 Guillaume Wenzek
|
fa85f05545
test flash_attn
|
1 rok temu |
|
Guillaume Wenzek
|
06d4ed1475
ggml.ne != np.shape
|
2 lat temu |
|
Guillaume Wenzek
|
c2e6384e29
nb
|
2 lat temu |
|
Guillaume Wenzek
|
3f5912b973
forward
|
2 lat temu |
|
Guillaume Wenzek
|
c0bec21155
export model size in hparams
|
2 lat temu |
|
Guillaume Wenzek
|
772f90dfdc
load_fairseq2_ggml_file
|
2 lat temu |
|
Guillaume Wenzek
|
506dee42d8
move layers to fairseq2.cpp
|
2 lat temu |
|
Guillaume Wenzek
|
6f32f3c06f
unity_graph -> unity_audio_encoder_graph
|
2 lat temu |
|
Guillaume Wenzek
|
6cf3dfeb05
pass the input tensor explicitly
|
2 lat temu |
|
Guillaume Wenzek
|
22f8430903
clearly split between vendored ggml.py and our utilities
|
2 lat temu |
|
Guillaume Wenzek
|
9021fad301
use same convention than ggml.c
|
2 lat temu |
|
Guillaume Wenzek
|
e7dc2b86fc
add ggml python bindings
|
2 lat temu |
