3c833b3f7f78807440fc36e5c61c0c780a5efebf2092941adc336a1253968f34  CHECKPOINTS.md
5b34869ed123ed745227f6f4a4e24d76f63199ddbfdd7edebb74d87f12e591a1  README.md
8f9bcbf4cc3b11b273f4207633f37783586cd040c25a7b33192b21d9a4cc7092  VALIDATION.md
467809f9f9a3a35b80f04f8c5850c02ed3af2f8b57c41654f99d7bf7e4eaffd4  e2e/analyze.py
2c30e982cf60bd7d4d8f6bcfe7785bcbeefe1ebaedc6d1f2eed5c7843a2ddc62  e2e/check_budget_cpu.py
d7c9092dc79a08518b16b06ae75147f8bcbb2bdb1deb92ca821e57e6d68d2bba  e2e/cpu_scheduler.py
4c1f48a7b63bf72524e9c7e89131f41765008bfccd4e6db0f4c0f7dc55ca210c  e2e/doctor.py
3e1f09d71055d6861458c5f29284dfdb86feae4100127e2241bfff8d4bfbdb53  e2e/download_model.py
e1a1ff0c917f2ee917dd34b76f8f862baf4dad317baa87bcbfed15b6248778d6  e2e/final_report.py
4140680582e84c10779f071f53f44203189133e4aca9d34add26fbd5b3abd5f9  e2e/logits_check.py
1d0024c7cfd102de1f573b28a1b5fe24717935f2fd51d405454aa73d0ca36ad9  e2e/principles.py
7a462eb0f994c1e07124bceef88d2be4b7e12b0f838b22c6e48e64dfa8dc2535  e2e/quick_walk.py
d77a9a54d03a20f9b3751c15bfd2e80836a034968a0152bcbcab1aa2ad3e884b  e2e/scheduler_practice.py
2a16fba34c0dec516e20a205ac1a9a98349b858cca39bef77cd6f384c31c2d0a  e2e/smoke.py
82f94ee28cde8bef701f950bc7a135b1912f60c7fdf81d6ce00da416cbb8bd8c  e2e/source-manifest.json
b206ed8f5e907e4bdbdd20c6deff2f3069a8c17e82e4768d8f7dede62bb5e34b  e2e/source_walk.py
b77023e505ca03c5dd4948c0cab6afba26c0dffbcbf432b60730c75dbd04b581  env.sh
f16a7abef78f7c6bccc5aec877942bc0ec820267ce1b676e45c00ca53f49ba53  labs/01_attention_numpy.py
6dd96d769f9b830e4f72b28f38efb8a02f481e6889f635d45e44d4046a7e8bc0  labs/02_cached_forward.py
04ca9577fb15fce92358dee8a00c5709f94f8813d5a84d047029bf1aa560d594  labs/03_trace_nanovllm.py
ddd4637defa675c26dfb9e41e1b289ebad1937e069dec4e064f7f4aa2d366320  labs/04_benchmark_nanovllm.py
76e7b19d70924408232476335ab3e8832bb1115952f117528192f71d04467266  notes/attention.md
ac16c7838401042d79412654231d0ba30c8fde6848250c28731b8bfa5c099416  notes/cached-forward.md
66cb17126c669b90576eff2beac6b0434ba50fb6b6866214f676d67a46a3402b  notes/cuda-graph.md
b99082a30a9efda36b9ade8164eeba127d0d3e47c29510b3ea27633cf2170aca  notes/engine-logits.md
857a6602ff031c52da5cbcea886c0cb97a81617a5651c45c80f1aded6fb40da8  notes/execution-change.md
72e20de6d10a9ba70bada57ce69b14ace9d1f0fe1d08b40e1f1faba56698b5c3  notes/generation-timeline.md
9ffa176352b235423753d8171aeed9fd0132336b1743785c246c2b4f4241017f  notes/graph-fallback.md
61608ad5ba47f743abe816f0db314227b1bb72785434710a1f061784f9391e68  notes/kv-cache-math.md
f28d5b54fe68daf62467d38fea7e6fc29fb26774d4e171da2971b0fdd27fdb13  notes/kv-pages.md
3bc200677e46f0ceae2acc836221558d0ecc3a1319dfc4fef3a98a551a1eb036  notes/lifecycle.md
91986495dbd9ee4a5536a8f7e5c34a86c19233efe72f1f9a593805cbb3953cad  notes/memory-budget.md
89d47b4652e31fc2aea784aae04d24383165dfd8984a248c37aef10196b28450  notes/model-execution.md
9edea610d547fdf6a6756d6848c7baf29dd45f1a9de3bc6d3a88b1c9970587e5  notes/parallel-service.md
1aa4a1d0c9ec365e2f6982c44c406d85ee07fcd541c28d3fb481bf2fdcda3005  notes/prefix-cache.md
61c082c204b67fb4316079140640b6287c5c2861e36272fa079fb38fde344add  notes/scheduler.md
11d3dfea9e5e46167d5b65100c304bae6f6744a3ca37266eb305a77d748d14cd  notes/scheduling-change.md
1c04dcb5cbaf6f1ce2d03b4233adcdafe7a452489b0651f9ba04ea70c096aaa2  notes/service-lifecycle.md
9c637d03dfa70037100f78c5ec411945085b2227a59306cc186b4c63f66586a0  notes/service-runtime.md
7c1c34e76af61c04d8f7e80104ca433f37b81330db0d06725b9a1c83fd478dee  notes/summary.md
05d2a51684a7d605ad5c74d00f6d61cc7be983f5dc28c24de287c7035840dfbc  notes/tensor-parallel.md
2e0e4c2588fa36a0395e1a192158f84eec3271d3f790d09eee4e914cffc0196a  notes/tokenizer.md
ef646d0db921a3474a51d7291779583dedde1b3debbcaaffe36b8aa3afb0cab8  patches/01-step-trace.patch
68591a8397e87e9920fdf3f8455dddec725ec0d2e8baea03c0208e24739e7a55  patches/02-graph-eager-fallback.patch
40ed6fde3ea4d7dbbd50f133b865384370545f71b3790965af6fdaa1fd1ab267  patches/03-fair-scheduler.patch
7ceb012b368ea8cbf9b963848f758a8d9238b6499fe1deeccf6ffc804e6f1a7e  patches/04-replica-isolation.patch
4a7e1a0cc16530f70b486d93696f211c18bbb19a11b87b95dfd32249477c7307  pedagogy/00_uncached_walkthrough.py
95de94908e3521e250526ecfa019c5882ccb24d53beaedc8de8dcd507e8d3651  practice/__init__.py
52243942f8bfa1f4d1e460662d11bbe65282f7ac2020dfba038acad0861a11b3  practice/report.py
22934dda35edef0356769cb80052dbcad081379866b40ac4ca5f321b4515df72  practice/run.py
b5c83ff33a3b92ff7630036c1d32e138c928b7c2255d6ee41dfbe7184decc8d0  practice/worker.py
3c2d047d24e943bf0497affa8ecbb3a5e0955666460a11f6f324689b46574cb6  requirements-cpu.txt
1378e250dfc972f3b2c636b5ce5d110bfd91fb1efd3a2c7d2b790cddc4b6c402  requirements-nano.txt
017df3bf8215693528372f317fd67b751125a2c187212d48b0b87f16fd6c8720  validation/attention.json
49b6f936db7fab9212a5daf07f391797b2773f94a0aa0ebd378710d5a3b9cbaf  validation/budget.json
08433fc005a4c549d84739ec107f5e32c18aa98c9cdffa31b57449f7ec048c03  validation/cache.json
c1e4a1b1bd9b47174246dc0a33feb6a3b7a67024e2964159eb1a90364066952e  validation/check_scheduler_baseline.py
ffdb8a82704eae61ee1045b97f1bcab931bff599e4375cbd4f96ef6c0af67848  validation/closure-cpu.json
59bcd43f45bb90ba68f5ab47c932cd1a5840d808a298a3ce35811f1d5d6e80b2  validation/memory.json
991e7fcbeb1e7cd7757f8d52fda1e2d7c3484adcd4fb1376ffe22a7d1ceaa6e6  validation/practice/balanced-cpu.summary.json
512c046515203eb963bcd9020deb7ca9c65549b9ae56fbe01832d6de7b3b5591  validation/practice/client-cancel-cpu.summary.json
f0f57e628fcc2f4e979f6a749091045810be548bac96078d807cd073541f708d  validation/practice/comparison-cpu.json
e50063576f7af18d8b25b1b50454d9c92139df6f91cc5ea7e997f07d1545dfb6  validation/practice/comparison-cpu.md
d53c7a7dfd78e8ea696bd525db1fffd58453e861153a512287760ecee58d03d2  validation/practice/prefill-first-cpu.summary.json
5ad63a39f2fa2a3c1e8566a83805f68e0098a2019b95a3436c8fa08435ef421a  validation/practice/replicas-cpu.summary.json
e0f1c9adcd2d64a5e738d20eb6de5303c5a47a767916b0fcf3715bb5c45805b6  validation/practice/replicas-report-cpu.json
2a90ace4048a807adcbd4e93bf6a3ff0b930362ac236618981d147843c82c3d9  validation/practice/replicas-report-cpu.md
bcc5ae01343902077536430e9268302c6527341f30f9013db0bde1626b57f3e0  validation/practice/round-cpu.summary.json
ba78cc2ce979d818b4b481d24bbf78f398a61abe17007ffbb2c1296539cfa7fd  validation/practice/runtime-cpu.summary.json
e9553d92dec7a934b5a2aace5eb665e0364674617e00603f4718ed523ac0a9f0  validation/preempt.json
c091c8618189cc19ff231a3ad9574c3e90afcfa86bfa1ae9f0df2f765e1bc80a  validation/quick-cpu.json
c1031a1bd4345b2a79ad30c14e32e3b303a422fafa88b3edf223884d2b2fca8f  validation/refs.json
44d5f1380cf3017b5cfc8eea4333d758ddfa802e46246a6db42f1e6b71b00230  validation/replica-isolation-cpu.json
88131bba587a729e7582db7011729d2e3237a16076e0d130dcd40fdec34334f1  validation/scheduler-baseline-validation.json
515ec3daad0cd2f48c957799ef70f341bd45c796d0452231dc3631679b4aa83e  validation/scheduler-patch-validation.json
efcd9c4cb407114e5b9eee880b911e0ab9c49ea0b7f28ac26930bf5b122c51ef  validation/scheduler-practice.json
e2aeddc886175a3ea378e642a16e743a74d77654d07da32336bcb8b877df855d  validation/shapes-tp2.json
9a98a65a29358cb025808d1490f35c9bf1c6926092b56b6e85fa0466d00508ed  validation/shapes.json
68904772a776e664e914acc873d4ae4bc844bda8764f63cf307ede76cbcb2fbe  validation/slots.json
a3c59f8b79c4c3b4e1a677b72d85cb3c33867c873e013ed7cffa06b0df8539de  validation/timeline.json
a8b8c12da33f47db72232cc52b143cdffb1b399c25f8f812e3e3ec87f617e4c1  validation/validation-summary.json
e7f5149f2c91e29bb558987664ebd1f35a5fdfdda45a442d9ae59db8406ab9e7  workloads/mixed.jsonl
8585ce5e462b7fa0e37fda7cbbedc15a2e8e264e31716816a7ab7771853724e9  实验手册.html
