Skip to content

Pull requests: antirez/ds4

New pull request New
56 Open 65 Closed
56 Open 65 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

fix: resilient DSML tool-call parsing and repair in long contexts
#192 opened May 18, 2026 by jackygurui Loading…
Fast CUDA backend: vendored mmq matmul, VMM weight layout, speculative-decoding plumbing
#187 opened May 18, 2026 by Entrpi Loading…
Compressed KV disk cache via streaming LZ4
#186 opened May 18, 2026 by jasontitus Loading…
eval: regrade traces and tighten answer extraction
#185 opened May 18, 2026 by fitchmultz Loading…
2
[codex] harden Metal failure recovery
#184 opened May 18, 2026 by Chedrian07 • Draft
rocm: add wmma indexer support
#180 opened May 17, 2026 by alantsev Loading…
1
[codex] Fix CUDA object rebuild on config changes
#178 opened May 17, 2026 by audreyt Contributor • Draft
fix(kv-cache): boost eviction score for cold/evict/shutdown checkpoints
#177 opened May 17, 2026 by unsaltedbutter-ai Contributor Loading…
4 tasks
1
fix(kv-cache): drop superseded continued snapshots on store
#175 opened May 17, 2026 by unsaltedbutter-ai Contributor Loading…
1
Metal: speed up M5 Max decode indexer
#169 opened May 16, 2026 by fitchmultz Loading…
16
Steering MoE LLMs via Expert (De)Activation
#168 opened May 16, 2026 by Chida82 Loading…
1
cuda: fix HMM path on coherent unified-memory systems (GB10 NVLink-C2C)
#158 opened May 15, 2026 by Mikehutu Loading…
4
Fix ROCm build/runtime naming and MTP model mapping
#156 opened May 15, 2026 by adv0r Loading…
cuda: add direct-model partial weight cache
#153 opened May 15, 2026 by ddxxlao Loading…
6
Metal: correctness-gate M5 Max 4096 prefill (+5%)
#149 opened May 14, 2026 by fitchmultz Loading…
2
server: add tool-safe directional steering policy ideas-to-explore steering
#148 opened May 14, 2026 by audreyt Contributor Loading…
9
A launcher for local coding-agent clients like Pi, Opencode, Copilot ideas-to-explore
#147 opened May 14, 2026 by Chida82 Loading…
3
server: guard disk KV cache with CRC32C
#146 opened May 14, 2026 by Ghatage Contributor Loading…
2
cuda: add launch-bounded tile8 MoE down path
#145 opened May 14, 2026 by amarrmb Loading…
docs(bench): add M5 max
#143 opened May 14, 2026 by fry69 Loading…
2
Guard KV cache against page-cache pressure kv-cache no-brainer 💃
#134 opened May 14, 2026 by Ghatage Contributor Loading…
1
[codex] Add Metal TurboQuant KV cache kv-cache likely-risky
#124 opened May 14, 2026 by Chedrian07 • Draft
1
cuda: skip ordered f16 matmul on Blackwell
#121 opened May 13, 2026 by amarrmb Loading…
flock: unlink lock file on release
#119 opened May 13, 2026 by nopjmp Loading…
ROCm fix: replace host-side rsqrtf with 1/sqrtf for HIP compatibility
#117 opened May 13, 2026 by reffdev Loading…
ProTip! What’s not been updated in a month: updated:<2026-04-18.