Skip to content

Pull requests: facebookresearch/generative-recommenders

New pull request New
20 Open 73 Closed
20 Open 73 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
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Pull requests list

update block size for standalone_cint_v4 CLA Signed This label is managed by the Meta Open Source bot. fb-exported
#145 opened Nov 22, 2024 by zhaozhul Loading…
2
standalone_cint_v4 CLA Signed This label is managed by the Meta Open Source bot. fb-exported
#144 opened Nov 22, 2024 by zhaozhul Loading…
4
loop unroll for hstu attn bwd CLA Signed This label is managed by the Meta Open Source bot. fb-exported
#143 opened Nov 21, 2024 by LinjianMa Loading…
1
Convert directory fbcode/hammer to use the Ruff Formatter CLA Signed This label is managed by the Meta Open Source bot. fb-exported
#142 opened Nov 20, 2024 by tpolasek Loading…
1
Prepare for "Fix type-safety of torch.nn.Module instances": fbcode/h* CLA Signed This label is managed by the Meta Open Source bot. fb-exported
#141 opened Nov 20, 2024 by ezyang Loading…
3
Add complete_cumsum cpu and meta ops CLA Signed This label is managed by the Meta Open Source bot. fb-exported
#140 opened Nov 20, 2024 by jiyuanzFB Loading…
2
add pytorch implementations for jagged operations CLA Signed This label is managed by the Meta Open Source bot. fb-exported
#137 opened Nov 19, 2024 by zhaozhul Loading…
3
Replace Triton addmm with torch.addmm for AMD to achieve better training performance CLA Signed This label is managed by the Meta Open Source bot. fb-exported
#133 opened Nov 19, 2024 by yoyoyocmu Loading…
1
Change autotune key for ragged_hstu_attention to support dynamic batch size CLA Signed This label is managed by the Meta Open Source bot. fb-exported
#132 opened Nov 18, 2024 by AlbertDachiChen Loading…
10
Bug fix: detecting contextual length in triton hstu attn code CLA Signed This label is managed by the Meta Open Source bot. fb-exported
#130 opened Nov 18, 2024 by lic225 Loading…
1
Fix type-safety of torch.nn.Module instances CLA Signed This label is managed by the Meta Open Source bot. fb-exported
#129 opened Nov 18, 2024 by ezyang Loading…
1
move sorted_kv_pairs from hammer/ops/cuda/ to hammer/ops/cpp/ CLA Signed This label is managed by the Meta Open Source bot. fb-exported
#128 opened Nov 18, 2024 by jiyuanzFB Loading…
3
copy complete_cumsum kernel to ops/cpp/ CLA Signed This label is managed by the Meta Open Source bot. fb-exported
#127 opened Nov 15, 2024 by jiyuanzFB Loading…
3
Redefine FBGEMM targets with gpu_cpp_library (Re-land attempt of D64863809) CLA Signed This label is managed by the Meta Open Source bot. fb-exported
#126 opened Nov 15, 2024 by q10 Loading…
1
num_stages=0 becomes num_stages=2 CLA Signed This label is managed by the Meta Open Source bot. fb-exported
#125 opened Nov 14, 2024 by nmacchioni Loading…
3
Fix max_attn_len numerical CLA Signed This label is managed by the Meta Open Source bot. fb-exported
#124 opened Nov 14, 2024 by hanli0612 Loading…
1
Enable HSTU SMEM for TW and PW with autotuning for both SMEM preload and maxnregs CLA Signed This label is managed by the Meta Open Source bot. fb-exported
#92 opened Oct 7, 2024 by plotfi Loading…
7
[Triton SMEM] Add not-yet-landed usage of Triton SMEM feature with autotuning CLA Signed This label is managed by the Meta Open Source bot.
#72 opened Aug 22, 2024 by plotfi • Draft
[WIP] TMA Version of HSTU (Autotuned) CLA Signed This label is managed by the Meta Open Source bot.
#71 opened Aug 21, 2024 by plotfi Loading…
20
TMA version of hstu CLA Signed This label is managed by the Meta Open Source bot.
#57 opened Jul 24, 2024 by manman-ren • Draft
1
ProTip! Follow long discussions with comments:>50.