mfritsche 0ebb855635 docs: int4 closed — no weight-only int4 dtype on RK3588 (hardware limit) ...

Probed rknn_matmul_create for every int4 dtype. RK3588 supports ONLY W8A8 and
W4A4. All weight-only int4 modes (int8xint4, fp16xint4) are rejected:
"unsupported matmul dtype ... in this platform". So int4 forces W4A4 = int4
activations, whose transformer outliers dont fit in 4 bits -

2026-07-10 16:58:33 +02:00

..

android

android: fix missing screenshots for Android.md (#18156)

2025-12-19 09:32:04 +02:00

backend

sycl : Improve SYCL doc (#23025)

2026-06-04 08:02:54 +03:00

development

docs : update HOWTO-add-model.md (#23883)

2026-06-02 11:40:22 +02:00

multimodal

docs : fix duplicated "the" in granitevision and model-conversion docs (#23767)

2026-05-27 09:34:06 +02:00

ops

[SYCL] Support Q4_1, Q5_0, Q5_1 in Flash-attention (#23812)

2026-06-01 09:53:53 +03:00

android.md

android: fix missing screenshots for Android.md (#18156)

2025-12-19 09:32:04 +02:00

autoparser.md

common : fix env names to all have LLAMA_ARG_ prefix (#23778)

2026-05-27 14:52:47 +03:00

build-riscv64-spacemit.md

ci : update spacemit toolchain url and enhance curl command (#23642)

2026-05-25 10:43:24 +02:00

build-s390x.md

docs: update s390x build docs (#19643)

2026-02-16 00:33:34 +08:00

build.md

docs : update ZenDNN docs for Q8 support (#23791)

2026-05-31 10:26:42 +02:00

docker.md

sycl : Improve SYCL doc (#23025)

2026-06-04 08:02:54 +03:00

function-calling.md

docs: Update documentation with Granite 4.0/4.1 (#23404)

2026-05-22 20:35:46 +08:00

install.md

docs : add "Quick start" section for new users (#13862)

2025-06-03 13:09:36 +02:00

llguidance.md

…

multi-gpu.md

Write a readme on Multi-GPU usage in llama.cpp (#22729)

2026-05-07 17:48:40 +02:00

multimodal.md

docs: listing qwen3-asr and qwen3-omni as supported (#21857)

2026-04-13 22:28:17 +02:00

ops.md

[SYCL] Add more types in GET_ROWS OP (#23710)

2026-06-01 09:53:04 +03:00

preset.md

preset: allow named remote preset (#18728)

2026-01-10 15:12:29 +01:00

rknpu2-campaign-results.md

docs: int4 closed — no weight-only int4 dtype on RK3588 (hardware limit)

2026-07-10 16:58:33 +02:00

speculative.md

server-bench : add speed-bench for speculative decoding benchmarking (#23869)

2026-05-29 23:09:47 +02:00