Skip to content

fix(serialization): Slice 1-D multibyte data as bytes for pwrite #181

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 4 commits into from
Nov 28, 2024
Merged

fix(serialization): Slice 1-D multibyte data as bytes for pwrite #181

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
9d35e68
fix(serialization): Slice 1-D multibyte data as bytes for `pwrite`
Eta0 Nov 25, 2024
11231b6
fix(serialization): Don't accumulate `os.pwrite` results <= 0
Eta0 Nov 27, 2024
a6f7a23
docs: Update changelog to mention 1-dimensional tensor serialization fix
Eta0 Nov 28, 2024
b1c7873
chore: Update code version to 2.9.1, and update the changelog
Eta0 Nov 28, 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
6 changes: 4 additions & 2 deletions CHANGELOG.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,10 +5,12 @@ All notable changes to this project will be documented in this file.
The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).

## [Unreleased]
## [2.9.1] - 2024-11-27

### Fixed

- `TensorSerializer` no longer sometimes fails to serialize very large
1-dimensional tensors with multibyte `dtype`s
- `RedisStreamFile.readable()` and `RedisStreamFile.seekable()` now correctly
return `True`

Expand Down Expand Up @@ -407,7 +409,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
- `get_gpu_name`
- `no_init_or_tensor`

[Unreleased]: https://github.com/coreweave/tensorizer/compare/v2.9.0...HEAD
[2.9.1]: https://github.com/coreweave/tensorizer/compare/v2.9.0...v2.9.1
[2.9.0]: https://github.com/coreweave/tensorizer/compare/v2.8.1...v2.9.0
[2.8.1]: https://github.com/coreweave/tensorizer/compare/v2.8.0...v2.8.1
[2.8.0]: https://github.com/coreweave/tensorizer/compare/v2.7.2...v2.8.0
Expand Down
2 changes: 1 addition & 1 deletion tensorizer/_version.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1 +1 @@
__version__ = "2.9.0"
__version__ = "2.9.1"
8 changes: 5 additions & 3 deletions tensorizer/serialization.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3510,7 +3510,7 @@ def _mv_suffix(data: "collections.abc.Buffer", start: int):
if not isinstance(data, memoryview):
data = memoryview(data)
try:
if data.ndim != 1:
if data.ndim != 1 or data.format != "B":
data = data.cast("B")
return data[start:]
finally:
Expand All @@ -3526,13 +3526,15 @@ def _pwrite_syscall(
verify if isinstance(verify, int) else self._buffer_size(data)
)
bytes_just_written: int = os.pwrite(self._fd, data, offset)
bytes_written += bytes_just_written
if bytes_just_written > 0:
bytes_written += bytes_just_written
while bytes_written < expected_bytes_written and bytes_just_written > 0:
# Writes larger than ~2 GiB may not complete in a single pwrite call
offset += bytes_just_written
with self._mv_suffix(data, bytes_written) as mv:
bytes_just_written = os.pwrite(self._fd, mv, offset)
bytes_written += bytes_just_written
if bytes_just_written > 0:
bytes_written += bytes_just_written
if isinstance(verify, int) or verify:
self._verify_bytes_written(bytes_written, expected_bytes_written)
return bytes_written
Expand Down
Loading