Skip to content

fix(ci): synchronous data loading for hanging t5 functional tests #5241

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
Connor-XY wants to merge 1 commit into from
Closed

fix(ci): synchronous data loading for hanging t5 functional tests #5241

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 6 additions & 0 deletions tests/functional_tests/test_cases/t5/t5_mcore_te_tp1_pp1_vp1_resume_torch/model_config.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -53,4 +53,10 @@ MODEL_ARGS:
--ckpt-format: torch
--attention-backend: unfused
--log-memory-to-tensorboard: true
# Synchronous data loading: the default 2 persistent dataloader workers
# (pin_memory, no DataLoader timeout) can stall on the CI data mount, leaving
# the rank blocked in get_batch -> queue.get forever -> the other ranks time
# out in the grad all-reduce (600s NCCL watchdog). In-process loading removes
# the worker-queue hang surface; test data is tiny so perf impact is nil.
--num-workers: 0
TEST_TYPE: ckpt-resume
6 changes: 6 additions & 0 deletions tests/functional_tests/test_cases/t5/t5_mcore_tp1_pp1_vp1/model_config.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -51,4 +51,10 @@ MODEL_ARGS:
--deterministic-mode: true
--ckpt-format: torch
--log-memory-to-tensorboard: true
# Synchronous data loading: the default 2 persistent dataloader workers
# (pin_memory, no DataLoader timeout) can stall on the CI data mount, leaving
# the rank blocked in get_batch -> queue.get forever -> the other ranks time
# out in the grad all-reduce (600s NCCL watchdog). In-process loading removes
# the worker-queue hang surface; test data is tiny so perf impact is nil.
--num-workers: 0
TEST_TYPE: regular
6 changes: 6 additions & 0 deletions tests/functional_tests/test_cases/t5/t5_mcore_tp1_pp1_vp1_resume_torch/model_config.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -51,4 +51,10 @@ MODEL_ARGS:
--deterministic-mode: true
--ckpt-format: torch
--log-memory-to-tensorboard: true
# Synchronous data loading: the default 2 persistent dataloader workers
# (pin_memory, no DataLoader timeout) can stall on the CI data mount, leaving
# the rank blocked in get_batch -> queue.get forever -> the other ranks time
# out in the grad all-reduce (600s NCCL watchdog). In-process loading removes
# the worker-queue hang surface; test data is tiny so perf impact is nil.
--num-workers: 0
TEST_TYPE: ckpt-resume
Loading