Skip to content

Updating dataset code to avoid creating multiple iterators from a DataPipe #1708

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

Merged
merged 1 commit into from
May 11, 2022
Merged

Updating dataset code to avoid creating multiple iterators from a DataPipe #1708

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
14 changes: 6 additions & 8 deletions torchtext/datasets/iwslt2016.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -125,7 +125,7 @@
# avoid additional conditional imports.
def _filter_clean_cache(cache_decompressed_dp, full_filepath, uncleaned_filename):
cache_inner_decompressed_dp = cache_decompressed_dp.on_disk_cache(filepath_fn=lambda x: full_filepath)
cache_inner_decompressed_dp = FileOpener(cache_inner_decompressed_dp, mode="b").load_from_tar()
cache_inner_decompressed_dp = cache_inner_decompressed_dp.open_files(mode="b").load_from_tar()
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

FYI, we recently added .open_files as the functional form of FileOpener

cache_inner_decompressed_dp = cache_inner_decompressed_dp.filter(
lambda x: os.path.basename(uncleaned_filename) in x[0]
)
Expand Down Expand Up @@ -261,12 +261,10 @@ def IWSLT2016(
)

cache_decompressed_dp = cache_compressed_dp.on_disk_cache(filepath_fn=lambda x: inner_iwslt_tar)
cache_decompressed_dp = (
FileOpener(cache_decompressed_dp, mode="b")
.load_from_tar()
.filter(lambda x: os.path.basename(inner_iwslt_tar) in x[0])
)
cache_decompressed_dp = cache_decompressed_dp.open_files(mode="b").load_from_tar()
cache_decompressed_dp = cache_decompressed_dp.filter(lambda x: os.path.basename(inner_iwslt_tar) in x[0])
cache_decompressed_dp = cache_decompressed_dp.end_caching(mode="wb", same_filepath_fn=True)
cache_decompressed_dp_1, cache_decompressed_dp_2 = cache_decompressed_dp.fork(num_instances=2)
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The recommended pattern is to use fork when you need the same DataPipe for multiple usages later.


src_filename = file_path_by_lang_and_split[src_language][split]
uncleaned_src_filename = uncleaned_filenames_by_lang_and_split[src_language][split]
Expand All @@ -276,7 +274,7 @@ def IWSLT2016(
full_src_filepath = os.path.join(root, "2016-01/texts/", src_language, tgt_language, languages, src_filename)

cache_inner_src_decompressed_dp = _filter_clean_cache(
cache_decompressed_dp, full_src_filepath, uncleaned_src_filename
cache_decompressed_dp_1, full_src_filepath, uncleaned_src_filename
)

tgt_filename = file_path_by_lang_and_split[tgt_language][split]
Expand All @@ -287,7 +285,7 @@ def IWSLT2016(
full_tgt_filepath = os.path.join(root, "2016-01/texts/", src_language, tgt_language, languages, tgt_filename)

cache_inner_tgt_decompressed_dp = _filter_clean_cache(
cache_decompressed_dp, full_tgt_filepath, uncleaned_tgt_filename
cache_decompressed_dp_2, full_tgt_filepath, uncleaned_tgt_filename
)

tgt_data_dp = FileOpener(cache_inner_tgt_decompressed_dp, encoding="utf-8")
Expand Down
9 changes: 5 additions & 4 deletions torchtext/datasets/iwslt2017.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -104,7 +104,7 @@
# avoid additional conditional imports.
def _filter_clean_cache(cache_decompressed_dp, full_filepath, uncleaned_filename):
cache_inner_decompressed_dp = cache_decompressed_dp.on_disk_cache(filepath_fn=lambda x: full_filepath)
cache_inner_decompressed_dp = FileOpener(cache_inner_decompressed_dp, mode="b").load_from_tar()
cache_inner_decompressed_dp = cache_inner_decompressed_dp.open_files(mode="b").load_from_tar()
cache_inner_decompressed_dp = cache_inner_decompressed_dp.filter(
lambda x: os.path.basename(uncleaned_filename) in x[0]
)
Expand Down Expand Up @@ -208,8 +208,9 @@ def IWSLT2017(root=".data", split=("train", "valid", "test"), language_pair=("de
)

cache_decompressed_dp = cache_compressed_dp.on_disk_cache(filepath_fn=lambda x: inner_iwslt_tar)
cache_decompressed_dp = FileOpener(cache_decompressed_dp, mode="b").load_from_tar()
cache_decompressed_dp = cache_decompressed_dp.open_files(mode="b").load_from_tar()
cache_decompressed_dp = cache_decompressed_dp.end_caching(mode="wb", same_filepath_fn=True)
cache_decompressed_dp_1, cache_decompressed_dp_2 = cache_decompressed_dp.fork(num_instances=2)

src_filename = file_path_by_lang_and_split[src_language][split]
uncleaned_src_filename = uncleaned_filenames_by_lang_and_split[src_language][split]
Expand All @@ -224,7 +225,7 @@ def IWSLT2017(root=".data", split=("train", "valid", "test"), language_pair=("de
)

cache_inner_src_decompressed_dp = _filter_clean_cache(
cache_decompressed_dp, full_src_filepath, uncleaned_src_filename
cache_decompressed_dp_1, full_src_filepath, uncleaned_src_filename
)

tgt_filename = file_path_by_lang_and_split[tgt_language][split]
Expand All @@ -240,7 +241,7 @@ def IWSLT2017(root=".data", split=("train", "valid", "test"), language_pair=("de
)

cache_inner_tgt_decompressed_dp = _filter_clean_cache(
cache_decompressed_dp, full_tgt_filepath, uncleaned_tgt_filename
cache_decompressed_dp_2, full_tgt_filepath, uncleaned_tgt_filename
)

tgt_data_dp = FileOpener(cache_inner_tgt_decompressed_dp, encoding="utf-8")
Expand Down
6 changes: 4 additions & 2 deletions torchtext/datasets/multi30k.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -79,7 +79,9 @@ def Multi30k(root: str, split: Union[Tuple[str], str], language_pair: Tuple[str]
)
cache_compressed_dp = HttpReader(cache_compressed_dp).end_caching(mode="wb", same_filepath_fn=True)

src_cache_decompressed_dp = cache_compressed_dp.on_disk_cache(
cache_compressed_dp_1, cache_compressed_dp_2 = cache_compressed_dp.fork(num_instances=2)

src_cache_decompressed_dp = cache_compressed_dp_1.on_disk_cache(
filepath_fn=lambda x: os.path.join(root, f"{_PREFIX[split]}.{language_pair[0]}")
)
src_cache_decompressed_dp = (
Expand All @@ -89,7 +91,7 @@ def Multi30k(root: str, split: Union[Tuple[str], str], language_pair: Tuple[str]
)
src_cache_decompressed_dp = src_cache_decompressed_dp.end_caching(mode="wb", same_filepath_fn=True)

tgt_cache_decompressed_dp = cache_compressed_dp.on_disk_cache(
tgt_cache_decompressed_dp = cache_compressed_dp_2.on_disk_cache(
filepath_fn=lambda x: os.path.join(root, f"{_PREFIX[split]}.{language_pair[1]}")
)
tgt_cache_decompressed_dp = (
Expand Down