Skip to content

Pull requests: radixark/Megatron-LM

New pull request New
39 Open 48 Closed
39 Open 48 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

Fuse NVFP4 fake-QAT quantize-dequantize with CuTe DSL
#87 opened Aug 24, 2026 by ziang-and • Draft
3 of 6 tasks
Support bounded distributed-optimizer main initialization
#86 opened Aug 18, 2026 by zianglih Loading…
2 of 6 tasks
Support packed sequences in Gated Delta Net
#85 opened Aug 17, 2026 by yapdianang • Draft
fix: restore PP process group on batched pipeline P2P ops
#83 opened Aug 14, 2026 by Laz4rz Loading…
Make MoE router-weight placement explicit in experts
#79 opened Aug 6, 2026 by zianglih • Draft
Extend FP32 MoE numerics to low-precision recipes and registered activations
#77 opened Aug 5, 2026 by zianglih • Draft
5 of 10 tasks
Fix for lora dgrad flatten in _linear_dgrad_matmul
#76 opened Aug 5, 2026 by williammtan Loading…
[WIP] Add NVFP4 fake QAT for grouped experts
#75 opened Jul 31, 2026 by zianglih • Draft
6 tasks
1
Add opt-in FlashInfer-fast activations for rollout parity
#73 opened Jul 29, 2026 by zianglih • Draft
2 of 6 tasks
TOP dense parity for Qwen3 on Blackwell: flashinfer attention, batch-invariant kernel seam, non-reentrant recompute
#70 opened Jul 20, 2026 by adrenaline21 Loading…
Handle step key correctly in dp_reshardable checkpoint save with --optimizer-cpu-offload
#69 opened Jul 16, 2026 by artkorenev Loading…
1
fix(true-on-policy): match SGLang's row-linear k-tile reduction order
#65 opened Jul 9, 2026 by zihaow211 Loading…
1
[optim] run plan/metadata coordination over a gloo group
#62 opened Jul 7, 2026 by yueming-yuan Loading…
[optim] bucket checkpoint save to avoid CPU memory spike during ckpt saving
#61 opened Jul 6, 2026 by yueming-yuan Loading…
Fix GB300 torch_dist checkpoint save crashes from forked local writers
#57 opened Jun 17, 2026 by zyzshishui Loading…
6 tasks
Add TV (total-variation) loss option for the MTP draft head
#56 opened Jun 16, 2026 by ElliotXinqiWang Loading…
fix(mtp): rename MTP submodule transformer_layer -> mtp_model_layer
#54 opened Jun 8, 2026 by Zhichenzzz Loading…
1
[15/15] Cover fast top-k routing contract
#48 opened May 26, 2026 by maocheng23 Loading…
[14/15] Cover true-on-policy SGLang extension wiring
#47 opened May 26, 2026 by maocheng23 Loading…
[13/15] Wire fused MoE backward kernel launcher
#46 opened May 26, 2026 by maocheng23 Loading…
[12/15] Add fused MoE top-k weight backward kernel
#45 opened May 26, 2026 by maocheng23 Loading…
[11/15] Add fused MoE backward weight kernel
#44 opened May 26, 2026 by maocheng23 Loading…
[10/15] Add fused MoE backward input kernel
#43 opened May 26, 2026 by maocheng23 Loading…
[9/15] Add autograd wrapper for SGLang fused MoE
#42 opened May 26, 2026 by maocheng23 Loading…
[8/15] Add SGLang fused MoE forward wrapper
#41 opened May 26, 2026 by maocheng23 Loading…
ProTip! Follow long discussions with comments:>50.