Repository navigation
contilab v0.3.0
Added
- Release workflow: pushing a tag
vX.Y.Z(or running the workflow with a tag) builds the sdist and wheel and
publishes a GitHub release with that version's CHANGELOG section as notes. Its PyPI job uses trusted publishing
and stays off until thePUBLISH_PYPIrepository variable is set (seeRELEASING.md). contilab bench --precision(defaultfp32,auto): throughput at the precision the profiles actually use.- A group that runs out of memory is re-run as two smaller groups (recorded as
memory_splitinenv.json).
A strategy that cannot be vmapped runs member by member, recorded asvmap_fallback. - Split CIFAR-100 (Class-IL, 10 × 10 classes) in the
cpuprofile and incpu_main.cpu_main_seedsadds
seeds 3–5 to the three MNIST benchmarks, so exact paired tests there can reach p < 0.05.
Fixed
torch.compileno longer recompiles for every batch size. The batch axis is marked dynamic, so A-GEM's
reference batches, filling replay buffers and the last batch of a task share one graph. On CPU, each ResNet
recompilation took about 25–50 s, and A-GEM runs hit Dynamo's limit of 8 graphs and then trained eagerly.- LoRA-per-task stays compiled beyond 8 tasks: the graph limit per compiled function is now the number of
tasks + 8.
Changed
- Less repeated work:
- the class table is cached per stream and device
- LwF runs one student forward per step instead of two
- distillation reuses the run's data pool
- probe-based metrics fit the probe matrix once
- Results (wiki and README) cover 6 benchmarks and 324 runs, with exact paired tests over 6 seeds on MNIST.
DER++ now has the best mean on Split MNIST Class-IL, not significantly different from multi-teacher
distillation.
Results and documentation: wiki · Results
Install
pip install https://github.com/JonPark0/contilab/releases/download/v0.3.0/contilab-0.3.0-py3-none-any.whl
# or from source: git clone --branch v0.3.0 https://github.com/JonPark0/contilab && pip install -e ./contilab