Skip to content

Pull requests: allenai/open-instruct

New pull request New
77 Open 1,495 Closed
77 Open 1,495 Closed
Author
Filter by author
Loading
Label
Filter by label
Loading
Use alt + click/return to exclude labels
or ⇧ + click/return for logical OR
Projects
Filter by project
Loading
Milestones
Filter by milestone
Loading
Reviews
Filter by reviews
No reviews Review required Approved review Changes requested
Assignee
Filter by who’s assigned
Assigned to nobody Loading
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Pull requests list

Fix argparse crash: restore deprecated --gradient_checkpointing for DPO
#1870 opened Sep 1, 2026 by divo12 Loading…
Honor --max_train_samples in dpo.py
#1868 opened Aug 31, 2026 by abhishekraok Collaborator Loading…
4
Bump datasets to >=4.8.4,<5
#1867 opened Aug 31, 2026 by abhishekraok Collaborator Loading…
4
Stop data_loader from importing vllm at runtime
#1866 opened Aug 31, 2026 by abhishekraok Collaborator Loading…
4
Install google-cloud-cli, not the dropped google-cloud-sdk package
#1865 opened Aug 31, 2026 by abhishekraok Collaborator Loading…
1
Increase Beaker CLI download timeout
#1864 opened Aug 31, 2026 by JimmyWang0417 Loading…
1
2
Fix OLMo-core SFT vocabulary bounds
#1862 opened Aug 30, 2026 by ReinforcedKnowledge Contributor Loading…
Add never_give_up (NGU) GRPO retry mechanism
#1861 opened Aug 29, 2026 by mnoukhov Contributor Loading…
Support DPO on olmo-core MoE checkpoints (OLMoDDPModel)
#1860 opened Aug 29, 2026 by abhishekraok Collaborator • Draft
2
Checking MOPD's diff with OPD
#1848 opened Aug 25, 2026 by shatu • Draft
On-policy distillation (OPD) support for grpo_fast
#1847 opened Aug 25, 2026 by shatu • Draft
Drop OLMoE-1B-7B from OLMo-core SFT support
#1846 opened Aug 24, 2026 by ajing Loading…
Fix Olmo Hybrid weight loading in GRPO
#1844 opened Aug 21, 2026 by primorLee Loading…
1
Publish wandb metrics from a background thread so a wedged SDK cannot stall training
#1841 opened Aug 20, 2026 by pdasigi Member Loading…
3
Add a script that checks an olmo-core config reproduces its HF logits
#1839 opened Aug 20, 2026 by abhishekraok Collaborator Loading…
5
Detect olmo-core checkpoints by config contents; let mason pass a GCS credential
#1838 opened Aug 19, 2026 by abhishekraok Collaborator Loading…
3
Fix factuality and refusal judge verdicts collapsing to 0.0
#1833 opened Aug 19, 2026 by arthi-arumugam-git Loading…
1
Fix four broken references in the docs
#1832 opened Aug 18, 2026 by VishnuR23 Loading…
Stop charging worker-process startup against max_execution_time
#1831 opened Aug 18, 2026 by VishnuR23 Loading…
Add final HF export to olmo_core_finetune.py; share export_to_hf via olmo_core_utils
#1830 opened Aug 17, 2026 by farhatkevin Collaborator Loading…
8
Record the Beaker experiment in wandb for olmo-core runs
#1829 opened Aug 17, 2026 by VishnuR23 Loading…
1
Update the installation docs to match the current toolchain
#1828 opened Aug 17, 2026 by VishnuR23 Loading…
Backend A/B benchmarks: DeepSpeed vs OLMo-core for SFT/DPO/GRPO
#1827 opened Aug 16, 2026 by farhatkevin Collaborator Loading…
20
fix: make numpy SFT cache key independent of tokenizer access order
#1826 opened Aug 14, 2026 by sylvesterkaczmarek Loading…
1
fix: record source provenance for numpy SFT caches
#1825 opened Aug 14, 2026 by sylvesterkaczmarek Loading…
1
ProTip! What’s not been updated in a month: updated:<2026-08-01.