Skip to content

Actions: kmbandy/llama.cpp

Actions

Server

Actions

Loading...
Loading

Show workflow options

Create status badge

Loading
server.yml
145 workflow runs
145 workflow runs

Filter by Event

⌥ + click/return to exclude

Filter by Status

⌥ + click/return to exclude

Filter by Branch

⌥ + click/return to exclude

Filter by Actor

⌥ + click/return to exclude
mt:: paged-attn — cold-tier (SSD) spill via fixed-slot per-layer file… Server #45: Commit 94df2ab pushed by kmbandy
26m 41s master
master
26m 41s
mt:: paged-attn — multi-slot admission control + whole-slot preemptio… Server #44: Commit 7af19e1 pushed by kmbandy
20m 33s master
master
20m 33s
mt:: paged-attn — multi-seq concurrent batching (MAD-124) Server #43: Commit d895d8c pushed by kmbandy
23m 50s master
master
23m 50s
mt:: paged-attn — hot↔warm tier infrastructure (MAD-120, partial) Server #42: Commit b54c5da pushed by kmbandy
12m 15s master
master
12m 15s
mt:: paged-attn — tier-aware GPU pool sizing (MAD-117) Server #41: Commit 5d65222 pushed by kmbandy
14m 2s master
master
14m 2s
mt:: paged-attn — quantized KV cache (Q8_0 + TURBO4_0) (MAD-116) Server #40: Commit c9d2c7c pushed by kmbandy
13m 59s master
master
13m 59s
mt:: paged-attn — chunked online-softmax, ctx > 16k unblocked (MAD-115) Server #39: Commit 7f62913 pushed by kmbandy
15m 23s master
master
15m 23s
mt:: paged-attn — separate scatter kernel as default + GGML_PAGED_FUS… Server #38: Commit 7cd6297 pushed by kmbandy
13m 17s master
master
13m 17s
mt:: paged-attn — fix WAR hazard on per-batch metadata tensors (MAD-114) Server #37: Commit 7ab1c77 pushed by kmbandy
26m 52s master
master
26m 52s
mt:: paged attn — LDS-overflow guard + honest --help text Server #36: Commit 4430198 pushed by kmbandy
15m 2s master
master
15m 2s
mt:: paged cache — accept layer filter + per-tensor type_k/type_v Server #35: Commit 4d55540 pushed by kmbandy
22m 39s master
master
22m 39s
mt:: tiered cache — per-seq tier metadata for --parallel > 1 Server #34: Commit 81e0b3e pushed by kmbandy
39m 25s master
master
39m 25s
mt:: paged cache — multi-seq compute_slot_mapping + init_batch Server #33: Commit 7ef076e pushed by kmbandy
12m 51s master
master
12m 51s
docs: update --kv-tier-paged-blocks help text + ignore .mneme-code Server #32: Commit d227a7a pushed by kmbandy
13m 57s master
master
13m 57s
mt:: paged attention — Phase 3.7 (kernel correctness, end-to-end cohe… Server #31: Commit ba3162d pushed by kmbandy
11m 53s master
master
11m 53s
mt:: paged attention — Phase 3.4a (model creation routes to llama_kv_… Server #30: Commit cf0bac1 pushed by kmbandy
1h 41m 8s master
master
1h 41m 8s
mt:: paged attention — Phase 3.1-3.3 (kernel + ggml op + cache class) Server #29: Commit 251ae6b pushed by kmbandy
1h 17m 7s master
master
1h 17m 7s
server: advance eviction cursor by range width, not positions returned Server #28: Commit b8bb63c pushed by kmbandy
1h 38m 41s master
master
1h 38m 41s
mt:: paged refactor — Phase 2b-C-3 (gated paged whole-seq wipe) Server #27: Commit 78ccad2 pushed by kmbandy
1h 26m 38s master
master
1h 26m 38s
mt:: paged refactor — Phase 2b-C-2 (gated paged restore read side) Server #26: Commit 75d2e4c pushed by kmbandy
1h 15m 0s master
master
1h 15m 0s
mt:: paged refactor — Phase 2b-C-1 (gated paged backup write side) Server #25: Commit 9f54b3b pushed by kmbandy
1h 2m 15s master
master
1h 2m 15s
mt:: paged refactor — Phase 2a wiring (BlockPool/BlockTable instantia… Server #24: Commit 77e5bcd pushed by kmbandy
50m 32s master
master
50m 32s
mt:: paged refactor — Phase 1 scaffolding (BlockPool + BlockTable) Server #23: Commit 19b71e7 pushed by kmbandy
42m 1s master
master
42m 1s
mt:: defensive assert — refuse to load with n_seq_max > 1 Server #22: Commit 01d8e29 pushed by kmbandy
12m 26s master
master
12m 26s
mt:: + server — wipe tier state on whole-seq seq_rm (agent task lifec… Server #21: Commit 64533a2 pushed by kmbandy
19m 12s master
master
19m 12s