arXiv:2607. 19058v1 Announce Type: cross Abstract: Optimizer state is the largest single line item in the memory budget of mixture-of-experts (MoE) training: on a 6.
arXiv:2608. 04407v1 Announce Type: cross Abstract: Memory-efficient matrix optimizers such as Sinkhorn gradient descent remove most AdamW optimizer state for dense Transformer matrices, but direct application to Mixture-of-Experts (MoE) training is unreliable.
arXiv:2609.36222v1 Announce Type: new Abstract: Large language models are increasingly expensive to serve. In large-scale serving systems, autoregressive decoding is often bottlenecked by transferrin...
The paper investigates whether training Mixture-of-Experts (MoE) routers can improve memory‑bandwidth locality on consumer GPUs. Using a new zero‑surgery telemetry tool, the authors measure that a large Qwen3‑235B model is bottlenecked by disk‑based expert access, and that an LRU cache can serve a majority of requests. They pre‑register experiments training 137 M‑parameter MoE models with locality‑aware losses, finding that while cache misses can drop up to 60 % (99 % static‑pin hit rate), every configuration fails to meet a strict 1 % perplexity threshold, indicating a tight coupling between cache efficiency and model quality.
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:2608. 20210v1 Announce Type: cross Abstract: Small language models are usually built like large ones and then squeezed onto a CPU afterwards.
arXiv:2606. 07404v1 Announce Type: new Abstract: This paper reports on training a hundred-billion-parameter sparse mixture of experts on a single eight-GPU node, end to end.
The study re‑examines a reported advantage of a routed ternary (1.58‑bit) language model over a full‑precision transformer at 60K parameters. By running controlled experiments with multiple seeds and a fixed training recipe, the authors find that the apparent benefit largely stems from the choice of baseline model shape rather than the ternary architecture itself. While the routed model does outperform other shapes at a larger 130M‑byte budget, its advantage diminishes when a plain gated diagonal‑SSM block is used, and the ternary penalty varies with architecture and quantization details.
arXiv:2609.08115v1 Announce Type: new Abstract: Mixture-of-Experts (MoE) pretraining relies on an auxiliary load-balancing loss (LBL) to drive per-expert utilization toward uniformity. Post-training...
The paper introduces Edge0, a streaming mixture‑of‑experts (MoE) inference engine that enables a 35‑billion‑parameter MoE model to run on consumer hardware by predicting routing decisions one token ahead. Edge0 uses a per‑layer prerouter to prefetch the necessary experts from SSD, and an unmerged recovery LoRA trained on the student path to recover quality lost to 4‑bit quantization and routing replacement. On a single 24‑GB machine, Edge0 serves the 35B MoE at 20 tokens per second while keeping peak active memory below 3 GiB, achieving performance close to its fp16 teacher across five public benchmarks.
Serving a 235B-parameter Mixture-of-Experts (MoE) model on a single 8 GB GPU is bottlenecked not by compute but by memory bandwidth: decode must stream each token's active experts from whichever tier holds them, and on consumer hardware most experts sit on an SSD far slower than RAM. We quantify this bandwidth wall on Qwen3-235B (Q4_K_M, 134 GB): measured decode is 0.
Small language models are usually built like large ones and then squeezed onto a CPU afterwards. We did the opposite: we fixed the target first, one user, one token at a time, 4-bit weights, ordinary CPU, and chose the architecture to suit it.
arXiv:2609.13922v1 Announce Type: new Abstract: Minibatch persistency reuses data instead of reading it: rather than drawing a fresh minibatch at every optimizer step, it takes K consecutive steps on...