github/gpt4all
1
0
Fork 0
mirror of https://github.com/nomic-ai/gpt4all.git synced 2025-08-12 05:12:07 +00:00
Code Issues Packages Projects Releases Wiki Activity

feat: pull from multiple datasets

Browse Source
This commit is contained in:
Zach Nussbaum 2023-04-17 20:00:19 +00:00
parent 0b4d45e57d
commit c76f6e33a9
1 changed files with 18 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

18
data.py
View File

@ -61,7 +61,23 @@ def tokenize_inputs(config, tokenizer, examples):
def load_data(config, tokenizer): def load_data(config, tokenizer):
dataset_path = config["dataset_path"] dataset_path = config["dataset_path"]
if os.path.exists(dataset_path): if isinstance(dataset_path, list):
all_datasets = []
for path in dataset_path:
dataset = load_dataset(path, split="train")
current_columns = dataset.column_names
columns_to_keep = ["prompt", "response"]
to_remove = set(current_columns) - set(columns_to_keep)
dataset = dataset.remove_columns(to_remove)
if "source" not in current_columns:
dataset = dataset.add_column("source", [path.split("/")[-1]] * len(dataset))
all_datasets.append(dataset)
dataset = concatenate_datasets(all_datasets)
elif os.path.exists(dataset_path):
if os.path.isdir(dataset_path): if os.path.isdir(dataset_path):
files = glob.glob(os.path.join(dataset_path, "*_clean.jsonl")) files = glob.glob(os.path.join(dataset_path, "*_clean.jsonl"))
else: else: