Skip to content

Add nvidia/OpenMathReasoning dataset (cot, tir, genselect splits) #2316

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
dlwh merged 1 commit into from
Jan 12, 2026
Merged

Add nvidia/OpenMathReasoning dataset (cot, tir, genselect splits) #2316

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
60 changes: 60 additions & 0 deletions experiments/posttrain/instruction_datasets.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -453,6 +453,66 @@ def __call__(self, row: dict[str, Any]) -> dict[str, Any]:
name="GeneralReasoning/GeneralThought-195K-modelreasoning",
splits=["train"],
),
# nvidia/OpenMathReasoning - CoT split (Chain of Thought reasoning)
"nvidia/OpenMathReasoning/cot": InstructionDatasetConfig(
hf_dataset_id="nvidia/OpenMathReasoning",
revision="d3d0866",
adapter=instruction_response_adapter(
instruction_column="problem",
response_column="generated_solution",
),
metadata_columns=[
"expected_answer",
"problem_source",
"problem_type",
"generation_model",
"inference_mode",
"pass_rate_72b_tir",
"used_in_kaggle",
],
name="nvidia/OpenMathReasoning/cot",
splits=["cot"],
),
# nvidia/OpenMathReasoning - TIR split (Tool-Integrated Reasoning)
"nvidia/OpenMathReasoning/tir": InstructionDatasetConfig(
hf_dataset_id="nvidia/OpenMathReasoning",
revision="d3d0866",
adapter=instruction_response_adapter(
instruction_column="problem",
response_column="generated_solution",
),
metadata_columns=[
"expected_answer",
"problem_source",
"problem_type",
"generation_model",
"inference_mode",
"pass_rate_72b_tir",
"used_in_kaggle",
],
name="nvidia/OpenMathReasoning/tir",
splits=["tir"],
),
# nvidia/OpenMathReasoning - genselect split (curated subset)
"nvidia/OpenMathReasoning/genselect": InstructionDatasetConfig(
hf_dataset_id="nvidia/OpenMathReasoning",
revision="d3d0866",
adapter=instruction_response_adapter(
instruction_column="problem",
response_column="generated_solution",
),
metadata_columns=[
"expected_answer",
"problem_source",
"problem_type",
"generation_model",
"inference_mode",
"pass_rate_72b_tir",
"used_in_kaggle",
],
name="nvidia/OpenMathReasoning/genselect",
splits=["genselect"],
),
}

for split_name in SMOLTALK2_SPLITS:
Expand Down