Tri Dao
|
c3b2196652
Add Alibi to MHA, test with Baichuan-13B
|
1 anno fa |
Tri Dao
|
3557e0bb8f
[MLP] Implement SwiGLU with torch jiterator
|
1 anno fa |
Tri Dao
|
f1a73d0740
Run isort and black on python files
|
1 anno fa |
Tri Dao
|
364a5b4a71
[MLP] Change the check for out_features being None
|
1 anno fa |
Tri Dao
|
4c98d0b41f
[MLP] Edit ParallelGatedMlp
|
1 anno fa |
Haodong Lyu
|
8ee62efca3
Implement ParallelGatedMlp (#251)
|
1 anno fa |
Tri Dao
|
75e334d407
[MLP] Add ParallelMLP
|
1 anno fa |
Tri Dao
|
96d10f6545
Implement LLaMa
|
1 anno fa |
Tri Dao
|
b630aef53f
Implement GatedMlp
|
1 anno fa |
Zhiyuan Chen
|
8c42415664
make mlp hidden_features defaults to 4*in_features
|
1 anno fa |
Tri Dao
|
88173a1aaf
[FusedDense] Support relu, rename FusedDenseGeluDense -> FusedMLP
|
1 anno fa |
Tri Dao
|
226a1b721d
Implement TensorParallel for FusedDense and FusedDenseGeluDense
|
2 anni fa |
Tri Dao
|
e68ebbe89a
Simplify FusedDense
|
2 anni fa |
Tri Dao
|
13cdceb377
Implement last_layer_subset optimization for BERT
|
2 anni fa |
Tri Dao
|
1feb94265c
[ViT] Use dropout_add_ln for the 1st layer norm
|
2 anni fa |
Tri Dao
|
d4b320b31f
Add MLP, MHA, Block, Embedding modules
|
2 anni fa |