
},
{
"name": "class_two",
"default_value": 2 // optional
},
{
"name": "class_three",
"default_value": 3 // optional
}
]
```
### Usage
```sh
farglot init
farglot set-classifers-path /path/to/class_configs.json
farglot set-hub-db-path /path/to/.rocks/rocks.hub._default
farglot new-training-set --out ./data/training-set.csv
```
### Tuning
TODO: Example of fine-tuning and uploading dataset and model to [Hugging Face](https://huggingface.co/)
### Tuning Resources
Not sure how to where to start? Check out the following blog posts on tuning an LLM:
- [Datasets and Preprocessing](https://michaelhly.com/posts/tune-llm-one)
- [Hyperparameters and Metrics](https://michaelhly.com/posts/tune-llm-two)
This largely is largely adapted off of [pysentimiento](https://github.com/pysentimiento/pysentimiento).