mirror of
https://github.com/hiyouga/LLaMA-Factory.git
synced 2026-05-28 02:48:54 +08:00
[v1] Add FlashAttention selection and implement normal / padding-free / dynamic batching (#10469)
This commit is contained in:
@@ -117,12 +117,3 @@ def test_pair_converter(num_samples: int):
|
||||
],
|
||||
}
|
||||
assert data_engine[index] == {"_dataset_name": "tiny_dataset", **expected_data}
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
"""
|
||||
python -m tests_v1.plugins.data_plugins.test_converter
|
||||
"""
|
||||
test_alpaca_converter(1)
|
||||
test_sharegpt_converter()
|
||||
test_pair_converter(1)
|
||||
|
||||
Reference in New Issue
Block a user