Commit History

Author SHA1 Message Date
  Guillaume Wenzek e7c3b7a4ba allow flash attn 2 years ago
  Guillaume Wenzek b07a08102a MultiheadAttention_forward 2 years ago
  Guillaume Wenzek 3f1d6992f3 fix to_numpy for transposed tensors 2 years ago
  Guillaume Wenzek a80a3b49f3 remove un-needed code 2 years ago
  Ning 4aaca94e11 Self attention (SDPA) in unity.cpp (#34) 2 years ago
  Guillaume Wenzek be236b8584 remove size/init functions 2 years ago
  Guillaume Wenzek fa85f05545 test flash_attn 2 years ago
  Guillaume Wenzek 06d4ed1475 ggml.ne != np.shape 2 years ago
  Guillaume Wenzek c2e6384e29 nb 2 years ago
  Guillaume Wenzek 3f5912b973 forward 2 years ago
  Guillaume Wenzek 2b6130855c handle 4 dims tensors and eof 2 years ago
  Guillaume Wenzek c0bec21155 export model size in hparams 2 years ago
  Guillaume Wenzek d6425f84b3 black + isort 2 years ago
  Guillaume Wenzek b1c4943c4e merge stuff into ggml_convert.py 2 years ago
  Guillaume Wenzek 772f90dfdc load_fairseq2_ggml_file 2 years ago
  Guillaume Wenzek 506dee42d8 move layers to fairseq2.cpp 2 years ago
  Guillaume Wenzek 5024f43b15 layer norm 2 years ago
  Guillaume Wenzek 6f32f3c06f unity_graph -> unity_audio_encoder_graph 2 years ago
  Guillaume Wenzek 6cf3dfeb05 pass the input tensor explicitly 2 years ago
  Guillaume Wenzek 22f8430903 clearly split between vendored ggml.py and our utilities 2 years ago
  Guillaume Wenzek 9021fad301 use same convention than ggml.c 2 years ago
  Guillaume Wenzek e7dc2b86fc add ggml python bindings 2 years ago
  Naji El Hachem 853b53bad0 start working out the loader api 2 years ago
  cndn d7fcd9ec44 pos enc (WIP SDPA) 2 years ago
  cndn ff5cbf57c9 More encoder states loading & pos enc 2 years ago
  cndn 625b69883d Matched results for 1 FFN layer 2 years ago
  cndn db2570ec38 model weight load & dev setup 2 years ago
  cndn e73d4eb3f1 init unity.cpp setup 2 years ago
  cndn d80093f9f8 Import ggml to SC 2 years ago
  Anna Sun 14b013315e [streaming] add s2s + s2t expressive demo (#153) 2 years ago