feat: add home and update TODOs

This commit is contained in:
2026-05-07 17:12:25 +08:00
parent a57afa86b4
commit 8036c9016c
21 changed files with 21579 additions and 1275 deletions

View File

@@ -451,7 +451,7 @@ TP1 + PP1 + DP8 + BLKSZ 32 + MAX_BATCH_TOKENS 16384 + MAX_SEQS 32
```
- [x] 更新 vllm to main to fix DP performance problem
- [ ] AITuner 添加一个性能 monitor 的角色,多次 iter 得到相似性能时,该 role 指示 LLM 做更激进的探索
- [x] AITuner 添加一个性能 monitor 的角色,多次 iter 得到相似性能时,该 role 指示 LLM 做更激进的探索
- [x] 支持 early stop当某个 config 性能已经明显爆炸时,提前 quit 测试,避免跑完整的 1h+
- [x] 让 LLM 显式给出每轮 iter 的从数据分析得到的理由以及预期优化目标,并和实际测试结果进行对比
- [x] 比较 10min/60min 效果,是否能缩短时间
@@ -1961,13 +1961,13 @@ on this workload under the given SLO constraints, based on real experimental evi
测试 TODO
- [ ] qwen3-30b-a3b | traceA | 60min
- [ ] qwen3-30b-a3b | thinking | 60min
- [ ] qwen3-30b-a3b | coder | 60min
- [ ] qwen3-235b-a22b | traceA | 60min
- [ ] qwen3-235b-a22b | thinking | 60min
- [ ] qwen3-235b-a22b | coder | 60min
- [x] qwen3-30b-a3b | traceA | 60min
- [x] qwen3-30b-a3b | thinking | 60min
- [x] qwen3-30b-a3b | coder | 60min
- [x] qwen3-235b-a22b | traceA | 60min
- [x] qwen3-235b-a22b | thinking | 60min
- [x] qwen3-235b-a22b | coder | 60min
#### Misc