Skip to content

Commit e3af9ed

Browse files
JamesBrianDclaude
andauthored
[blog]Serving Ling-2.6-1T on TPU with SGLang-JAX (#348)
* blog: add Ling-2.6 TPU serving post * blog: revise Ling-2.6 TPU post — title, language polish, structure - Retitle to highlight the Pallas fused MoE kernel core (hiding data movement behind compute) - Fix non-native phrasing, dangling modifiers, and naming consistency (Fused MoE V1/V2, FusedEPMoE, fp8, hidden-dimension slices) - Credit Fused MoE V1 authors (tpu-inference) and add SGLang-JAX adapted-kernel reference; renumber references - Move full TPU-vs-GPU comparison (incl. prefill gap) to benchmarks next to the GLA prefill note; renumber figures - Deduplicate: GLA section, memory pools, DP, future-work bullets, AIME result; fold Accuracy section into appendix - Clarify measurement scopes (in-kernel vs standalone all-to-all) and per-device vs per-chip specs; restore TPU v7x spec note in appendix Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> * blog: update Ling-2.6 TPU post acknowledgments Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> * blog: address review feedback on Ling-2.6 TPU post - Remove em-dashes from prose (use commas/colons/semicolons), per review - Spell out MoE/MXU/VPU on first use; add TPU system architecture link - Tighten model intro and MoE section heading; drop duplicated latency line - Note the SGLang `random` (ShareGPT) benchmark dataset in Fig. 1 + config - Rephrase DP/EP scaling and EPLB/GLA bullets without notation-heavy dashes Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> * blog: align Ling-2.6 references with house style Use [n] [Title](url) format (matching 2026-05-28-mori), making the descriptive title the link and dropping the em-dash + duplicated raw URL. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> * blog: clarify activation quant, JAX device term, and MoE cost wording - Note that activation quantization is dynamic per-token fp8 with no observed accuracy regression (addresses review question) - Introduce "JAX devices (two per v7x chip)" on first use, shorten later mentions to "devices" - "MoE cost" -> "MoE's operational cost" per review nit Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> * blog: fix author/ack name order and add 0xaskr - Author: Fu Haolin -> Haolin Fu (given-name-first, consistent with the rest) - Acknowledgments: YuHong Guo -> Yuhong Guo; add 0xaskr to SGLang-JAX team Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> * blog: set Ling-2.6 TPU post publish date to June 17, 2026 Rename post file and image directory from 2026-06-11 to 2026-06-17 (URL slug and image paths), and update the frontmatter date accordingly. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> * blog: annotate all-to-all benchmark payload size Add the payload definition (16384 × 8192 × dtype_size) to the Mode column header of the flat-vs-hierarchical all-to-all table. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
1 parent 16b9e9a commit e3af9ed

9 files changed

Lines changed: 498 additions & 0 deletions

blog/2026-06-17-ling-2-6-tpu.md

Lines changed: 498 additions & 0 deletions
Large diffs are not rendered by default.
115 KB
Loading
109 KB
Loading
13.3 KB
Loading
102 KB
Loading
75.8 KB
Loading
98.2 KB
Loading
98.2 KB
Loading
105 KB
Loading

0 commit comments

Comments
 (0)