Hugging Face's logo

HelloSun
/
ling30flash-moe-expert-paging-q4_EPQ

Text Generation
GGUF
bailingmoe3
ling
Mixture of Experts
expert-paging
ssd-to-ram
cpu
Model card Files
xet
 Community
ling30flash-moe-expert-paging-q4_EPQ

Commit History

16 days ago

  1. Add consolidated results report; mark EOG-era tables with correction notice

    HelloSun
    Verified signature
    d7d8b12

  2. Fix first-token drop; rerun policy sweep with raw completion; add 8 GiB 10-question QA check

    HelloSun
    Verified signature
    effa948

  3. Add LRU vs least_stale eviction-policy sweep at 8/16/32/56/64 GiB

    HelloSun
    Verified signature
    5a74a18

24 days ago

  1. 修正 arena sizing 預留量並加入核心層硬上限實測⁠

    opencode
    71c36bd

26 days ago

  1. 整理中文 model card 與 EPQ 專家分頁定位

    HelloSun
    16b1f1d

28 days ago

  1. README: llama-server section (serve.sh, port 22222), sdq_buft regex lifetime pitfall, results/server_audit

    HelloSun
    Verified signature
    c9167aa

  2. update sdq-ling/ (llama-server support + sdq_buft regex lifetime fix)

    HelloSun
    Verified signature
    b6e48af

  3. update results/ (llama-server support + sdq_buft regex lifetime fix)

    HelloSun
    Verified signature
    cdcaeb5

  4. add I/O honesty audit (mincore + O_DIRECT control run), fix RSS reporting (smaps_rollup has no RssFile key), re-run memory matrix with anon/file RSS split

    HelloSun
    Verified signature
    e4e298b

  5. Delete sdq-ling/__pycache__/verify_ling.cpython-312.pyc with huggingface_hub

    HelloSun
    Verified signature
    22929b5

  6. fix README encoding, drop __pycache__

    HelloSun
    Verified signature
    53ecda5

  7. Delete sdq-ling/__pycache__/verify_ling.cpython-312.pyc with huggingface_hub

    HelloSun
    Verified signature
    4559043

  8. Ling-3.0-flash Q4_K_M: SSD->RAM MoE expert pager (sdq-ling) + 8/16/32/56 GiB memory experiments and NumPy verification

    HelloSun
    Verified signature
    6d28d45

  9. initial commit

    HelloSun
    Verified signature
    b9395de