Skip to content

Pull requests: pytorch/torchtitan

New pull request New
190 Open 1,535 Closed
190 Open 1,535 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: enable torch.autocast for TP parallelism without FSDP CLA Signed This label is managed by the Meta Open Source bot.
#2213 opened Jan 8, 2026 by eous Loading…
5
feat(moe): add topk_before_score routing and use_router_bias support CLA Signed This label is managed by the Meta Open Source bot.
#2212 opened Jan 8, 2026 by eous Loading…
@shuhuayu
27
fix(gpt-oss): correct attention sink from sigmoid to LSE renormalization CLA Signed This label is managed by the Meta Open Source bot.
#2211 opened Jan 8, 2026 by eous Loading…
@wwwjn
4
feat: add differential learning rate and weight decay support CLA Signed This label is managed by the Meta Open Source bot.
#2210 opened Jan 8, 2026 by eous Loading…
13
[HybridEP] Support hybridEP for GB200 with NVL72 CLA Signed This label is managed by the Meta Open Source bot.
#2207 opened Jan 8, 2026 by elfiegg Loading…
@shuhuayu
17
feat(gpt-oss): Add CPU offload optimizer, differential LR/WD, and more CLA Signed This label is managed by the Meta Open Source bot.
#2205 opened Jan 7, 2026 by eous Loading…
31
[rl] Use vllm.Attention for trainer. ciflow/8gpu CLA Signed This label is managed by the Meta Open Source bot.
#2198 opened Jan 5, 2026 by zhxchen17 • Draft
2
[rl] refactor model registery ciflow/8gpu CLA Signed This label is managed by the Meta Open Source bot.
#2194 opened Jan 2, 2026 by wwwjn Loading…
2
[rl] Using JobConfig as the centralized config system for inference and simple GRPO ciflow/8gpu CLA Signed This label is managed by the Meta Open Source bot.
#2191 opened Jan 2, 2026 by wwwjn Loading…
31
use comms in compiler toolkit ciflow/8gpu CLA Signed This label is managed by the Meta Open Source bot.
#2190 opened Dec 31, 2025 by dolpm • Draft
3
experiments: add nemotron3 model to experiments folder CLA Signed This label is managed by the Meta Open Source bot.
#2187 opened Dec 30, 2025 by aghilann Loading…
4 tasks
15
auto-chunk unembed & loss ciflow/8gpu CLA Signed This label is managed by the Meta Open Source bot.
#2186 opened Dec 29, 2025 by shunting314 Loading…
11
Add epoch-based training support
#2182 opened Dec 28, 2025 by yurekami Loading…
5 tasks
1
8
[rl] Update callsite to init_batch_invariance to pass attention backend. ciflow/8gpu CLA Signed This label is managed by the Meta Open Source bot.
#2176 opened Dec 24, 2025 by zhxchen17 Loading…
15
compiler_toolkit: inputs are not DTensor if TP is not enabled CLA Signed This label is managed by the Meta Open Source bot.
#2175 opened Dec 24, 2025 by yanboliang Loading…
4
Add Flex flash backend to flex attention module ciflow/8gpu CLA Signed This label is managed by the Meta Open Source bot.
#2171 opened Dec 22, 2025 by drisspg • Draft
11
[do not land] trying invoke_subgraph on torchtitan ciflow/8gpu CLA Signed This label is managed by the Meta Open Source bot.
#2169 opened Dec 19, 2025 by yushangdi • Draft
2
[transformers_modeling_backend] Upgrade transformers from 4.57.1 to 5.0.0rc0 CLA Signed This label is managed by the Meta Open Source bot.
#2154 opened Dec 15, 2025 by 3outeille Loading…
1
5
[WIP] Use all DTensor for Qwen3 and llama4 at TP region ciflow/8gpu CLA Signed This label is managed by the Meta Open Source bot.
#2149 opened Dec 12, 2025 by wwwjn Loading…
13
Staging SFT training CLA Signed This label is managed by the Meta Open Source bot.
#2148 opened Dec 12, 2025 by rakkit Loading…
106
Add repeated_subgraphs option in AutoParallel example CLA Signed This label is managed by the Meta Open Source bot.
#2138 opened Dec 10, 2025 by fmassa Loading…
1
[Not Ready] Enable Async TP CI ciflow/8gpu CLA Signed This label is managed by the Meta Open Source bot.
#2125 opened Dec 8, 2025 by fegin • Draft
improve throughput of HF dense model (no need actually) CLA Signed This label is managed by the Meta Open Source bot.
#2122 opened Dec 8, 2025 by 3outeille • Draft
perf(pipeline): implement auto-partition algorithm CLA Signed This label is managed by the Meta Open Source bot. enhancement New feature or request
#2113 opened Dec 5, 2025 by TXacs Loading…
24
[simple_fsdp] Turn on bucketing by default CLA Signed This label is managed by the Meta Open Source bot.
#2103 opened Dec 3, 2025 by IvanKobzarev Loading…
11
ProTip! What’s not been updated in a month: updated:<2025-12-30.