Skip to content
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

Jump to bottom

CU-8694hukwm: Document the materialising of generator when multiproce… #433

Merged
merged 3 commits into from
May 22, 2024
Merged

CU-8694hukwm: Document the materialising of generator when multiproce… #433

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
e95e3a5
CU-8694hukwm: Document the materialising of generator when multiproce…
mart-r May 13, 2024
0728626
CU-8694hukwm: Add TODO note for where the generator is materialised
mart-r May 13, 2024
82d705d
CU-8694hukwm: Add warning from large amounts of generator data (10k i…
mart-r May 14, 2024
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
22 changes: 21 additions & 1 deletion medcat/cat.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -49,6 +49,8 @@

HAS_NEW_SPACY = has_new_spacy()

MIN_GEN_LEN_FOR_WARN = 10_000


class CAT(object):
"""The main MedCAT class used to annotate documents, it is built on top of spaCy
Expand Down Expand Up @@ -1526,6 +1528,11 @@ def multiprocessing_batch_docs_size(self,

This method batches the data based on the number of documents as specified by the user.

NOTE: When providing a generator for `data`, the generator is evaluated (`list(in_data)`)
and thus all the data is kept in memory and (potentially) duplicated for use in
multiple threads. So if you're using a lot of data, it may be better to use
`CAT.multiprocessing_batch_char_size` instead.

PS:
This method supports Windows.

Expand All @@ -1550,7 +1557,20 @@ def multiprocessing_batch_docs_size(self,
if nproc == 0:
raise ValueError("nproc cannot be set to zero")

in_data = list(in_data) if isinstance(in_data, Iterable) else in_data
# TODO: Surely there's a way to not materialise all of the incoming data in memory?
# This is counter productive for allowing the passing of generators.
if isinstance(in_data, Iterable):
in_data = list(in_data)
in_data_len = len(in_data)
if in_data_len > MIN_GEN_LEN_FOR_WARN:
# only point this out when it's relevant, i.e over 10k items
logger.warning("The `CAT.multiprocessing_batch_docs_size` method just "
f"materialised {in_data_len} items from the generator it "
"was provided. This may use up a considerable amount of "
"RAM, especially since the data may be duplicated across "
"multiple threads when multiprocessing is used. If the "
"process is kiled after this warning, please use the "
"alternative method `multiprocessing_batch_char_size` instead")
n_process = nproc if nproc is not None else min(max(cpu_count() - 1, 1), math.ceil(len(in_data) / batch_factor))
batch_size = batch_size if batch_size is not None else math.ceil(len(in_data) / (batch_factor * abs(n_process)))

Expand Down
Loading