Niko1221 7be42963e6 Version 0.2.12 / engine 0.1.58: Terminal and app 0.2.12, Control and every engine package 0.1.58 (the 0.2.11 bump list, plus the v0.1.58 line in the setup.py version history)

2026-11-06 19:21:19 +01:00

..

include

Mac DeepSeek: the vision tower on Metal (kernels, BF16 GEMM, host side), capability marker STRATA_DS4_METAL_VISION_V1; credit to antirez/ds4 (MIT, 890898cf)

2026-11-02 19:16:42 +01:00

kernels

merge f-0215-mac-pp (host-row pipeline, qsa_merge_h, ragged expert tiles) into f-0215-mac-int

2026-11-06 01:25:09 +01:00

probe

Metal probe: busy_flag_probe shows the DS4 busy-spin kernel does not see the host's flag write (#4771)

2026-10-31 13:34:26 +01:00

src

Admit the PCIe share's fetched misses to the GPU cache (opt-in STRATA_ADAPT_FETCH=1|2)

2026-11-06 15:47:43 +01:00

tests

Metal prompt: host rows beside the GPU (embedding + PLE rows of the next chunk on a thread, parallel PLE gather on Macs), qsa_merge_h (merge writes the fp16 operand), ragged expert tiles (mm_gu6/mm_down6/mm_scan6, STRATA_PROMPT_V6), STRATA_PROMPT_KPROF per-dispatch profile, matmul2d probes

2026-11-05 20:36:54 +01:00

tools

Merge f-dsv41-mac (DeepSeek on Metal) into v0.2.7

2026-10-27 16:44:27 +01:00

CMakeLists.txt

Version 0.2.12 / engine 0.1.58: Terminal and app 0.2.12, Control and every engine package 0.1.58 (the 0.2.11 bump list, plus the v0.1.58 line in the setup.py version history)

2026-11-06 19:21:19 +01:00

setup_mac.py

Mac: DFlash2 is the default drafter for UD-Q4_K_XL and UD-IQ4_XS-v31, with Strata's tuned drafter at --draft-conf 0.7

2026-11-06 07:52:31 +01:00