yangzhuxinyzxandCodex b504a224bc [Bugfix] Restore Triton state on the first vLLM AOT reload ...

Serialize referenced Triton side-table entries with fixed launch settings and remap them without overwriting other graphs. Ignore frozen pseudo-files in backend hashes so the first reload reuses its original artifacts. Record first-reload timings and the still-open output-parity gate.

Co-authored-by: Codex <noreply@openai.com>
Signed-off-by: yangzhuxinyzx <153831768+yangzhuxinyzx@users.noreply.github.com>

2026-09-22 13:34:28 +08:00

..

check_boolean_context_manager.py

…

check_forbidden_imports.py

[Bugfix] Restore Triton state on the first vLLM AOT reload

2026-09-22 13:34:28 +08:00

check_init_lazy_imports.py

…

check_spdx_header.py

…

check_torch_cuda.py

Revert "[Release] Restore public source snapshot to v1.3.0 (#329)"

2026-08-27 22:24:17 +08:00

generate_attention_backend_docs.py

[Docs] Fix MLA prefill backend default docs (#43697)

2026-05-27 10:13:22 +00:00

generate_nightly_torch_test.py

…

mypy.py

…

png-lint.sh

…

rust-check.sh

…

shellcheck.sh

…

update-dockerfile-graph.sh

…

validate_config.py

…