ChaiVerse is the community's one-stop repository for large language model training, deployment and LIVE USERS evaluation package. Train models and win prizes!

A Collaboration Between

Quick Start With Run Pod

Just watch the following 3 minutes tutorial, our repo is made to run on runpod 🚀🚀

Model Training

To train a LLaMA7b model and push it to huggingface it's just 5 lines of code 🥳

import chaiverse as cv

dataset = cv.load_dataset('ChaiML/davinci_150_examples', 'chatml')

model = cv.LLaMA7b()
model.fit(dataset, output_dir='./my_first_llama', num_epochs=1)

model_url = 'ChaiML/llama7b_dummy' # your huggingface URL
model.push_to_hub(model_url, private=True)

Winning Prizes with Chai Guanaco

chai-guanaco login

Now submit the model you have just trained!

import chai_guanaco as chai

model_url = "ChaiML/llama7b_dummy"

generation_params = {
        'temperature': 1.0,
        'repetition_penalty': 1.13,
        'top_p': 0.2,
        "top_k": 40,
        "stopping_words": ['\n'],
        "presence_penalty": 1.5,
        "frequency_penalty": 1.5
        }
submission_parameters = {'model_repo': model_url, 'generation_params': generation_params, 'model_name': 'my-awesome-llama'}

submitter = chai.ModelSubmitter()
submission_id = submitter.submit(submission_parameters)

Sponsored By