Add vLLM v0.18.1 source tree with KV transfer abort fix
third_party/vllm/ now tracked in git for direct patch management.
Based on vLLM v0.18.1 release with one patch applied:
vllm/v1/core/sched/scheduler.py:
Replace fatal assert with graceful skip when KV transfer callback
arrives for an already-aborted request during PD disaggregated serving.
Future vLLM modifications should be made directly in third_party/vllm/
and committed normally. The patches/ directory is kept as documentation
of what changed from upstream.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
BIN
third_party/vllm/docs/assets/features/disagg_encoder/disagg_encoder_flow.png
vendored
Normal file
|
After Width: | Height: | Size: 84 KiB |
BIN
third_party/vllm/docs/assets/features/disagg_prefill/abstraction.jpg
vendored
Normal file
|
After Width: | Height: | Size: 102 KiB |
BIN
third_party/vllm/docs/assets/features/disagg_prefill/high_level_design.png
vendored
Normal file
|
After Width: | Height: | Size: 91 KiB |
BIN
third_party/vllm/docs/assets/features/disagg_prefill/overview.jpg
vendored
Normal file
|
After Width: | Height: | Size: 173 KiB |
BIN
third_party/vllm/docs/assets/features/disagg_prefill/workflow.png
vendored
Normal file
|
After Width: | Height: | Size: 88 KiB |
321
third_party/vllm/docs/assets/features/speculative_decoding/speculators-user-flow-dark.svg
vendored
Normal file
|
After Width: | Height: | Size: 339 KiB |
275
third_party/vllm/docs/assets/features/speculative_decoding/speculators-user-flow-light.svg
vendored
Normal file
|
After Width: | Height: | Size: 374 KiB |