Skip to content

fix: make mcore lr scheduler configuration consistent with dtensor #681

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
terrykong merged 5 commits into from
Jul 17, 2025
Merged

fix: make mcore lr scheduler configuration consistent with dtensor #681

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
0fbc0b2
take one scheduler step per rollout batch
ashors1 Jul 16, 2025
c0c7590
update configs
ashors1 Jul 16, 2025
cae0933
remove unused variable
ashors1 Jul 16, 2025
8310511
Remove old comment
ashors1 Jul 16, 2025
fd69656
increase learning rate for test
ashors1 Jul 17, 2025
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
2 changes: 1 addition & 1 deletion examples/configs/grpo_math_1B_megatron.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -116,7 +116,7 @@ policy:
weight_decay_incr_style: "constant"
lr_decay_style: "constant"
lr_decay_iters: null
lr_warmup_iters: 50
lr_warmup_iters: 13
lr_warmup_init: 5.0e-7

distributed_data_parallel_config:
Expand Down
2 changes: 1 addition & 1 deletion examples/configs/grpo_math_70B_megatron.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -49,7 +49,7 @@ policy:
weight_decay_incr_style: "constant"
lr_decay_style: "constant"
lr_decay_iters: null
lr_warmup_iters: 50
lr_warmup_iters: 13
lr_warmup_init: 3.0e-8

generation:
Expand Down
2 changes: 1 addition & 1 deletion examples/configs/grpo_math_qwen30ba3b_megatron.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -52,7 +52,7 @@ policy:
weight_decay_incr_style: "constant"
lr_decay_style: "constant"
lr_decay_iters: null
lr_warmup_iters: 50
lr_warmup_iters: 13
lr_warmup_init: 3.0e-8

env_vars:
Expand Down
11 changes: 6 additions & 5 deletions nemo_rl/models/policy/megatron_policy_worker.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -857,13 +857,8 @@ def train(
num_zeros_in_grad
)

# Update learning rate.
if update_successful:
increment = total_dataset_size.item()
self.scheduler.step(increment=increment)
skipped_iter = 0
curr_lr = self.scheduler.get_lr(self.optimizer.param_groups[0])
curr_wd = self.scheduler.get_wd()
else:
skipped_iter = 1

Expand All @@ -880,6 +875,8 @@ def train(
for k in x.keys():
loss_metrics[k] = x[k] / num_global_batches
gb_loss_metrics.append(loss_metrics)
curr_lr = self.scheduler.get_lr(self.optimizer.param_groups[0])
curr_wd = self.scheduler.get_wd()
loss_metrics["lr"] = curr_lr
loss_metrics["wd"] = curr_wd
loss_metrics["grad_norm"] = grad_norm
Expand All @@ -905,6 +902,10 @@ def train(
all_mb_metrics.extend(gb_loss_metrics)
losses.append(torch.tensor(mb_losses).sum().item())

if not eval_mode:
# take one LR step every rollout batch
self.scheduler.step(increment=1)

# Aggregate metrics across all microbatches
mb_metrics = defaultdict(list)
for m in all_mb_metrics:
Expand Down
4 changes: 2 additions & 2 deletions tests/unit/models/policy/test_megatron_worker.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -777,8 +777,8 @@ def test_megatron_reference_policy_functionality():
)

config = create_megatron_test_config()
config["megatron_cfg"]["optimizer"]["lr"] = 1e-3 # Increase from 5e-6 to 1e-3
config["megatron_cfg"]["optimizer"]["min_lr"] = 1e-4 # Increase min_lr as well
config["megatron_cfg"]["optimizer"]["lr"] = 1e-2 # Increase from 5e-6 to 1e-2
config["megatron_cfg"]["optimizer"]["min_lr"] = 1e-3 # Increase min_lr as well

tokenizer = get_tokenizer(config["tokenizer"])
config["generation"] = configure_generation_config(config["generation"], tokenizer)
Expand Down
Loading