Super 3.5 vllm model tuning (Qwen3.5-122B-A10B, Inkling-Small) - #2599

Merged

bxyu-nvidia merged 30 commits into

mainfrom

fsiino/super35-vllm-model-tuning

Aug 26, 2026

Commits on Aug 12, 2026

Fix eval container profiling dependency
fsiino-nvidia
committed
693e911
Harden external vllm jobs
fsiino-nvidia
committed
e367f34
Init Qwen config
fsiino-nvidia
committed
7c5ecb9
Fix concurrent Tau2 data setup
fsiino-nvidia
committed
dbd666d
Fail fast during Super 3.5 eval
fsiino-nvidia
committed
900d181
Make Super 3.5 vllm launcher portable
fsiino-nvidia
committed
9df2268

Commits on Aug 13, 2026

Stabilize Qwen3.5 vLLM serving under load
fsiino-nvidia
committed
474b4ef
Add opt-in eval concurrency and safe node exclusions
fsiino-nvidia
committed
c11be37
Enable Qwen3.5 CUDA graphs with BF16 KV cache
fsiino-nvidia
committed
a728408

Commits on Aug 14, 2026

Commits on Aug 17, 2026

Commits on Aug 19, 2026

Init inkling small config
fsiino-nvidia
committed
7b8433f
Handle malformed Tau2 tool calls without aborting evals
fsiino-nvidia
committed
495dc92
Tune Inkling Small prefill and MTP settings
fsiino-nvidia
committed
c1f3def
Merge remote-tracking branch 'origin/main' into fsiino/super35-vllm-model-tuning
fsiino-nvidia
committed
ef1dd91
Fix eval container dependencies
fsiino-nvidia
committed
c685f1f
Remove summarize vllm metrics script
fsiino-nvidia
committed
18b1a74

Commits on Aug 20, 2026

Commits on Aug 21, 2026

lint
fsiino-nvidia
committed
7c0dafe
Lint again
fsiino-nvidia
committed
2ca1aa8

Commits on Aug 26, 2026