feat(phase7+8): Production LLaDA2.0 model + vLLM-native torch.compile optimization#38

Open

AlonKellner-RedHat wants to merge 95 commits intomainfrom

feat/phase7-llada2-real-model

Commits on May 5, 2026

Commits on May 6, 2026

feat: add LLaDA2.0 benchmark tests with TPS, TTFT, ITL, E2E metrics
AlonKellner-RedHat
committed
feat: add GuideLLM benchmark test for LLaDA2.0-mini

AlonKellner-RedHat
and
claude
committed
fix: benchmark test trust_remote_code and async-scheduling args

AlonKellner-RedHat
and
claude
committed
fix: correct GuideLLM CLI command structure

AlonKellner-RedHat
and
claude
committed
fix: use correct GuideLLM synthetic data format

AlonKellner-RedHat
and
claude
committed
fix: add processor argument for GuideLLM synthetic data

AlonKellner-RedHat
and
claude
committed
fix(scheduler): skip spec decode metrics to prevent assertion failures

AlonKellner-RedHat
and
claude
committed
feat(phase8): add GPU capability detection and torch.compile optimization

AlonKellner-RedHat
and
claude
committed
feat(phase8): use vLLM-native torch.compile via @support_torch_compile

AlonKellner-RedHat
and
claude
committed
chore: clean up root directory and document Phase 8 benchmarks

AlonKellner-RedHat
and
claude
committed
fix: address code review feedback (P0-P1 issues)

AlonKellner-RedHat
and
claude
committed
feat(phase7): implement virtual batch pattern for block-style attention (WIP)

AlonKellner-RedHat
and
claude
committed

Commits on May 7, 2026

Commits on May 9, 2026