neptune_ai
Experiment tracker purpose-built for foundation model training.
Monitor thousands of per-layer metrics—losses, gradients, and activations—at any scale. Visualize them with no lag and no missed spikes. Drill down into logs and debug training issues fast. Keep your model training stable while reducing wasted GPU cycles.
What Researchers Told Us at ICML: Experiment Tracking Nightmares
We Tried Training Foundation Models — From Pretraining to Eval, Fully Tracked in Neptune
Inside Bioptimus: Training (and Tracking) Foundation Models for Biology
neptune.ai Review – One source of truth for experiments | Hiren @ Vicon
neptune.ai Review – Powerful experiments comparison | Nikola @ ETH
neptune.ai Review – Neptune vs competitors | Austin @ Weill Cornell Medicine
neptune.ai Review – A Tracker for Deep Learning research | Filippo @ UiT
neptune.ai Review – Monitoring of GPU utilization | Hiren @ Vicon
neptune.ai Review – One source of truth for the team | Eugene @ ZOZO
SKEWACT: Red-Teaming LLMs via Activation-Skewed Adversarial Prompt Optimization
neptune.ai Review – Controlled access to projects | Filippo @ UiT
neptune.ai Review – Integration, onboarding & support | Hiren @ Vicon
neptune.ai Review – Logging from different environments | Aaron @ poolside
Exploration of LLM-Guided Conversation in Reminiscence Therapy
MMed-RAG: Versatile Multimodal RAG System for Med-LVLMs
neptune.ai Review – Monitoring large-scale pre-training | Carlos @ poolside
Efficient Asynchronous Parallel SGD in the Presence of Heterogeneous and Random Worker Compute Times
Optimizing Attention
Scalable Data Selection for Fine-tuning LLMs
Instruction Tuning LLMs to Understand Electronic Health Records
Demo: Fork Runs From Stable Checkpoints After Spotting Training Issues
Demo: Identify and Debug Issues in Foundation Model Traning
Demo: Monitor Large-Scale Long Training Jobs in Real Time
Demo: Find the Best Model for the Long Training (From Hundreds of Experiments)
Generating Diverse Negations from Affirmative Sentences
Instant Transformer Adaption via HyperLoRA
Interpretable Zero-Shot Predictions on Clinical Tabular Data
Adaptive LoRA Merging for Efficient Domain Incremental Learning
Realistic Threat Model for LLM Jailbreaks
Monitor & Debug Foundation Model Training – neptune.ai Demo