Stuck on an issue?

Lightrun Answers was designed to reduce the constant googling that comes with debugging 3rd party libraries. It collects links to all the places you might be looking at while hunting down a tough bug.

And, if you’re still stuck at the end, we’re happy to hop on a call to see how we can help out.

Core50 benchmark with ICARL strategy

See original GitHub issue

ICarl strategy returns an error with Core50 benchmark: scenario = CORe50(scenario="nc", run=9)

The error is: AttributeError: 'GenericCLScenario' object has no attribute 'n_classes_per_exp'

Other nc benchmarks belong to “NCScenario” class and this issue does not appear.

Issue Analytics

State:
Created 2 years ago
Comments:6 (2 by maintainers)

Top GitHub Comments

1reaction

AntonioCartacommented, Oct 18, 2021

As a quick fix, you can easily compute n_classes_per_exp yourself by using exp.dataset.targets.

0reactions

Melika-Ayoughicommented, Sep 16, 2022

I have the same problem with ego objects benchmark and icarl. Is it still not fixed?

Top Results From Across the Web

Continuous Learning Benchmarks - CORe50 - GitHub Pages

Continuous Learning Benchmarks. In this constantly updated page we keep track and categorize the most popular benchmarks for Continuous Learning (CL).

Accuracy results on the CORe50 NICv2 -391 benchmark of ...

In Figure 6, we show the iCaRL accuracy over time and compare it with AR1*free (conv5 4/dw), AR1* (pool6) as well as the...

CORe50: a New Dataset and Benchmark for Continuous ...

In the context of real-world object recognition applications (e.g., robotic vision), where continuous learning is crucial, very few datasets and ...

arXiv:2203.10317v1 [cs.LG] 19 Mar 2022 - Zenodo

sively compare replay-based strategies on different benchmarks and settings to ... ICarl. Fig. 4. CORe50-NC memory-accuracy.

avalanche.benchmarks.classic.CORe50

Creates a CL benchmark for CORe50. If the dataset is not present in the computer, this method will automatically download and store it....