Second version of degub/deterministic configs.#1807

Merged

tianyu-l merged 47 commits intopytorch:mainfrom

githubsgi:deterministic02

Oct 29, 2025

Commits on Oct 29, 2025

Second version of degub/deterinistic configs.
githubsgi
committed
Review relaeted updates.
githubsgi
committed
Review 2 related changes.
githubsgi
committed
[DSV3] Offload dequantization process to DCP QuantizedHFReader (#1804)

wwwjn
authored and
githubsgi
committed
Disable FlexAttention max-autotune when deterministic is used (#1808)

fegin
authored and
githubsgi
committed
Fix num of layers for deepseek-v3 (#1845)

wwwjn
authored and
githubsgi
committed
[VLM] Add token-imbalance loss (#1803)

lkhphuc
authored and
githubsgi
committed
refactor TrainSpec to remove the name field (#1850)

tianyu-l
authored and
githubsgi
committed
Refactor attention and make attention mask an argument to the model (#1776)

fegin
authored and
githubsgi
committed
minor refactor over EP (#1854)

tianyu-l
authored and
githubsgi
committed
Graduate qwen3 from experiment to core (#1860)

wwwjn
authored and
githubsgi
committed
Review related updates.
githubsgi
committed
Rebasing and adding MATH attention kernel.
githubsgi
committed
Indent issue fix.
githubsgi
committed
Removing ipex.
githubsgi
committed
Review updates.
githubsgi
committed
Fixing linter error.
githubsgi
committed
graduate llama4 to core (#1865)

tianyu-l
authored and
githubsgi
committed
consolidate experiments/deepseek_v3 (#1869)

tianyu-l
authored and
githubsgi
committed
add auto_eager_graph_pass (#1813)

ruisizhang123
authored and
githubsgi
committed
Second version of degub/deterinistic configs.
githubsgi
committed
Review relaeted updates.
githubsgi
committed
Review 2 related changes.
githubsgi
committed
[DSV3] Offload dequantization process to DCP QuantizedHFReader (#1804)

wwwjn
authored and
githubsgi
committed
Disable FlexAttention max-autotune when deterministic is used (#1808)

fegin
authored and
githubsgi
committed
[VLM] Add token-imbalance loss (#1803)

lkhphuc
authored and
githubsgi
committed
refactor TrainSpec to remove the name field (#1850)

tianyu-l
authored and
githubsgi
committed
Refactor attention and make attention mask an argument to the model (#1776)

fegin
authored and
githubsgi
committed
add script to train with ft (#1812)

tushar00jain
authored and
githubsgi
committed
Indent issue fix.
githubsgi
committed
Post rebase changes.
githubsgi
committed
Second version of degub/deterinistic configs.
githubsgi
committed
Review relaeted updates.
githubsgi
committed
Review 2 related changes.
githubsgi
committed
add script to train with ft (#1812)

tushar00jain
authored and
githubsgi
committed
minor refactor over EP (#1854)

tianyu-l
authored and
githubsgi
committed
[vlm] Add light-weight CI for experimental models (#1848)

wwwjn
authored and
githubsgi
committed
add owners and CI status for experiments (#1859)

tianyu-l
authored and
githubsgi
committed
move PP API to model agnostic file (#1868)

tianyu-l
authored and
githubsgi
committed
[refactor] graduate custom_config_module and unify args/config naming (#1871)

tianyu-l
authored and
githubsgi
committed
Rebase misses.
githubsgi
committed
Rebase mistakes.
githubsgi
committed
Lint'er error fixes.
githubsgi
committed
Review updates.
githubsgi
committed
Review unpdates.
githubsgi
committed
ufmt fixes.
githubsgi
committed
Update scripts/generate/test_generate.py

githubsgi
and
tianyu-l
authored