Primus-SaFE(Stability and Fault Endurance)
-
Updated
Sep 18, 2026 - Go
#
training-stability
Here are 10 public repositories matching this topic...
Code and data of the EMNLP 2022 Main Conference paper "Reduce Catastrophic Forgetting of Dense Retrieval Training with Teleportation Negatives".
information-retrieval catastrophic-forgetting hard-negative-mining dense-retrieval training-stability
-
Updated
Mar 25, 2024 - Python
Global L2 norm adaptive gradient clipping engine to mitigate exploding gradients during deep neural network training.
deep-learning gradient-clipping l2-norm training-stability agentic-ai genpark-skill exploding-gradients
-
Updated
Sep 10, 2026 - Python
Global L2 norm adaptive gradient clipping engine to mitigate exploding gradients during deep neural network training.
deep-learning gradient-clipping l2-norm training-stability agentic-ai genpark-skill exploding-gradients
-
Updated
Sep 10, 2026 - Python
PyTorch NaNs are silent killers. This hook catches them at the exact layer and batch — with ~3 ms overhead vs ~7 ms for set_detect_anomaly.
pytorch autograd model-debugging training-stability gradient-explosion nan-detection deep-learning-debugging forward-hooks
-
Updated
Apr 30, 2026 - Python
Drift-Aware Adaptive Aggregation (DAA) for federated learning on CIFAR-10 under heterogeneous client partitions.
machine-learning deep-learning optimization heterogeneous-data federated-learning non-iid robust-federated-learning training-stability model-aggregation
-
Updated
Mar 15, 2026 - Python
RVAV: a physics-informed PyTorch optimizer for energy-stable, high-LR training—with a simple closure API, tests, CI, and quickstart.
benchmarking machine-learning deep-learning ci optimizer pytorch scientific-computing reproducibility numerical-optimization physics-informed energy-based training-stability large-learning-rates closure-api
-
Updated
Aug 18, 2025 - Python
An unofficial extended version for ICLR 2021"Gradient Descent on Neural Networks Typically Occurs at the Edge of Stability"
-
Updated
Nov 21, 2023 - Python
Benchmarking GAN optimizers (Adam, RMSprop, SGD, Lookahead) on CIFAR-10 using WGAN-GP and FID evaluation.
python benchmark machine-learning deep-learning optimizer pytorch generative-adversarial-network gan sgd neural-networks image-generation cifar10 adam rmsprop lookahead wgan-gp training-stability
-
Updated
Jul 12, 2025 - Jupyter Notebook
nlp deep-learning simulation pytorch research-tool language-model early-stopping pretraining llm training-stability divergence-detection loss-dynamics
-
Updated
Apr 9, 2026 - Python
Add this topic to your repo
To associate your repository with the training-stability topic, visit your repo's landing page and select "manage topics."