[Qwen3.5 B300] preserve power workload on native SRT / 保留原生 SRT 功耗工作负载 - #3379
Draft
edwingao28 wants to merge 8 commits into
Draft
edwingao28 wants to merge 8 commits into
edwingao28 wants to merge 8 commits into
Conversation
补齐 Qwen3.5 B300 的 c1/c2 并要求有效功耗，保留既有覆盖。
Contributor
|
Thanks for the contribution!
中文感谢你的贡献！
|
记录 Qwen3.5 B300 的真实 PR 链接，完成交付溯源。
Contributor
|
View unofficial run (performance): https://inferencex.semianalysis.com/inference?unofficialRun=36058352540 View unofficial run (accuracy): https://inferencex.semianalysis.com/evaluation?unofficialRun=36058352540 |
将 Qwen3.5 FP8 STP 接入现有 B300 batch 分配路径，显式校验路由输入并添加行为回归测试；保留服务参数、测量和功耗验证。
edwingao28
changed the title
同步已合入的 Python 3.12 结果处理前置修复；保留 Qwen3.5 B300 已批准的 batch 路由、测量设置和完整验证矩阵。仅合并追加的性能变更记录。
仅为 Qwen B300 c256 eval 启用完整请求日志，记录结束原因和 token 数；保持模型、生成设置和准确率阈值不变。
将已验证的 B300 DeepGEMM-off 工作负载迁移到原生 SRT，保留聊天模板、功耗验收和完整并发覆盖。
edwingao28
changed the title
…eep] 合并 main 并把本 PR 的 changelog 条目重新追加到末尾，历史条目保持不变。 Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01EDrdQJ2wJHwcswqaU8ii19
Collaborator
Author
|
/use 36058352540 Generated by Claude Code |
This branch has not been deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Description
Preserve B300 TP4 power coverage on native SRT, retain chat formatting, and disable JIT DeepGEMM. Request logging stays limited to c256 evaluation.
Testing: 75 local tests pass. The legacy diagnostic passed strict 96.36% and flexible 95.75% against unchanged 94% gates.
Delivery: Native SRT qualification requires nine throughput points and two full evaluations.
中文
让 B300 TP4 功耗覆盖适配原生 SRT，保留 chat 格式并关闭 JIT DeepGEMM。请求日志仍仅用于 c256 评估。
测试： 75 项本地测试通过。旧路径诊断 strict 为96.36%、flexible 为95.75%，均通过原有94%门槛。
交付： 原生 SRT 仍需完成九个吞吐点和两次完整评估。
GPT-6 负责实现和只读委派核验，精确变体不可用。无关联 issue。
AI model disclosure
Related Issue
None.
Type of Change
Checklist
perf-changelog.yamland have not edited historical entriesOWNER/MEMBER/COLLABORATOR) has commented/use <run_id>(or the legacy/reuse-sweep-run) on this PR. Do this only once there is a final full sweep that is all green with evals passing, since after this comment the sweep label will no longer automatically kick off new sweeps. Remove and re-add the label to force one.