为了安全，强烈建议开启2FA双因子认证：User Settings -> Account -> Enable two-factor authentication！！！

Tags

Tags give the ability to mark specific points in history as being important

b4666

ed926d88 · llama : fix defrag logic (#11707) · Feb 07, 2025
b4663

c026ba3c · vulkan: print shared memory size (#11719) · Feb 07, 2025
b4662

7ee953a6 · llama : add llama_sampler_init for safe usage of llama_sampler_free (#11727) · Feb 07, 2025
b4661

ec3bc827 · SYCL: remove XMX info from print devices (#11712) · Feb 07, 2025
b4660

b7552cfc · common : add default embeddings presets (#11677) · Feb 07, 2025
b4659

225bbbfa · ggml : optimize and build warning fix for LoongArch (#11709) · Feb 07, 2025
b4658

855cd073 · llama : fix old glm4 models (#11670) · Feb 06, 2025
b4657

8a59053f · sync : ggml · Feb 06, 2025
b4651

c0d48432 · build : fix llama.pc (#11658) · Feb 06, 2025
b4649

2c6c8df5 · vulkan: optimize coopmat2 iq2/iq3 callbacks (#11521) · Feb 06, 2025
b4648

8a7e3bf1 · vulkan: initial support for IQ4_XS quantization (#11501) · Feb 06, 2025
b4647

1b598b30 · vulkan: use smaller combined allocations to avoid fragmentation (#11551) · Feb 06, 2025
b4646

902368a0 · metal : avoid breaking build when metal API predates TARGET_OS_VISION (#11690) · Feb 06, 2025
b4644

d774ab3a · metal : adjust support conditions for norm operators (#11671) · Feb 05, 2025
b4643

fa62da9b · CUDA: support for mat. mul. with ne03 != ne13 (#11656) · Feb 05, 2025
b4642

1ec20808 · llava: add quantization for the visual projector LLAVA, Qwen2VL (#11644) · Feb 05, 2025
b4641

9f4cc8f8 · `sync`: minja (#11641) · Feb 05, 2025
b4640

fd08255d · CUDA: non-contiguous (RMS) norm support (#11659) · Feb 04, 2025
b4639

3ec9fd4b · HIP: force max threads per block to be 1024 (#11621) · Feb 04, 2025
b4637

1bef571f · arg : list RPC devices first when using --list-devices (#11655) · Feb 04, 2025

1
…
22
23
24
25
26
27
28
29
30
…
178