Metraipolis › archive › dc_ae742bf8632d
Block Parallelism For Efficient Distributed Long-Context Diffusion Language Model Training — arXiv:2609.19242
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.19242
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-10T04:08:34.223Z
- SHA-256 of the archived text
12ed31b4b8773caf6855311e8cb1ce4ba700b606ac6f21ba1ed905e01f6d61dc- Stable URL of this page
https://metraipolis.com/documento/dc_ae742bf8632d
How to cite
Block Parallelism For Efficient Distributed Long-Context Diffusion Language Model Training — arXiv:2609.19242 — arXiv (en), https://arxiv.org/abs/2609.19242 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_ae742bf8632d (Metraipolis archive dc_ae742bf8632d, sha256 12ed31b4b8773caf6855311e8cb1ce4ba700b606ac6f21ba1ed905e01f6d61dc)
BibTeX
@misc{dcae742bf8632d,
title = {Block Parallelism For Efficient Distributed Long-Context Diffusion Language Model Training — arXiv:2609.19242},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.19242},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_ae742bf8632d; sha256 12ed31b4b8773caf6855311e8cb1ce4ba700b606ac6f21ba1ed905e01f6d61dc},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.19242. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Block diffusion language models (BDLMs) combine autoregressive dependencies across blocks with parallel denoising within blocks, but long-context training is constrained by distributed attention communication and activation memory. Conventional context parallelism (CP) shards the combined clean-plus-corrupted sequence by position, communicating shared clean K/V together with block-specific corrupted K/V and their gradients. We observe that the BDLM objective separates over target blocks. We introduce block parallelism (BP), a new distributed parallelism dimension that assigns each corrupted-block computation to one rank. To scale BP to long contexts, we introduce context-sharded block parallelism (CSBP), which also shards the shared clean sequence across those ranks. CSBP keeps corrupted K/V and gradients local, avoids replicated clean prefixes, and preserves BDLM training semantics. On 16 H200 GPUs at 256K context, CSBP improves throughput over the best baseline by 1.18-1.45x for supervised fine-tuning and 1.27-1.33x for conversion of autoregressive models to BDLMs, while matching or reducing peak HBM. Full-model speedup reaches 1.61x at 512K. On eight H100 GPUs, CSBP accelerates DFlash2 speculative-decoder training by 2.48x at 512K and 7.59x at 1M. In matched 12-hour DiffusionGemma 26B-A4B SFT runs, CSBP achieves higher pass rates at every trained checkpoint on SWE-bench Verified and Terminal-Bench Lite. Code: this https URL Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-18 in cs.LG). Categorie arXiv: Machine Learning (cs.LG). Pagina del lavoro: https://arxiv.org/abs/2609.19242 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-10-10T03:48:18.239Z dalla pagina https://arxiv.org/catchup/cs.LG/2026-09-18?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).