jam gpt

Jam-gpt

about :

note : its a simple toy llm model builder framework to play around llm models and understand its working principles

its not a llm model but you can put ur data and train it to build llm models

An Experimental implementation of language model (LLM) architecture for research and development of it architectures, design process to build, training, and fine-tuning efficient Generative Pretrained Transformers (GPT) models

for more ai related tools and framework look into OX-AI an open source AI project

Installation :

latest version

github pull will be clean if encountered with bugs please report issues

pip install git+https://github.com/Lokeshwaran-M/jam-gpt.git

stable release

jam-gpt==0.0.4 may not have fine tuning as its still under development and may contain bug please report issues if any

pip install jam-gpt

Usage :

Refere Docs and test-gptLM.ipynb for code examples

from jam_gpt.tokenizer import Tokenizer
from jam_gpt import config
from jam_gpt import lm
from jam_gpt.model import Model

md_name = "md-name"

tok = Tokenizer()
tok.get_encoding(md_name)

# model initilization
model = Model()

# load pretrined model
model.load_model(md_name)

# Generate data using Model
pmt = tok.encode("user prompt")
res = tok.decode(model.generate(pmt))
print(res)

Docs :

Jam-gpt docs will give you the complete useage and explanation of the jam-gpt library

1 setup
2 Collecting data
3 Tokenization
4 configuration
5 Language Model ( LM , Model )
6 Model Fine Tuning

Contribution :

for contribution guidelines and terms and condition to contribute refere jam-contribution by rasing the PR you are accepting the terms and condition

Any form of contribution is accepted here

Submitting :
    Issues
    pull requests
    feature requests
    bug reports
    documentation

credits :

• kudos to Andrej karpathy for his lectures on deep learning
• Open AI for GPT-2
• paper "Attention Is All You Need"