Skip to content

fix: change default batch sizes for dp workloads #528

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
binaryaaron merged 4 commits into from
May 28, 2026
Merged

fix: change default batch sizes for dp workloads #528

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
54b6543
chore: change default batch sizes to 4 + 16 grad accumulation
binaryaaron May 27, 2026
f113815
fix: align remaining DP Slurm configs
binaryaaron May 27, 2026
f535cb6
chore: set mistral dp to size 2
mckornfield May 28, 2026
972d223
fix: udpate mistral-dp e2e
binaryaaron May 28, 2026
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 script/slurm/configs/mistral-dp.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2,7 +2,8 @@
# SPDX-License-Identifier: Apache-2.0

training:
batch_size: 8 # DP is not constrained to batch size of 1
batch_size: 4
gradient_accumulation_steps: 16
Comment thread
binaryaaron marked this conversation as resolved.
Resolved
Comment thread
binaryaaron marked this conversation as resolved.
Resolved
pretrained_model: mistralai/Mistral-7B-Instruct-v0.3
privacy:
dp_enabled: true
Expand Down
3 changes: 2 additions & 1 deletion script/slurm/configs/tinyllama-dp.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2,7 +2,8 @@
# SPDX-License-Identifier: Apache-2.0

training:
batch_size: 8 # DP is not constrained to batch size of 1
batch_size: 4 # DP is not constrained to batch size of 1
gradient_accumulation_steps: 16
pretrained_model: TinyLlama/TinyLlama-1.1B-Chat-v1.0
privacy:
dp_enabled: true
Expand Down
3 changes: 2 additions & 1 deletion tests/e2e/required_configs/mistral-dp.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -10,6 +10,7 @@ generation:
privacy:
dp_enabled: true
training:
batch_size: 8
batch_size: 4
gradient_accumulation_steps: 16
Comment thread
binaryaaron marked this conversation as resolved.
Resolved
pretrained_model: mistralai/Mistral-7B-Instruct-v0.3
learning_rate: 0.0001
3 changes: 2 additions & 1 deletion tests/e2e/required_configs/smollm3-dp.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -9,5 +9,6 @@ generation:
privacy:
dp_enabled: true
training:
batch_size: 8
batch_size: 4
gradient_accumulation_steps: 16
pretrained_model: HuggingFaceTB/SmolLM3-3B
3 changes: 2 additions & 1 deletion tests/e2e/required_configs/tinyllama-dp.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -10,5 +10,6 @@ generation:
privacy:
dp_enabled: true
training:
batch_size: 8
batch_size: 4
gradient_accumulation_steps: 16
pretrained_model: TinyLlama/TinyLlama-1.1B-Chat-v1.0
Loading