mirror of https://github.com/tinygrad/tinygrad.git synced 2026-02-07 21:26:21 -05:00

Files

qazal d6f4a61c42 graph LBScheduleItem [run_process_replay] (#5960 )

* add toposort key to LBScheduleItem

* use dedup

* graph LBScheduleItem

* make that comment beautiful again

* diff_schedule utils

* update fuzz_schedule

2024-08-07 19:59:11 +03:00

diff_schedule.py

graph LBScheduleItem [run_process_replay] (#5960 )

2024-08-07 19:59:11 +03:00

process_replay.py

apply opts.extra_matcher in process replay [run_process_replay] (#5877 )

2024-08-02 18:07:58 +03:00

README.md

process replay docs improvements (#5481 )

2024-07-15 00:09:28 +03:00

reset.py

make process replay unique in retries [run_process_replay] (#5773 )

2024-07-28 20:44:15 +03:00

restore.py

add llama3 BEAM=2 failures to test_linearizer_failures (#5553 )

2024-07-30 00:37:32 +03:00

test_diff_schedule.py

diff_schedule tests [run_process_replay] (#5958 )

2024-08-07 13:50:27 +03:00

test_process_replay.sh

assert process replay asserts (#5737 )

2024-07-27 12:07:50 +03:00

README.md

Process replay tests

Process replay is a tool for creating a diff of generated kernels between two commits. By default, process replay doesn't assert kernel diffs.

Refactor and speedup prs must enable the assert by including [run_process_replay] in the pr title.

Note that process replay early stops when over 20% of kernels change, for speed..

Running locally

To run process replay locally:

(optional: clear previous process replay runs with test/external/process_replay/reset.py)

Run tests with RUN_PROCESS_REPLAY=1 in your branch. This will capture the kernels.
Checkout master
Run test/external/process_replay/process_replay.py