Add H100 MiniMax-M3 NVMe and tiered AgentX sweep - #2775

Closed

cquil11 wants to merge 22 commits into

mainfrom

codex/h100-minimaxm3-nvme-agentx

Commits on Aug 28, 2026

feat(agentx): add H100 NVMe KV offload sweep
cquil11
committed
5d79351
feat(agentx): add tiered DRAM and NVMe sweep
cquil11
committed
44d1af3
test(agentx): cover tiered KV offload matrices
cquil11
committed
923f60d
refactor(agentx): represent tiered offload as tier list
cquil11
committed
3ebc733
perf: schedule H100 MiniMax-M3 offload sweep
cquil11
committed
0e41fd1
fix(agentx): extend H100 allocation time
cquil11
committed
ef3c36e
fix(h100): clean NVMe before releasing allocation
cquil11
committed
366d271
fix(agentx): reject multinode NVMe offload
cquil11
committed
0545bda
fix(h100): exclude node with broken NVLink P2P
cquil11
committed
8de9e47
fix(h100): clean stale vLLM offload memory
cquil11
committed
ae673c5
fix(h100): purge stale NVMe offload directories
cquil11
committed
d83dd9d
fix(h100): bound offload cleanup steps
cquil11
committed
f436e23
simplify H100 offload lifecycle
cquil11
committed
e631356
test updated cache-source image
cquil11
committed
fc9433a
Merge remote-tracking branch 'origin/main' into codex/h100-minimaxm3-nvme-agentx
cquil11
committed
48389ac
bound in-container NVMe cleanup
cquil11
committed
1a0793b
simplify H100 offload launcher
cquil11
committed
3cb3641
test tiering timeout image
cquil11
committed
893cd1f

Commits on Aug 31, 2026

clean stale vLLM offload mmap files
cquil11
committed
8a00c0a
document stale offload cleanup
cquil11
committed
a4c2a6c
Merge remote-tracking branch 'origin/main' into codex/h100-minimaxm3-nvme-agentx
cquil11
committed
b52c639

Commits on Sep 1, 2026

Narrow H100 cache-source validation sweep
cquil11
committed
cb0fec2