shishi ad4fd2b878 fix: INVALID_WORKLOAD 只跳过当前shape而不中止所有配置 ...

原来 ADAPTIVE_FATAL_WORKLOAD=1 会设 abort_all=1,导致所有剩余TP/DP配置
全部被跳过。改为 continue 只跳过当前 shape,让后续 shape 和配置正常
继续执行。

这修复了 TP=8/DP=2 在 isl=4096 osl=128 因 KV cache 不足导致 workload
验证失败后,TP=16/DP=1 被跳过的 bug。

2026-07-30 10:05:28 +08:00

..

adaptive_bench_lib.sh

fix: INVALID_WORKLOAD 只跳过当前shape而不中止所有配置

2026-07-30 10:05:28 +08:00

adaptive_concurrency.py

fix(910c/glm52): 修复sglang0.5.2解析兼容性+TP=16设备挂载+health超时

2026-07-28 16:43:05 +08:00

adaptive_heartbeat.sh

bench: add adaptive run heartbeat monitor

2026-07-11 13:04:35 +00:00

bench_client_docker.sh

Add P800 long-context matrix and tune benchmark/health settings

2026-07-08 15:02:22 +00:00

compare.py

refactor(common): centralize parse/compare/warmup; add TEMPLATE and EXPERIMENT_GUIDE

2026-07-08 10:48:43 +00:00

lib.sh

fix(lib.sh): avoid cd side-effect in git_commit/git_dirty and safe JSON in write_metadata_json

2026-07-22 03:44:18 +00:00

npu_smi_sampler.py

feat(platform): add Ascend 910C NPU platform support

2026-07-27 22:00:05 +08:00

parse_backend.py

fix(910c/glm52): 修复sglang0.5.2解析兼容性+TP=16设备挂载+health超时

2026-07-28 16:43:05 +08:00

platform.sh

feat(platform): add Ascend 910C NPU platform support

2026-07-27 22:00:05 +08:00

server_docker.sh

Fix P800 baseline server startup issues

2026-07-08 15:27:31 +00:00

warmup.py

refactor(common): centralize parse/compare/warmup; add TEMPLATE and EXPERIMENT_GUIDE

2026-07-08 10:48:43 +00:00