Skip to content

Pull requests: THUDM/slime

New pull request New
56 Open 1,153 Closed
56 Open 1,153 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: resolve OOM in long-sequence training via conditional entropy gradient tracking run-ci-megatron
#1524 opened Jan 30, 2026 by ppraneth Loading…
2
Make DataSource implement __len__ to standardize the data source contract
#1518 opened Jan 29, 2026 by TSunny007 Loading…
2
fix: pp broadcast_object_list patch bug of megatron bridge
#1517 opened Jan 29, 2026 by Yangruipis Loading…
[Multimodal] make multimodal processing robust
#1516 opened Jan 29, 2026 by coding-famer Loading…
refactor: replace internal Ray API calls with public interfaces
#1508 opened Jan 28, 2026 by HappyCpp Loading…
[Feature] Support nvfp4 RL
#1505 opened Jan 27, 2026 by fy1214 Loading…
[FIX] model provider compatibility with Megatron-LM
#1500 opened Jan 27, 2026 by simondong1 Loading…
1
fix: fp16 config and grad scaling
#1495 opened Jan 26, 2026 by yxyOo Loading…
3
[Fix] support qwen3vl and qwen3vlmoe torch_dist to HF
#1492 opened Jan 26, 2026 by sxl1993 Loading…
add concurrent loading for eager dataset
#1485 opened Jan 23, 2026 by weixiao-zhan Loading…
[Fix] Allow default TIS function usage when get_mismatch_metrics is set
#1483 opened Jan 23, 2026 by eecspan Loading…
[Fix] Fix some tiny bugs in fault tolerance run-ci-megatron
#1480 opened Jan 23, 2026 by yitianlian Loading…
[Feature] Distribute Prefill on different node run-ci-megatron
#1478 opened Jan 22, 2026 by yitianlian Loading…
fix(rollout): raise error when buffer is insufficient without global dataset
#1474 opened Jan 21, 2026 by JackXu0 Loading…
perf(rollout): compress loss masks with run-length encoding
#1473 opened Jan 20, 2026 by JackXu0 Loading…
fix: remove enforced CPU initialization assertion for AMD GPUs
#1471 opened Jan 20, 2026 by Vivicai1005 Loading…
fix: validate reward_key when RM returns dict
#1470 opened Jan 20, 2026 by zpqiu Loading…
1
[VLM multi-turn] delete fsdp (#1141)
#1463 opened Jan 20, 2026 by gxlvera Loading…
fix: rename samples to seq_lens and remove dead code
#1453 opened Jan 19, 2026 by JackXu0 Loading…
[debug] fixed bug in sft_rollout
#1451 opened Jan 18, 2026 by MikaStars39 Loading…
@yitianlian
[Feature] Add multi task logging in slime
#1439 opened Jan 17, 2026 by yitianlian Loading…
Fix multi‑turn loss masking, clarify qwen25/qwen3 types, and strengthen mask tests
#1418 opened Jan 14, 2026 by Daucloud Loading…
issue #1414 fix: correct reward normalization for unequal group sizes
#1415 opened Jan 14, 2026 by ccggddmm Loading…
4
[1/2] feat/add_fault_torlance
#1404 opened Jan 13, 2026 by lilei199908 Loading…
8
support Opencua click dataset
#1393 opened Jan 12, 2026 by kafkayu Loading…
3 of 4 tasks
ProTip! Type g i on any issue or pull request to go back to the issue listing page.