Niko1221 c62d5e3a15 DeepSeek plan: the NVIDIA cards it counts are the ones setup runs (a card of 8 GB or less starts Off beside a bigger one), and the plan says which card is not used

2026-10-31 03:23:01 +01:00

..

core

engine: wait up to 20 s for the RAM the expert arena needs, then load with a plainer warning (#4051)

2026-10-29 00:01:54 +01:00

cuda

diagnostics: STRATA_QSA_STAT (overlap of consecutive queries' selections), hip_pa_bench target and PA_BENCH_REF (compare two variants of the prompt attention on a recorded selection)

2026-10-30 15:41:53 +01:00

data

Two GTX 1080s are a working PC: an 8 GB card starts Off only beside a bigger card, and a PC whose only NVIDIA cards are Pascal uses them on the CUDA 12 engine (#3655)

2026-10-22 13:47:41 +02:00

desktop

Delete a model: every part of a multi-part GGUF goes with it (the run config names only the first), and a part another model names stays (#3716)

2026-10-26 18:19:38 +01:00

dsv41

Merge f-034-cand (0.2.9 candidate) into f-036-strix

2026-10-30 10:52:31 +01:00

fixtures/dsv41

Accept ISTA-DASLab's six-shard DeepSeek V4.1 Flash GGUF layout

2026-10-30 00:16:56 +01:00

gdn

GDN: production-linked parity/bench harness, dispatch probe, FP64 fixtures, contract doc; opt-in STRATA_GDN_TC on 99-KB cards (split v/T/P buffering)

2026-10-30 01:12:19 +01:00

hip

GDN gfx12: conv + SiLU + q/k norm written as FP16 and read by the chunked kernels (one entry, STRATA_GDN_FUSE=0 off); ablation hooks removed

2026-10-30 01:51:05 +01:00

platform

Merge branch 'f-142-mgpu' into f-142

2026-10-09 14:47:08 +02:00

prefill

CPU prefill share: skip the default when host memory is short, and never fail a chat on its buffers

2026-10-09 15:10:32 +02:00

program

Later-card VRAM: audit fixes for failed allocations, the reserve arithmetic as a tested helper, a per-card out-of-memory report (#3638 #3702 #3797)

2026-10-27 02:02:46 +01:00

sycl

Setup and the launcher write --spec 2 for exactly two Arc B-series cards

2026-10-29 15:52:37 +01:00

terminal

DeepSeek plan: the NVIDIA cards it counts are the ones setup runs (a card of 8 GB or less starts Off beside a bigger one), and the plan says which card is not used

2026-10-31 03:23:01 +01:00

time_skip

Time-skip and tune job records are replaced with a wait for a reader (Windows 'Access is denied' when the host reads the file at that moment: it ended jobs at random)

2026-10-16 13:29:12 +02:00

tuning

Presets: a value outside the range the tuner tries is kept with a warning, not dropped (#3844)

2026-10-27 02:02:46 +01:00