Default Branch

4fd59e8427 · ggml-cuda: use CMAKE_CUDA_ARCHITECTURES if set when GGML_NATIVE=ON (#18413) · Updated 2025-12-27 17:33:14 -08:00

Branches

3cb1c348b3 · metal : try to improve batched decoding · Updated 2023-12-01 12:01:58 -08:00    happyz

5955
2

eb594c0f7d · alloc : fix build with debug · Updated 2023-12-01 00:46:05 -08:00    happyz

5979
14

5b74310e6e · build : enable libstdc++ assertions for debug builds · Updated 2023-11-30 15:18:24 -08:00    happyz

5964
1

bb39b87964 · ggml : restore abort() in GGML_ASSERT · Updated 2023-11-27 16:27:09 -08:00    happyz

5983
1

87f4102a70 · llama : revert n_threads_batch logic · Updated 2023-11-27 11:47:35 -08:00    happyz

5984
3

6272b6764a · use stride=128 if built for tensor cores · Updated 2023-11-27 10:09:14 -08:00    happyz

5987
3

8d8b76d469 · lookahead : add comments · Updated 2023-11-26 01:26:55 -08:00    happyz

5999
9

21b70babf7 · straightforward /v1/models endpoint · Updated 2023-11-24 08:22:39 -08:00    happyz

6000
12

f8e9f11428 · common : add -dkvc arg for enabling kv cache dumps · Updated 2023-11-23 08:47:56 -08:00    happyz

6006
4

f824902623 · YaRN : correction to GPT-NeoX implementation · Updated 2023-11-15 14:10:52 -08:00    happyz

6038
1

d0445a2eff · better documentation · Updated 2023-11-09 16:38:20 -08:00    happyz

6055
3

47d604fa2d · fix issues · Updated 2023-11-05 04:20:22 -08:00    happyz

6069
3

3ef358fffd · Revert "cuda : use CUDA memory pool with async memory allocation/deallocation when available (#3903)" · Updated 2023-11-04 13:26:51 -07:00    happyz

6073
2

46868a499e · metal : multi-simd softmax · Updated 2023-11-01 12:16:34 -07:00    happyz

6098
1

a8796f9609 · llm : cleanup + comments · Updated 2023-11-01 11:08:02 -07:00    happyz

6107
4

7420bef83e · wip wip wip · Updated 2023-10-31 23:51:43 -07:00    happyz

6107
1

afb3929279 · Merge branch 'master' into llama-refactor · Updated 2023-10-31 11:35:31 -07:00    happyz

6109
21

29fe516913 · wip · Updated 2023-10-31 09:36:37 -07:00    happyz

6110
1

dab42893c9 · scripts : working curl pipe · Updated 2023-10-31 08:03:56 -07:00    happyz

6110
3

7923b70cb8 · llama : add llm_build_inp_embd helper · Updated 2023-10-31 07:43:08 -07:00    happyz

6115
37