Skip to content

fix: enable gradient checkpointing by default #689

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
garrett4wade merged 1 commit into from
Dec 9, 2025
Merged

fix: enable gradient checkpointing by default #689

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
2 changes: 1 addition & 1 deletion examples/alignment/hhrlhf_rw.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -23,7 +23,7 @@ model:
init_from_scratch: false
is_critic: true
disable_dropout: true
gradient_checkpointing: false
gradient_checkpointing: true
dtype: bfloat16
mb_spec:
max_tokens_per_mb: 4096
Expand Down
2 changes: 1 addition & 1 deletion examples/camel/config.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -41,7 +41,7 @@ actor:
path: Qwen/Qwen2-1.5B-Instruct
init_from_scratch: false
disable_dropout: true
gradient_checkpointing: false
gradient_checkpointing: true
dtype: bfloat16
mb_spec:
max_tokens_per_mb: 16384
Expand Down
2 changes: 1 addition & 1 deletion examples/countdown/train_config.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -38,7 +38,7 @@ actor:
path: Qwen/Qwen2.5-3B-Instruct
init_from_scratch: false
disable_dropout: true
gradient_checkpointing: false
gradient_checkpointing: true
dtype: bfloat16
mb_spec:
max_tokens_per_mb: 8192
Expand Down
2 changes: 1 addition & 1 deletion examples/experimental/prox_approx/gsm8k_grpo_prox_approx.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -37,7 +37,7 @@ actor:
path: Qwen/Qwen2.5-1.5B-Instruct
init_from_scratch: false
disable_dropout: true
gradient_checkpointing: false
gradient_checkpointing: true
dtype: bfloat16
mb_spec:
max_tokens_per_mb: 10240
Expand Down
2 changes: 1 addition & 1 deletion examples/experimental/prox_approx/gsm8k_grpo_prox_approx_eval.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -37,7 +37,7 @@ actor:
path: Qwen/Qwen2.5-1.5B-Instruct
init_from_scratch: false
disable_dropout: true
gradient_checkpointing: false
gradient_checkpointing: true
dtype: bfloat16
mb_spec:
max_tokens_per_mb: 10240
Expand Down
2 changes: 1 addition & 1 deletion examples/lora/gsm8k_grpo_lora.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -39,7 +39,7 @@ actor:
path: Qwen/Qwen2.5-1.5B-Instruct
init_from_scratch: false
disable_dropout: true
gradient_checkpointing: false
gradient_checkpointing: true
dtype: bfloat16
mb_spec:
max_tokens_per_mb: 10240
Expand Down
2 changes: 1 addition & 1 deletion examples/lora/gsm8k_grpo_lora_vllm.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -40,7 +40,7 @@ actor:
path: ./model/Qwen3-0.6B
init_from_scratch: false
disable_dropout: true
gradient_checkpointing: false
gradient_checkpointing: true
dtype: bfloat16
mb_spec:
max_tokens_per_mb: 10240
Expand Down
2 changes: 1 addition & 1 deletion examples/math/gsm8k_dapo_dynamic_bs.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -38,7 +38,7 @@ actor:
path: Qwen/Qwen2.5-1.5B
init_from_scratch: false
disable_dropout: true
gradient_checkpointing: false
gradient_checkpointing: true
dtype: bfloat16
mb_spec:
max_tokens_per_mb: 10240
Expand Down
2 changes: 1 addition & 1 deletion examples/math/gsm8k_drgrpo.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -38,7 +38,7 @@ actor:
path: Qwen/Qwen2.5-1.5B-Instruct
init_from_scratch: false
disable_dropout: true
gradient_checkpointing: false
gradient_checkpointing: true
dtype: bfloat16
mb_spec:
max_tokens_per_mb: 10240
Expand Down
2 changes: 1 addition & 1 deletion examples/math/gsm8k_grpo.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -38,7 +38,7 @@ actor:
path: Qwen/Qwen2.5-1.5B-Instruct
init_from_scratch: false
disable_dropout: true
gradient_checkpointing: false
gradient_checkpointing: true
dtype: bfloat16
mb_spec:
max_tokens_per_mb: 10240
Expand Down
2 changes: 1 addition & 1 deletion examples/math/gsm8k_gspo.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -38,7 +38,7 @@ actor:
path: Qwen/Qwen2.5-1.5B-Instruct
init_from_scratch: false
disable_dropout: true
gradient_checkpointing: false
gradient_checkpointing: true
dtype: bfloat16
mb_spec:
max_tokens_per_mb: 10240
Expand Down
2 changes: 1 addition & 1 deletion examples/math/gsm8k_liteppo.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -38,7 +38,7 @@ actor:
path: Qwen/Qwen2.5-1.5B-Instruct
init_from_scratch: false
disable_dropout: true
gradient_checkpointing: false
gradient_checkpointing: true
dtype: bfloat16
mb_spec:
max_tokens_per_mb: 10240
Expand Down
2 changes: 1 addition & 1 deletion examples/math/gsm8k_m2po.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -38,7 +38,7 @@ actor:
path: Qwen/Qwen2.5-1.5B-Instruct
init_from_scratch: false
disable_dropout: true
gradient_checkpointing: false
gradient_checkpointing: true
dtype: bfloat16
mb_spec:
max_tokens_per_mb: 10240
Expand Down
2 changes: 1 addition & 1 deletion examples/math/gsm8k_ppo.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -38,7 +38,7 @@ actor:
path: Qwen/Qwen2.5-1.5B-Instruct
init_from_scratch: false
disable_dropout: true
gradient_checkpointing: false
gradient_checkpointing: true
dtype: bfloat16
mb_spec:
max_tokens_per_mb: 10240
Expand Down
2 changes: 1 addition & 1 deletion examples/math/gsm8k_reinforce.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -38,7 +38,7 @@ actor:
path: Qwen/Qwen2.5-1.5B-Instruct
init_from_scratch: false
disable_dropout: true
gradient_checkpointing: false
gradient_checkpointing: true
dtype: bfloat16
mb_spec:
max_tokens_per_mb: 10240
Expand Down
2 changes: 1 addition & 1 deletion examples/math/gsm8k_reinforce_baseline.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -38,7 +38,7 @@ actor:
path: Qwen/Qwen2.5-1.5B-Instruct
init_from_scratch: false
disable_dropout: true
gradient_checkpointing: false
gradient_checkpointing: true
dtype: bfloat16
mb_spec:
max_tokens_per_mb: 10240
Expand Down
2 changes: 1 addition & 1 deletion examples/math/gsm8k_rloo.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -38,7 +38,7 @@ actor:
path: Qwen/Qwen2.5-1.5B-Instruct
init_from_scratch: false
disable_dropout: true
gradient_checkpointing: false
gradient_checkpointing: true
dtype: bfloat16
mb_spec:
max_tokens_per_mb: 10240
Expand Down
2 changes: 1 addition & 1 deletion examples/math/gsm8k_sft.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -21,7 +21,7 @@ model:
trial_name: ${trial_name}
path: Qwen/Qwen3-1.7B
init_from_scratch: false
gradient_checkpointing: false
gradient_checkpointing: true
dtype: bfloat16
mb_spec:
max_tokens_per_mb: 4096
Expand Down
2 changes: 1 addition & 1 deletion examples/multi-turn-math/config.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -38,7 +38,7 @@ actor:
path: Qwen/Qwen2-1.5B-Instruct
init_from_scratch: false
disable_dropout: true
gradient_checkpointing: false
gradient_checkpointing: true
dtype: bfloat16
mb_spec:
max_tokens_per_mb: 16384
Expand Down
2 changes: 1 addition & 1 deletion examples/openai-agents/config.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -41,7 +41,7 @@ actor:
path: Qwen/Qwen2-1.5B-Instruct
init_from_scratch: false
disable_dropout: true
gradient_checkpointing: false
gradient_checkpointing: true
dtype: bfloat16
mb_spec:
max_tokens_per_mb: 16384
Expand Down
2 changes: 1 addition & 1 deletion examples/single-controller/gsm8k_grpo.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -44,7 +44,7 @@ actor:
path: Qwen/Qwen2.5-1.5B-Instruct
init_from_scratch: false
disable_dropout: true
gradient_checkpointing: false
gradient_checkpointing: true
dtype: bfloat16
mb_spec:
max_tokens_per_mb: 10240
Expand Down
2 changes: 1 addition & 1 deletion examples/single-controller/gsm8k_sft.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -20,7 +20,7 @@ model:
trial_name: ${trial_name}
path: Qwen/Qwen3-1.7B
init_from_scratch: false
gradient_checkpointing: false
gradient_checkpointing: true
dtype: bfloat16
mb_spec:
max_tokens_per_mb: 4096
Expand Down
2 changes: 1 addition & 1 deletion examples/skypilot/gsm8k_grpo_ray.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -38,7 +38,7 @@ actor:
path: Qwen/Qwen2.5-1.5B-Instruct
init_from_scratch: false
disable_dropout: true
gradient_checkpointing: false
gradient_checkpointing: true
dtype: bfloat16
mb_spec:
max_tokens_per_mb: 4096
Expand Down
2 changes: 1 addition & 1 deletion examples/tir/tir_math_config.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -35,7 +35,7 @@ actor:
path: Qwen/Qwen2.5-Math-1.5B
init_from_scratch: false
disable_dropout: true
gradient_checkpointing: false
gradient_checkpointing: true
dtype: bfloat16
mb_spec:
max_tokens_per_mb: 10240
Expand Down
2 changes: 1 addition & 1 deletion examples/vlm/clevr_count_70k_grpo.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -38,7 +38,7 @@ actor:
path: Qwen/Qwen2.5-VL-3B-Instruct
init_from_scratch: false
disable_dropout: true
gradient_checkpointing: false
gradient_checkpointing: true
dtype: bfloat16
mb_spec:
max_tokens_per_mb: 4096
Expand Down
2 changes: 1 addition & 1 deletion examples/vlm/clevr_count_70k_sft.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -21,7 +21,7 @@ model:
trial_name: ${trial_name}
path: Qwen/Qwen2-VL-7B
init_from_scratch: false
gradient_checkpointing: false
gradient_checkpointing: true
dtype: bfloat16
mb_spec:
max_tokens_per_mb: 4096
Expand Down
2 changes: 1 addition & 1 deletion recipe/AEnt/configs/gsm8k_aent_grpo.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -38,7 +38,7 @@ actor:
path: Qwen/Qwen2.5-1.5B-Instruct
init_from_scratch: false
disable_dropout: true
gradient_checkpointing: false
gradient_checkpointing: true
dtype: bfloat16
mb_spec:
max_tokens_per_mb: 10240
Expand Down