- 22 Oct, 2020 1 commit
-
-
Ran Chen authored
This fix tensorflow_models/official/nlp/bert/run_pretraining failure with explicit allreduce. PiperOrigin-RevId: 338521478
-
- 21 Oct, 2020 4 commits
-
-
A. Unique TensorFlower authored
PiperOrigin-RevId: 338364871
-
Yin Cui authored
PiperOrigin-RevId: 338348738
-
Hongkun Yu authored
PiperOrigin-RevId: 338309524
-
Chen Chen authored
PiperOrigin-RevId: 338249187
-
- 20 Oct, 2020 2 commits
-
-
Chen Chen authored
PiperOrigin-RevId: 338095907
-
Abdullah Rashwan authored
PiperOrigin-RevId: 338094579
-
- 19 Oct, 2020 3 commits
-
-
A. Unique TensorFlower authored
PiperOrigin-RevId: 337931726
-
Hongkun Yu authored
PiperOrigin-RevId: 337927475
-
Abdullah Rashwan authored
PiperOrigin-RevId: 337910656
-
- 17 Oct, 2020 2 commits
-
-
Reed Wanderman-Milne authored
Soon, AutoCastVariable.dtype will refer to the variable dtype, not the compute dtype. This change stops using AutoCastVariable.dtype. PiperOrigin-RevId: 337670544
-
Ruoxin Sang authored
PiperOrigin-RevId: 337609198
-
- 16 Oct, 2020 3 commits
-
-
Abdullah Rashwan authored
PiperOrigin-RevId: 337550702
-
Le Hou authored
PiperOrigin-RevId: 337540237
-
Abdullah Rashwan authored
PiperOrigin-RevId: 337456452
-
- 15 Oct, 2020 3 commits
-
-
Allen Wang authored
PiperOrigin-RevId: 337366497
-
A. Unique TensorFlower authored
PiperOrigin-RevId: 337251846
-
Hongkun Yu authored
PiperOrigin-RevId: 337242735
-
- 14 Oct, 2020 2 commits
-
-
Allen Wang authored
PiperOrigin-RevId: 337134894
-
Hongkun Yu authored
PiperOrigin-RevId: 336991213
-
- 13 Oct, 2020 10 commits
-
-
A. Unique TensorFlower authored
PiperOrigin-RevId: 336973004
-
A. Unique TensorFlower authored
PiperOrigin-RevId: 336967249
-
Hongkun Yu authored
PiperOrigin-RevId: 336960641
-
Abdullah Rashwan authored
PiperOrigin-RevId: 336951637
-
A. Unique TensorFlower authored
PiperOrigin-RevId: 336942308
-
Abdullah Rashwan authored
PiperOrigin-RevId: 336939703
-
A. Unique TensorFlower authored
Lambda layers are fundamentally non-portable since they serialize Python bytecode. PiperOrigin-RevId: 336917172
-
A. Unique TensorFlower authored
PiperOrigin-RevId: 336914823
-
A. Unique TensorFlower authored
Lambda layers are fundamentally non-portable since they serialize Python bytecode. PiperOrigin-RevId: 336910013
-
Hongkun Yu authored
PiperOrigin-RevId: 336795303
-
- 12 Oct, 2020 4 commits
-
-
A. Unique TensorFlower authored
PiperOrigin-RevId: 336771016
-
A. Unique TensorFlower authored
PiperOrigin-RevId: 336717785
-
Zongwei Zhou authored
PiperOrigin-RevId: 336715941
-
Allen Wang authored
PiperOrigin-RevId: 336709640
-
- 09 Oct, 2020 4 commits
-
-
Zhenyu Tan authored
PiperOrigin-RevId: 336353171
-
Hongkun Yu authored
PiperOrigin-RevId: 336339472
-
Abdullah Rashwan authored
PiperOrigin-RevId: 336229644
-
Ran Chen authored
PiperOrigin-RevId: 336215069
-
- 08 Oct, 2020 2 commits
-
-
Abdullah Rashwan authored
PiperOrigin-RevId: 336132820
-
Hongkun Yu authored
Factor out optimizer creation logic from the trainer, so as to avoid access runtime config inside Trainer. Add validation method on the ExperimentConfig as clients may use other containers instead of the model garden config dataclass. PiperOrigin-RevId: 336132223
-