CogitoNTNU/DeepTactics-Muzero
Feb 2025 – Apr 2025
3 Months active
Languages Used
PythonNumpyShell
Technical Skills
Algorithm ImplementationBackend DevelopmentBuffer ManagementData StructuresGame DevelopmentReinforcement Learning
Vegardhgr
PROFILE
Vegardhgr
Worked on the CogitoNTNU/DeepTactics-Muzero repository to build and refine the core data pipeline for scalable self-play reinforcement learning. Developed a Python-based ReplayBuffer for storing and retrieving game trajectories, ensuring robust data integrity and efficient training data collection. Introduced orchestration utilities to coordinate multiple games and manage network versions, enabling iterative training cycles. Enhanced reliability through comprehensive unit testing and refactoring, improving maintainability and CI readiness. Established a reproducible cluster execution workflow by adding SLURM job scripts for GPU-backed runs and automating environment setup. Demonstrated strong skills in Python, backend development, buffer management, and high-performance computing environments.
Overall Statistics
Feature vs Bugs
80%Features
Repository Contributions
7Total
Bugs
1
Commits
7
Features
4
Lines of code
396
Activity Months3
Your Network
12 peopleWork History
April 2025
1 Commits • 1 Features
Apr 1, 2025April 2025 monthly summary for CogitoNTNU/DeepTactics-Muzero focused on delivering cluster-enabled execution for GPU-accelerated workloads and setting up a reproducible environment for the Python-based application.
1 Commits • 1 Features
Apr 1, 2025April 2025 monthly summary for CogitoNTNU/DeepTactics-Muzero focused on delivering cluster-enabled execution for GPU-accelerated workloads and setting up a reproducible environment for the Python-based application.
April 2025
March 2025
3 Commits • 1 Features
Mar 1, 2025March 2025 (2025-03) - CogitoNTNU/DeepTactics-Muzero delivered foundational work to enable scalable training workflows and improved core data reliability for model training. A Self-Play Training Loop Placeholder was introduced to prepare for future integration of a training mechanism into the gameplay loop, and the ReplayBuffer was strengthened with robust history handling, correct next-state selection, and comprehensive tests. Refactoring cleaned up unused methods and aligned tests with the new structure, improving maintainability and CI readiness. These changes establish a stable, testable foundation for upcoming training iterations and reduce risk in the data pipeline.
March 2025
3 Commits • 1 Features
Mar 1, 2025March 2025 (2025-03) - CogitoNTNU/DeepTactics-Muzero delivered foundational work to enable scalable training workflows and improved core data reliability for model training. A Self-Play Training Loop Placeholder was introduced to prepare for future integration of a training mechanism into the gameplay loop, and the ReplayBuffer was strengthened with robust history handling, correct next-state selection, and comprehensive tests. Refactoring cleaned up unused methods and aligned tests with the new structure, improving maintainability and CI readiness. These changes establish a stable, testable foundation for upcoming training iterations and reduce risk in the data pipeline.
February 2025
3 Commits • 2 Features
Feb 1, 2025February 2025 – CogitoNTNU/DeepTactics-Muzero: Delivered the core data collection and orchestration components to enable scalable self-play training. Implemented a Python ReplayBuffer to store trajectories (states, actions, rewards, policies, values) with update/retrieval, refined for robust storage and data quality, enabling efficient training data collection. Added a self-play script and SharedStorage utility to coordinate multiple games, feed game data into the buffer, and manage network versions for iterative training. Fixed key issues in the replay buffer to improve data integrity and reliability of the training data pipeline. These changes establish an end-to-end data pipeline foundation, unlocking faster, more reliable training iterations and demonstrating strong Python engineering, data pipeline design, and system coordination skills.
3 Commits • 2 Features
Feb 1, 2025February 2025 – CogitoNTNU/DeepTactics-Muzero: Delivered the core data collection and orchestration components to enable scalable self-play training. Implemented a Python ReplayBuffer to store trajectories (states, actions, rewards, policies, values) with update/retrieval, refined for robust storage and data quality, enabling efficient training data collection. Added a self-play script and SharedStorage utility to coordinate multiple games, feed game data into the buffer, and manage network versions for iterative training. Fixed key issues in the replay buffer to improve data integrity and reliability of the training data pipeline. These changes establish an end-to-end data pipeline foundation, unlocking faster, more reliable training iterations and demonstrating strong Python engineering, data pipeline design, and system coordination skills.
February 2025
Activity
Loading activity data...
Quality Metrics
Correctness75.8%
Maintainability77.2%
Architecture71.4%
Performance65.6%
AI Usage20.0%
Skills & Technologies
Programming Languages
NumpyPythonShell
Technical Skills
Algorithm ImplementationBackend DevelopmentBuffer ManagementData StructuresEnvironment SetupGame DevelopmentHPCPythonReinforcement LearningSLURMTesting
Repositories Contributed To
Overview of all repositories you've contributed to across your timeline