Skip to content

Pull requests: hw-native-sys/pypto-serving

New pull request New
32 Open 70 Closed
32 Open 70 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

feat(deepseek): support MTP across chunked prefill
#132 opened Aug 1, 2026 by vegetabledoww Loading…
1
8
feat(serving): support async scheduling for DeepSeek MTP decode
#130 opened Jul 30, 2026 by ndleslx Collaborator Loading…
1
feat: add DeepSeek V4 W8A8 checkpoint converter
#128 opened Jul 30, 2026 by vegetabledoww Loading…
11
feat(qwen): size KV cache with simpler committed_device_memory
#126 opened Jul 30, 2026 by sunghajung6688 Contributor Loading…
1
4
Add DeepSeek V4 serving profiling skill
#125 opened Jul 30, 2026 by high-cloud Contributor Loading…
8
Optimize DeepSeek V4 startup with reusable programs and prepacked weights
#124 opened Jul 29, 2026 by puddingfjz Contributor Loading…
6
feat(skills): add DeepSeek V4 online SA_PROFILE perf skill
#121 opened Jul 28, 2026 by sunghajung6688 Contributor Loading…
5
feat(deepseek): support arbitrary MTP draft depth
#120 opened Jul 28, 2026 by hashiqiqixian Loading…
6
[feature] TQ serving: add TurboQuant prefill/decode branch alongside …
#114 opened Jul 24, 2026 by sunghajung6688 Contributor Loading…
1
4
ci: extract shared self-hosted setup into composite actions
#110 opened Jul 23, 2026 by luohuan19 Loading…
4
fix(deepseek): follow pypto-lib kernel dir rename v4 -> v4-flash
#108 opened Jul 22, 2026 by lwDavid Contributor • Draft
6
fix(qwen): default PTO2_RING_HEAP to 512MB so a plain Qwen3-14B launch doesn't deadlock
#106 opened Jul 21, 2026 by lterrac Collaborator Loading…
4
feat: add stage-one radix prefix cache
#99 opened Jul 20, 2026 by zmnobug Contributor Loading…
1
7
fix offline batch chunked prefill
#96 opened Jul 16, 2026 by vegetabledoww Loading…
20
ci: prevent orphaned DeepSeek V4 engines on test interruption
#86 opened Jul 14, 2026 by luohuan19 Loading…
14
decouple qwen3 serving from lib contract
#85 opened Jul 14, 2026 by bumble0918 Collaborator Loading…
9
feat(platform): add heartbeat module for replica health monitoring
#78 opened Jul 13, 2026 by lterrac Collaborator • Draft
3 tasks
6
feat(platform): execution modules and co-located coordinator/replica
#76 opened Jul 13, 2026 by lterrac Collaborator • Draft
4 tasks
11
perf(qwen3): compile prefill/decode/greedy straight from the signature
#73 opened Jul 13, 2026 by lyfne123 Contributor Loading…
3
perf(qwen3): skip redundant decode d2h/host copies in greedy path
#71 opened Jul 13, 2026 by lyfne123 Contributor Loading…
4
Integrate Qwen3 CANN FAI decode
#69 opened Jul 13, 2026 by ChaoWao • Draft
4
Add serving progress and adjust deepseek weight load
#63 opened Jul 9, 2026 by superxf Collaborator Loading…
19
Add Qwen3 A8W8 L2 runner foundation
#61 opened Jul 9, 2026 by vegetabledoww Loading…
2
Optimize Qwen3 device sampling outputs
#57 opened Jul 6, 2026 by zmnobug Contributor Loading…
10
feat(platform): add configuration types, module layer, and host control plane
#53 opened Jul 2, 2026 by lterrac Collaborator Loading…
60
ProTip! Type g i on any issue or pull request to go back to the issue listing page.