Fix for Incorrect ex_iterable used with multi num_worker #6582
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Corrects an issue where `self._ex_iterable` was erroneously used instead of `ex_iterable`, when both Distributed Data Parallel (DDP) and multi num_worker are used concurrently. This improper usage led to the generation of incorrect `shards_indices`, subsequently causing issues with the control flow responsible for worker creation. The fix ensures the appropriate iterable is used, thus providing a more accurate determination of whether a new worker should be instantiated or not.
|
A toy example to reveal the bug. """
DATASETS_VERBOSITY=debug torchrun --nproc-per-node 2 main.py
"""
import torch.utils.data
import torch.distributed
import datasets.distributed
import datasets
# num shards = 4
shards = [(0, 100), (100, 200), (200, 300), (300, 400)]
def gen(shards):
for st, ed in shards:
yield from range(st, ed)
torch.distributed.init_process_group()
# want to create total worker = world_size * 8
ds = datasets.IterableDataset.from_generator(gen, gen_kwargs={'shards': shards})
ds = datasets.distributed.split_dataset_by_node(
ds,
rank=torch.distributed.get_rank(),
world_size=torch.distributed.get_world_size(),
)
dl = torch.utils.data.DataLoader(ds, batch_size=10, num_workers=8)
for x in dl:
print(f"RANK={torch.distributed.get_rank()} {x}") |
lhoestq
approved these changes
Mar 1, 2024
Show benchmarksPyArrow==8.0.0 Show updated benchmarks!Benchmark: benchmark_array_xd.json
Benchmark: benchmark_getitem_100B.json
Benchmark: benchmark_indices_mapping.json
Benchmark: benchmark_iterating.json
Benchmark: benchmark_map_filter.json
Show updated benchmarks!Benchmark: benchmark_array_xd.json
Benchmark: benchmark_getitem_100B.json
Benchmark: benchmark_indices_mapping.json
Benchmark: benchmark_iterating.json
Benchmark: benchmark_map_filter.json
|
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Corrects an issue where
self._ex_iterablewas erroneously used instead ofex_iterable, when both Distributed Data Parallel (DDP) and multi num_worker are used concurrently. This improper usage led to the generation of incorrectshards_indices, subsequently causing issues with the control flow responsible for worker creation. The fix ensures the appropriate iterable is used, thus providing a more accurate determination of whether a new worker should be instantiated or not.