The paper introduces Batch Cost Lookahead (BCL), a coordination‑free data‑loading strategy that buffers a bounded window of records per rank and repartitions them into fixed‑size packs with balanced estimated cost. BCL maintains the trainer’s record‑count contract without requiring cross‑rank communication or runtime changes, and in simulations it reduces per‑rank batch‑cost variability by up to 82.8%. Experiments on 13 NeMo‑RL SFT corpora across eight H200 GPUs show speedups ranging from 1.050× to 1.167× over pooled MFFD, with further gains up to 1.343× on multi‑node setups, while also identifying a memory‑feasibility boundary that limits outliers before balancing.
arXiv:2609.37899v1 Announce Type: new Abstract: Zero-order optimization (ZO) trains without backpropagation, making it relevant to forward-only hardware and non-differentiable loss, but its gradient...
arXiv:2606. 19989v1 Announce Type: cross Abstract: Modern LLM training breaks a core assumption behind offline batch samplers: the true training cost of a sample is only observable after preprocessing, augmentation, templating, tokenization, and multimodal visual-token expansion.
The paper introduces an open midtraining recipe that primes base models for agentic post‑training, filling a gap between pretraining and full fine‑tuning. It trains on 256K context with hundreds of billions of tokens drawn from natural software‑engineering data and a small share of agentic traces. Applying this midtraining to Nemotron 3.5 Lightning and Super models yields measurable gains on SWE‑bench and DeepSWE benchmarks after identical full SFT, and improves robustness across agent harnesses.
Zephon is a data loader designed for foundation model training that ensures deterministic ordering of training data batches even when GPU resources change, checkpoints are resumed, or different processing backends are used. It handles online, stateful pipelines—where tokenization, packing, and mixing of samples create complex n‑to‑m transformations—by partitioning the data stream into topology‑independent lanes, serializing ordering decisions, and checkpointing only bounded in‑flight state. Experiments on text and vision‑language tasks show Zephon delivers competitive throughput while offering guarantees that existing loaders lack for such pipelines.
arXiv:2608. 19488v1 Announce Type: new Abstract: Production machine learning systems degrade under concept drift, yet practitioners have little principled guidance on when to retrain.
arXiv:2606. 04101v1 Announce Type: cross Abstract: Large-scale expert parallelism (EP) is becoming pivotal for training and serving frontier MoE models, but it also amplifies device-level expert load imbalance into compute stragglers, token all-to-all bottlenecks, and activation-memory spikes.
NeMo-DCR is a delta‑compressed refit method that sends only the changes in a trillion‑parameter policy model, achieving bit‑exact synchronization across AWS regions. By using fixed affine mappings, XOR masks, and a relay‑tree object storage, it avoids full checkpoint transfers and cross‑cluster collectives, reducing refit times from 87.5 min to 150 s for a 1 T model at 3% change rate. The approach scales to 30 B–1 T models, delivering 12–40× speedups over traditional full‑checkpoint refits.
arXiv:2610.19975v1 Announce Type: new Abstract: Looped Mixture-of-Experts (MoE) models combine parameter reuse with sparse computation to improve performance under matched budgets. Prior work demonst...
arXiv:2606. 06302v2 Announce Type: replace Abstract: Multi-turn LLM serving accumulates dialogue history whose Key-Value (KV) cache grows with every turn and every user, quickly exceeding the model weights themselves and making memory -- not compute -- the binding constraint on throughput.
arXiv:2604. 26968v2 Announce Type: replace-cross Abstract: Key-value (KV) cache memory management is the primary bottleneck limiting throughput and cost-efficiency in large-scale GPU inference serving.
arXiv:2606. 19004v1 Announce Type: cross Abstract: Reinforcement learning (RL) post-training of Diffusion Transformers (DiTs) is prohibitively expensive, requiring thousands of high-end GPUs.
arXiv:2610.19822v1 Announce Type: cross Abstract: LLM agents are evolving from a single model calling tools to planners that orchestrate multiple expert models whose capabilities live in distinct wei...