Skip to content

Accept NumPy integer indices in indexed datasets #8553

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
vineethsaivs wants to merge 1 commit into
base: master
Choose a base branch
from
+32 −6
Open

Accept NumPy integer indices in indexed datasets #8553

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
12 changes: 6 additions & 6 deletions deepspeed/runtime/data_pipeline/data_sampling/indexed_dataset.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -177,8 +177,8 @@ def __del__(self):
def __getitem__(self, idx):
if not self.data_file:
self.read_data(self.path)
if isinstance(idx, int):
i = idx
if isinstance(idx, (int, np.integer)):
i = int(idx)
self.check_index(i)
tensor_size = self.sizes[self.dim_offsets[i]:self.dim_offsets[i + 1]]
a = np.empty(tensor_size, dtype=self.dtype)
Expand Down Expand Up @@ -253,8 +253,8 @@ def prefetch(self, indices):

# @lru_cache(maxsize=8)
def __getitem__(self, idx):
if isinstance(idx, int):
i = idx
if isinstance(idx, (int, np.integer)):
i = int(idx)
self.check_index(i)
tensor_size = self.sizes[self.dim_offsets[i]:self.dim_offsets[i + 1]]
a = np.empty(tensor_size, dtype=self.dtype)
Expand Down Expand Up @@ -513,8 +513,8 @@ def __len__(self):

# @lru_cache(maxsize=8)
def __getitem__(self, idx):
if isinstance(idx, int):
ptr, size = self._index[idx]
if isinstance(idx, (int, np.integer)):
ptr, size = self._index[int(idx)]
np_array = np.frombuffer(self._bin_buffer, dtype=self._index.dtype, count=size, offset=ptr)
return np_array
elif isinstance(idx, slice):
Expand Down
26 changes: 26 additions & 0 deletions tests/unit/runtime/test_indexed_dataset.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,26 @@
# SPDX-License-Identifier: Apache-2.0
# DeepSpeed Team

Comment thread
vineethsaivs marked this conversation as resolved.
import numpy as np
import pytest
import torch

from deepspeed.runtime.data_pipeline.data_sampling.indexed_dataset import make_builder, make_dataset


@pytest.mark.parametrize("impl", ["lazy", "cached", "mmap"])
@pytest.mark.parametrize("index_dtype", [np.int32, np.int64, np.uint32, np.uint64])
def test_numpy_indices_through_subset_dataloader(tmp_path, impl, index_dtype):
prefix = str(tmp_path / "samples")
builder = make_builder(prefix + ".bin", impl=impl)
rows = [torch.tensor([1, 2, 3]), torch.tensor([4, 5, 6]), torch.tensor([7, 8, 9])]
for row in rows:
builder.add_item(row)
builder.finalize(prefix + ".idx")
dataset = make_dataset(prefix, impl=impl, skip_warmup=True)
indices = np.array([2, 0], dtype=index_dtype)
if dataset.supports_prefetch:
dataset.prefetch([0, 2])
subset = torch.utils.data.Subset(dataset, indices)
batch = next(iter(torch.utils.data.DataLoader(subset, batch_size=2)))
torch.testing.assert_close(batch.long(), torch.stack([rows[2], rows[0]]))
Loading