auto-round
Reduce peak gpu memory usage and support moe estimation
#981

Merged

Reduce peak gpu memory usage and support moe estimation #981

xin3he merged 21 commits into main from xinhe/fix

add moe support

0055d30e

use low_gpu_mem_usage to cache best params

29df3579

add warning for memory estimation

d8917f9d

xin3he changed the title ~~Xinhe/fix~~ Reduce peak gpu memory usage and support moe estimation 352 days ago

xin3he requested a review from

wenhuach21 352 days ago

xin3he removed review request from

wenhuach21 352 days ago

xin3he requested a review from

n1ck-guo 352 days ago

xin3he requested a review from

wenhuach21 352 days ago

xin3he requested a review from

yiliu30 352 days ago

fix bug

e04db30c

xin3he force pushed from f98c8119 to c1568dcf 352 days ago

support ds on CUDA and 70b on XPU

c1568dcf

fix oom of deepseek

550439d8

fix pylint

cbd02db7

threshold is 0.8

f4666772

wenhuach21 commented on 2025-11-04

Tighten constraints to avoid performance degradation cases

8942f58f

xin3he requested a review from

copilot-pull-request-reviewer 351 days ago

copilot-pull-request-reviewer commented on 2025-11-04

wenhuach21 commented on 2025-11-04

update per review comments

a5d455e5

xin3he commented on 2025-11-04

add clear_memory back

cde0aa10

xin3he requested a review from

wenhuach21 351 days ago

fix bug

518078bd

move loss device to the second card if card_0_in_high_risk

a27e1dd0

Merge branch 'main' into xinhe/fix

f77405de

wenhuach21 commented on 2025-11-05

update per review comments

74c45716

add comments

f155941e

add sync

9e209cd4

xin3he requested a review from

wenhuach21 349 days ago

xin3he requested a review from

WeiweiZhang1 349 days ago

wenhuach21 commented on 2025-11-06

remove duplicate clear_memory and add warning message

b7d13915

wenhuach21 commented on 2025-11-06

wenhuach21 approved these changes on 2025-11-06

update per review comments

0aaaa09b

Merge remote-tracking branch 'origin/main' into xinhe/fix

b8418964

rename pick samples to batch size

f530032c

xin3he merged 84e9a776 into main 349 days ago

xin3he deleted the xinhe/fix branch 349 days ago

xin3he added this to the 0.9.1 milestone 349 days ago

xin3he added Intel GPU

xin3he removed this from to the 0.9.1 milestone 349 days ago

xin3he added this to the 0.9.0 milestone 349 days ago

Reviewers

wenhuach21

copilot-pull-request-reviewer

n1ck-guo

yiliu30

WeiweiZhang1

Assignees

No one assigned

Labels

Intel GPU

Milestone

0.9.0

auto-round Reduce peak gpu memory usage and support moe estimation #981 Merged

Reduce peak gpu memory usage and support moe estimation #981

auto-round
Reduce peak gpu memory usage and support moe estimation
#981

Merged