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

b5045

35e592eb · vulkan: set cmake minimum and project name in vulkan-shaders (#12744) · Apr 04, 2025
b5043

c262bedd · CUDA: Prefer vector flash decoding kernel for Gemma models (#12738) · Apr 03, 2025
b5041

1c059995 · vulkan: Fix missing cmake logic for dot product extension (#12721) · Apr 03, 2025
b5039

5f696e88 · sync : minja (inclusionAI/Ling) and update tests (#12699) · Apr 03, 2025
b5038

193c3e03 · fix MUSA compiler warning (#12704) · Apr 03, 2025
b5037

65cfe136 · CANN: Support operator SIN COS ARGMAX (#12709) · Apr 03, 2025
b5036

3f9da22c · Simplify and improve CUDA graphs through use of indirect copy pointers (#9017) · Apr 03, 2025
b5035

2a0dc97e · CANN: Fix failed test cases (#12708) · Apr 03, 2025
b5034

97a20c01 · opencl: use `max_alloc_size` in backend ctx instead of querying again (#12705) · Apr 02, 2025
b5033

f01bd023 · vulkan: Implement split_k for coopmat2 flash attention. (#12627) · Apr 02, 2025
b5032

6f3bd386 · cmake: remove caching from vulkan coopmat checks (#12719) · Apr 02, 2025
b5031

be0a0f8c · vulkan: Implement grouped query attention in the coopmat2 FA shader (#12559) · Apr 02, 2025
b5030

92e3006b · Vulkan: Fix mmq int dot float cache size (#12722) · Apr 02, 2025
b5029

833e2b74 · model : print tensor size during load (#12711) · Apr 02, 2025
b5028

e0e912f4 · llama : add option to override model tensor buffers (#11397) · Apr 02, 2025
b5026

83a88bd6 · vocab : BailingMoE : change possessive quantifiers to greedy (#12677) · Apr 02, 2025
b5025

42eb248f · common : remove json.hpp from common.cpp (#12697) · Apr 02, 2025
b5022

f423981a · opencl : fix memory allocation size (#12649) · Apr 01, 2025
b5021

e39e727e · llama : use LLM_KV_GENERAL_FILE_TYPE instead of gguf_find_key (#12672) · Apr 01, 2025
b5019

3fd072a5 · metal : use F32 prec in FA kernels (#12688) · Apr 01, 2025

1
…
9
10
11
12
13
14
15
16
17
…
178