ayushdg

ayushdg

ayushdg

Git bisect to identify bad commit (here dask is used as an example)

Prerequisites to git bisect:

• Know how to build dask from source
• Known good commit eg: Dask version 2.4
• Known bad commit eg: Dask version 2.5
• Have a script/test that fails with bad commit but works with good commit
• Ensure that the test indeed fails with dask built from source with bad commit and passes for dask built from source with good commit. (In the existing environment)

ayushdg

import dask_xgboost as dxgb_gpu
import dask
import dask_cudf
from dask_cuda import LocalCUDACluster
from dask.delayed import delayed
from dask.distributed import Client, wait
import cudf

import numpy as np
import xgboost as xgb

ayushdg

ayushdg

def concat_dfs(df_list):
    """
    Concat a list of cudf dataframes
    """
    return cudf.concat(df_list)

def get_delayed_dict(ddf):
    """
    Returns a dicitionary with the dataframe tasks as keys and 
    the dataframe delayed objects as values

ayushdg

ayushdg

ayushdg