feat: i18n CN/EN, ops kernel labels, UI polish, compare improvements - #60
Merged
Conversation
Embed distribution chart on index.html (keep distribution.js data), add Community pages, submit wizard, share/cite modal, and recipe-first home layout. Remove standalone distribution.html and unused distribution view. Co-authored-by: Cursor <cursoragent@cursor.com>
- Add full Chinese/English i18n across all pages (60+ keys) - Add implOpsLabel() showing backend/kernel info in suite cards & rankings - Fix compare page: suite title/tagline i18n, vendor color tiles, scroll position - Fix citation page: hardcoded text → i18n keys - Fix reproduce page: countLine → i18n, community-nav entries → i18n - Hero chart: A100 vs 5090 ranking reversal visualization - Compare basket: framework selector dropdown for fair comparisons
✅ AccelMark Validation: All submissions validSee the workflow run for details. |
Collaborator
Author
|
你好,你的邮件我已经收到,辛苦了!这是来自QQ邮箱的假期自动回复邮件。
|
added 6 commits
August 3, 2026 18:07
…Error defense, Ascend config example - Remove license classifier from pyproject.toml (PEP 639 conflict with SPDX expression) - Replace pip install -e . with direct dependency install in README/CONTRIBUTING/submit.js - Add --privileged to docker run example in Ascend requirements.txt (npu-smi requires it) - Add defensive check in benchmark_runner._compute_run_id() for empty accelerators list - Add missing runner_ascend_vllm_ascend_d4aa9fda.yaml.example - 910C variant identification via Subsystem Device ID and board info (ascend.py)
…elism - New runner: ascend_vllm_ascend_bad8ded1 with dual-die auto-detection - Auto-detect dies_per_card via npu-smi board info Chip Count field - When dual-die detected and model fits per die: dp_size=2 by default - DP orchestration via subprocesses, each bound to one die via ASCEND_RT_VISIBLE_DEVICES - Merged result includes card_count, dies_per_card, per_card throughput - benchmark_runner.py: chip_info extended with card_count/dies_per_card - _compute_run_id() hash key now includes dies_per_card - --data-parallel-size CLI flag to override auto-detection
fix: CANN toolkit version detection + DP runner refinement (7abb8282)
1. requirements: docker run 加 --privileged + Ascend 设备挂载说明,
修复容器内 npu-smi 返回 -9005 的问题
2. pyproject.toml: PEP 639 — 删除与 SPDX license = "Apache-2.0"
冲突的 license classifier,消除新版 setuptools 构建报错
3. configs: 新增 runner_ascend_vllm_ascend_7abb8282.yaml.example,
全局默认 + suite 级覆盖(B/D/E/F/G 设 data_parallel_size: 1),
替代旧的 bad8ded1 config example
4. ascend.py platform plugin:
- 910C 变种识别:_910C_VARIANT_MAP,解析 Subsystem Device ID、
chip_count、board_id,torch_npu 后填 cube/vector core 等属性
- CANN 版本修复:detect_runtime_version() 从读 npu-smi driver
版本(25.2.3)改为读 version.cfg 的 CANN toolkit 版本(8.5.1),
与 NVIDIA CUDA toolkit version 语义一致
5. DP runner (ascend_vllm_ascend_7abb8282):
- 910C 双 die 自动检测:npu-smi info -t board Chip Count >= 2
- _run_with_dp(): fork dp_size 子进程,ASCEND_RT_VISIBLE_DEVICES
绑定 die,合并吞吐 per_chip / per_card
- max_num_seqs 自动除 dp_size,避免两引擎共享内存广播块争抢
导致的 shm_broadcast 超时
- --data-parallel-size CLI 参数,优先级 CLI > yaml > auto-detect
6. benchmark_runner.py: chip_info 增加 card_count / dies_per_card;
空 accelerators 防御 IndexError → sys.exit(1) 带提示
@
JuhaoLiang1997
approved these changes
Aug 12, 2026
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
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.
Summary
implOpsLabel()extracts backend + kernel features from impl data, shown in suite cards and rankings table