Release 0.9.2

0.9.2 Release

We're excited to announce the release of Furiosa SDK version 0.9.2. This is a hotfix release including a bug fix.

Changelog

• fix: assume the model is nuxmodel when no config (5b85aad)

You can find the full changelogs at v0.9.1...v0.9.2.

Documents

• 0.9.x Release Notes
• 0.9.x Upgrade Instruction
• 0.9.x SDK Document (English)
• 0.9.x SDK Document (Korean)

Installation

pip install furiosa-sdk==0.9.2

Release 0.9.1

0.9.1 Release

We're excited to announce the release of Furiosa SDK version 0.9.1. This is a minor release. You can find the full changelogs at v0.9.0...v0.9.1.

• 0.9.x Release Notes
• 0.9.x Upgrade Instruction
• 0.9.x SDK Document (English)
• 0.9.x SDK Document (Korean)

Installation

pip install furiosa-sdk==0.9.1

Release 0.9.0

0.9.0 Release

We're excited to announce the release of Furiosa SDK version 0.9.0. This release includes a number of important updates and new features, including. More details about 0.9.0 release can be found below:

• 0.9.0 Release Notes
• 0.9.0 Upgrade Instruction
• 0.9.0 SDK Document (English)
• 0.9.0 SDK Document (Korean)

Installation

pip install furiosa-sdk==0.9.0

Release 0.8.2

Full Changelog: v0.7.0...0.8.2

Release 0.8.0

Set the version to 0.8.0

Release 0.6.2

Installation

pip install --upgrade 'furiosa-sdk==0.6.2'

Change logs

Changed

• Fix the missing git_version files and README.md in some modules #318
• litmus should use furiosa-tools compile rather than session.create() #317
• More abundant version information in CLI tools #320

Fixed

• NAVL-11: Infer squeeze axes #321
• NAVL-9: case when bias scale is zero #315

Release 0.6.0

Change logs

• Reorganize the python structure
  • furiosa-sdk-runtime -> furiosa-sdk
  • furiosa-sdk-quantizer -> furiosa-quantizer
  • furiosa-sdk-validator -> furiosa-litmus
• Add furiosa-serving (FastApi based advanced serving API library)
  • Please checkout https://github.com/furiosa-ai/furiosa-sdk/tree/main/python/furiosa-serving
• (Quantizer) Guarantee Idempotence when applying quantization repeatedly
• (Quantizer) Remove PyTorch dependency
• (Quantizer) Improve code quality by cleaning up many Pylint warnings
• (Quantizer) Update dependencies to their latest versions (e.g., Numpy -> 1.21.5, Pyyaml -> 6.0.0)

Installation

pip uninstall furiosa-sdk-quantizer furiosa-sdk-runtime furiosa-sdk-validator && \
pip install --upgrade 'furiosa-sdk==0.6.0'

Release 0.5.2

Bug fix:

• Fix the constraints of Conv + Add pattern (7042f7c)

Installation

pip install --upgrade 'furiosa-sdk[full]==0.5.2'

Release 0.5.1

Bug fix:

• rewrite Pattern_3.pattern_condition_checker to filter (4e993b7)

Installation

pip install --upgrade 'furiosa-sdk[full]==0.5.1'

Release 0.5.0

v0.5.0

Reformat tests/tests_compile.py