Contlearn
C 62 completed
Library
cli / python · small
121
Files
21,937
LOC
1
Frameworks
5
Languages
Pipeline State
completedRun ID
#307439Phase
doneProgress
1%Started
Finished
2026-04-13 01:31:02LLM tokens
0Pipeline Metadata
Stage
SkippedDecision
skip_scaffold_dupNovelty
51.47Framework unique
—Isolation
—Last stage change
2026-04-16 18:15:42Deduplication group #47941
Member of a group with 1 similar repo(s) — canonical #9446 view group →
Top concepts (9)
Project DescriptiontestingTestingFactorySearchTestingDatabaseConfigurationAnalytics
Repobility (the analyzer behind this table) · https://repobility.com
🧪 Code Distillation
Browse all specs →Sample distilled functions (click for full spec)
create_awb_operationsDetermines and constructs the correct operations object for an AWB pipeline based on the structure of an input model. It accepts a model object, which is expected to be an Equinox model type. The function returns an instance of MLPAWBOps, CNNAWBOps, or GCNAWBOps, depending on which specific attribut
evaluate_on_loaderCalculates the average performance metric by iterating over a provided data loader. It accepts a trainer instance, model parameters, static components, the data loader, and a problem type string. Internally, it conditionally transforms the batch data—converting PyTorch tensors to JAX arrays for vect
compute_task_lrComputes the appropriate learning rate for a given task based on a specified decay schedule. It accepts a configuration dictionary, the current task identifier, and an optional loss ratio. The function returns a floating-point learning rate, which is guaranteed to be at least the adaptively calculat
AI Prompt
Create a command-line tool for continual learning in machine learning models. I need it to support various architectures like MLP, CNN, and GCN. The core functionality should utilize Hamiltonian gradients and Adaptive Weight Basis (AWB) for knowledge transfer. The tool should allow running experiments using JSON configuration files, supporting different learning rate schedules, and tracking metrics. Please structure it so I can run examples like sine regression or MNIST classification directly from the command line.
python cli machine-learning continual-learning jax pytorch gradient cnn gcn
Generated by gemma4:latest
Catalog Information
The contlearn project enables continual learning in machine learning models using Hamiltonian gradients and adaptive weight basis.
Description
Contlearn is a Python library that implements continual learning techniques, specifically using Hamiltonian gradients and adaptive weight basis (AWB). This approach allows neural networks to learn from new tasks without forgetting previously learned information. The project leverages popular deep learning frameworks such as PyTorch and JAX to provide efficient and scalable implementation.
الوصف
يعد مشروع contlearn مكتبة برمجة لتعلم المستمر في النماذج الحاسوبية باستخدام gradients هاميلتونيان و basis وزن متكيف (AWB). يتيح هذا النهج للنماذج العصبونية التعلم من المهام الجديدة دون فقدان المعلومات السابقة. يستفيد المشروع من إطاريات التعلم العميق الشهيرة مثل PyTorch و JAX لتقديم تنفيذ فعال ومستدام.
Novelty
7/10Tags
continual-learning machine-learning neural-networks deep-learning adaptive-weight-basis hamiltonian-gradients
Technologies
jax matplotlib numpy pytorch scikit-learn
Claude Models
claude-opus-4.5 claude-sonnet-4.5
Quality Score
C
62.5/100
Structure
75
Code Quality
43
Documentation
80
Testing
70
Practices
46
Security
72
Dependencies
90
Strengths
- Good test coverage (200% test-to-source ratio)
- Code linting configured (ruff (possible))
- Consistent naming conventions (snake_case)
Weaknesses
- No LICENSE file \u2014 legal ambiguity for contributors
- No CI/CD configuration \u2014 manual testing and deployment
- 9 bare except/catch blocks swallowing errors
- 2228 duplicate lines detected \u2014 consider DRY refactoring
- 7 'god files' with >500 LOC need decomposition
Recommendations
- Set up CI/CD (GitHub Actions recommended) to automate testing and deployment
- Add a LICENSE file (MIT recommended for open source)
- Replace bare except/catch blocks with specific exception types
Security & Health
5.8h
Tech Debt (A)
Medium
DORA Rating
A
OWASP (100%)
Repobility · severity-and-effort ranking · https://repobility.com
PASS
Quality Gate
A
Risk (1)
MIT
License
12.0%
Duplication
Languages
Frameworks
pytest
Symbols
method204
constant160
function113
variable77
class40
property29
Concepts (9)
| Category | Name | Description | Confidence | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Want fix-PRs on findings? Install Repobility's GitHub App · github.com/apps/repobility-bot | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| auto_description | Project Description | A JAX/Equinox framework for continual learning using Hamiltonian-based gradient computation and Adaptive Weight Basis (AWB) for architecture adaptation. | 80% | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| arch_layer | testing | Detected testing layer | 70% | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| auto_category | Testing | testing | 70% | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| design_pattern | Factory | Found factory/create_ naming patterns | 60% | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| business_logic | Search | Detected from 9 related files | 50% | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| business_logic | Testing | Detected from 72 related files | 50% | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| business_logic | Database | Detected from 6 related files | 50% | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| business_logic | Configuration | Detected from 33 related files | 50% | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| business_logic | Analytics | Detected from 18 related files | 50% | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Repobility — same analyzer, your code, free for public repos · /scan/
Embed Badge
Add to your README:
BinComp Dependency Hardening
All packages →5 of this repo's dependencies have been scanned for binary hardening. Grade reflects RELRO / stack canary / FORTIFY / PIE coverage.