I filtered out instructions with input context from sahil2801/CodeAlpaca-20k (CC BY) resulting in 9764 prompts that I then fed to WaveCoder Ultra at a low temperature.
This is the code:
from datasets import load_dataset
import json
from ollama import Client
from tqdm import tqdm
code_alpaca_dataset = load_dataset("sahil2801/CodeAlpaca-20k")
ollama_client = Client(host='
http://localhost:11434')
def filter_no_input(hf_dataset):
filtered_dataset = []
for i, prompt… See the full description on the dataset page:
https://huggingface.co/datasets/trollek/CodeMouse.
