Skip to content

[bugfix] ascend qwen3-30b fsdp2 model update bugfix and update yaml #451

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

Merged
PanAndy merged 1 commit into from
May 26, 2026
Merged

[bugfix] ascend qwen3-30b fsdp2 model update bugfix and update yaml #451

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
4 changes: 2 additions & 2 deletions ...ascend_examples/qwen3_30b_rlvl_fsdp2.yaml → ...ascend_examples/qwen3_30b_rlvr_fsdp2.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -58,8 +58,8 @@ add_token_level_kl: false

whiten_advantages: true

pretrain: /home/l00691321/models/Qwen3-30B-A3B
reward_pretrain: /home/l00691321/models/Qwen3-30B-A3B
pretrain: Qwen/Qwen3-30B-A3B
reward_pretrain: Qwen/Qwen3-30B-A3B

actor_train:
model_args:
Expand Down
2 changes: 1 addition & 1 deletion roll/third_party/fsdp2/model_update.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -223,7 +223,7 @@ def _colocated_model_update(self):
infer_parallel_tensors = [None] * infer_parallel_size if co_infer_rank == 0 else None
global_dst_rank = dist.get_global_rank(self._infer_parallel_cpu_group, 0)
dist.gather_object(
serialized_tensors, infer_parallel_tensors, group_dst=global_dst_rank, group=self._infer_parallel_cpu_group
serialized_tensors, infer_parallel_tensors, dst=global_dst_rank, group=self._infer_parallel_cpu_group
)
if refs:
ray.get(refs)
Expand Down
29 changes: 20 additions & 9 deletions roll/third_party/vllm/ray_distributed_executor.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -88,15 +88,26 @@ def _init_workers_ray(self, placement_group: "PlacementGroup", **ray_remote_kwar
runtime_env = RuntimeEnv(env_vars=env_vars)
assert current_platform.ray_device_key == "GPU" or "NPU"
# NV+AMD GPUs, and Intel XPUs
worker = ray.remote(
num_cpus=0,
num_gpus={current_platform.ray_device_key: 0.01},
runtime_env=runtime_env,
scheduling_strategy=PlacementGroupSchedulingStrategy(
placement_group=pg,
),
**ray_remote_kwargs,
)(RayWorkerWrapper).remote(rpc_rank=rank)
if current_platform.ray_device_key == "GPU":
worker = ray.remote(
num_cpus=0,
num_gpus=0.01,
runtime_env=runtime_env,
scheduling_strategy=PlacementGroupSchedulingStrategy(
placement_group=pg,
),
**ray_remote_kwargs,
)(RayWorkerWrapper).remote(rpc_rank=rank)
else:
worker = ray.remote(
num_cpus=0,
resources={current_platform.ray_device_key: 0.01},
runtime_env=runtime_env,
scheduling_strategy=PlacementGroupSchedulingStrategy(
placement_group=pg,
),
**ray_remote_kwargs,
)(RayWorkerWrapper).remote(rpc_rank=rank)
worker_metadata.append(RayWorkerMetaData(worker=worker, created_rank=rank))

worker_ips = ray.get(
Expand Down
Loading