-
Notifications
You must be signed in to change notification settings - Fork 55
Pull requests: hw-native-sys/pypto-lib
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
Refactor: fold DeepSeek-V4 Flash MTP split-K peels into matmul_acc init_cond
#1048
opened Aug 26, 2026 by
Hzfengsy
Member
Loading…
Perf: compact DeepSeek V4 CSA indexer pair arena
#1046
opened Aug 26, 2026 by
sjduan
Contributor
Loading…
Perf: pipeline HCA AllToAll into O-A
#1040
opened Aug 25, 2026 by
wangqin1723-max
Collaborator
Loading…
Perf: coarsen HCA stream merge workers
#1038
opened Aug 25, 2026 by
wangqin1723-max
Collaborator
Loading…
Perf: eliminate HCA raw KV GM transpose
#1035
opened Aug 25, 2026 by
wangqin1723-max
Collaborator
Loading…
Perf: fuse attention output with AllToAll
#1034
opened Aug 25, 2026 by
wangqin1723-max
Collaborator
Loading…
Add: use PyPTO paged attention in Qwen3 decode
#1031
opened Aug 25, 2026 by
Little-oil
Contributor
Loading…
Perf: fuse DeepSeek V4 SWA and HCA O projection reduce-scatter
#1025
opened Aug 24, 2026 by
wangqin1723-max
Collaborator
Loading…
Add: expose DeepSeek V4 serving capabilities
#1023
opened Aug 24, 2026 by
vegetabledoww
Contributor
Loading…
Add: DeepSeek-V4 DSpark confidence head
#1020
opened Aug 24, 2026 by
hashiqiqixian
Contributor
Loading…
Add: dump DeepSeek V4 MoE expert load statistics
#1016
opened Aug 23, 2026 by
high-cloud
Collaborator
Loading…
Add: DeepSeek V4 Flash temperature sampling
#1015
opened Aug 23, 2026 by
high-cloud
Collaborator
Loading…
CI: install with uv and default the toolchain to bundle
#952
opened Aug 13, 2026 by
luohuan19
Contributor
Loading…
Add: replace Qwen3 decode paged attention with PyPTO
#933
opened Aug 11, 2026 by
Little-oil
Contributor
Loading…
Fix deterministic DeepSeek split-K reductions
#927
opened Aug 10, 2026 by
high-cloud
Collaborator
•
Draft
Add: experimental shared-expert weight TP on fixed SP
#913
opened Aug 7, 2026 by
wangqin1723-max
Collaborator
Loading…
Add: use MXFP8 and MXFP4 in DeepSeek V4 Pro
#911
opened Aug 6, 2026 by
yanghaoran29
Contributor
•
Draft
EXPERIMENT (do not merge): A/B the A5 daily sweep against ptoas op fusion
wontfix
This will not be worked on
Perf: assemble SWA decode window in L1, drop the GM gather
#882
opened Jul 31, 2026 by
Hzfengsy
Member
Loading…
CI: bench decode_attention_csa on fixed even/odd cards
#881
opened Jul 31, 2026 by
luohuan19
Contributor
Loading…
Fix(deepseek-v4 pro): let MoE post_ffn use automatic dependency tracking
#879
opened Jul 31, 2026 by
yanghaoran29
Contributor
Loading…
2 of 3 tasks
Previous Next
ProTip!
Follow long discussions with comments:>50.