Skip to content

fix(rollout): separate transfer and policy versions #401

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

Open
quinnarnold wants to merge 1 commit into
base: main
Choose a base branch
from
Open

fix(rollout): separate transfer and policy versions #401

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
3 changes: 2 additions & 1 deletion unirl/rollout/engine/fastvideo/engine.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -181,6 +181,7 @@ def __init__(
)

self._version = 0
self._weight_update_calls = 0
self._generate_lock = threading.Lock()
self._shutdown_lock = threading.Lock()
self._shutdown_requested = False
Expand Down Expand Up @@ -620,7 +621,7 @@ def update_weights_from_path(self, checkpoint_path: str, *, track_prefix: str =
require(self._generator is not None, "fastvideo engine is offloaded/not initialized")
self._generator.update_transformer_weights_from_path(checkpoint_path)
self._last_weights_path = checkpoint_path
self._version += 1
self._weight_update_calls += 1
logger.info("fastvideo transformer weights updated from %s", checkpoint_path)


Expand Down
5 changes: 3 additions & 2 deletions unirl/rollout/engine/sglang/engine.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -152,6 +152,7 @@ def __init__(
)

self._version = 0
self._weight_update_calls = 0

def _prepare_generation(self, sample: Sample) -> Any:
require(
Expand Down Expand Up @@ -282,7 +283,7 @@ def update_weights_from_tensor(
load_format=load_format,
flush_cache=flush_cache,
)
self._version += 1
self._weight_update_calls += 1

def init_weights_update_group(
self,
Expand Down Expand Up @@ -329,7 +330,7 @@ def update_weights_from_distributed(
group_name=group_name,
flush_cache=flush_cache,
)
self._version += 1
self._weight_update_calls += 1

def destroy_weights_update_group(
self,
Expand Down
5 changes: 3 additions & 2 deletions unirl/rollout/engine/sglang_diffusion/engine.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -98,6 +98,7 @@ def __init__(
self.schedule_policy = self.adapter.schedule_policy()

self._version = 0
self._weight_update_calls = 0
self._generate_lock = threading.Lock()
self._shutdown_lock = threading.Lock()
self._shutdown_requested = False
Expand Down Expand Up @@ -236,7 +237,7 @@ def update_weights_from_tensor(
load_format=load_format,
flush_cache=flush_cache,
)
self._version += 1
self._weight_update_calls += 1

def init_weights_update_group(
self,
Expand Down Expand Up @@ -279,7 +280,7 @@ def update_weights_from_distributed(
target_modules=target_modules,
flush_cache=flush_cache,
)
self._version += 1
self._weight_update_calls += 1

def destroy_weights_update_group(
self,
Expand Down
7 changes: 4 additions & 3 deletions unirl/rollout/engine/vllm_omni/engine.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -38,6 +38,7 @@ def __init__(
) -> None:
self.cfg = config
self._version = 0
self._weight_update_calls = 0
self._generate_lock = threading.Lock()
self._shutdown_lock = threading.Lock()
self._shutdown_requested = False
Expand Down Expand Up @@ -256,7 +257,7 @@ def update_weights_from_ipc(
use_shm=use_shm,
replica_rank=replica_rank,
)
self._version += 1
self._weight_update_calls += 1

def init_weights_update_group(
self,
Expand Down Expand Up @@ -299,7 +300,7 @@ def update_weights_from_distributed(
target_modules=target_modules,
flush_cache=flush_cache,
)
self._version += 1
self._weight_update_calls += 1

def destroy_weights_update_group(
self,
Expand All @@ -326,7 +327,7 @@ def update_weights_from_tensor(
load_format=load_format,
flush_cache=flush_cache,
)
self._version += 1
self._weight_update_calls += 1

def set_lora_from_tensors(
self,
Expand Down
Loading