experiments/STT-tensorflow
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Merge pull request #42526 from xingyu-long:benchmark-fix-readme

Browse Source 
PiperOrigin-RevId: 328283554
Change-Id: I6eb06cb27648ce2442c7ce5fd4f4720bb211986e
This commit is contained in:
TensorFlower Gardener 2020-08-25 00:48:14 -07:00
commit 5b3cd9ce71
1 changed files with 29 additions and 29 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

58
tensorflow/python/keras/benchmarks/keras_examples_benchmarks/README.md
View File

@ -84,56 +84,56 @@ Metrics for following benchmarks:</br>
#### Cifar10 CNN benchmark
| ----- | Batch_size | Wall_time | Avg_epoch_time | Exp_per_sec |
Distribution_Strategy | | :---: | :--------: | :-------: | :------------:
| :---------: | :-------------------: | | CPU | 256 | 1393.4896 | 3.21 |
15397.69 | `off` | | GPU:2 | 256 | 76.49 | 2.59 | 18758.01 | `mirrored` |
| Batch_size | Wall_time | Avg_epoch_time | Exp_per_sec | Distribution_Strategy
:---: | :--------: | :-------: | :------------: | :---------: | :-------------------:
CPU | 256 | 1393.4896 | 3.21 | 15397.69 | `off`
GPU:2 | 256 | 76.49 | 2.59 | 18758.01 | `mirrored`
#### MNIST Conv benchmark
| ----- | Batch_size | Wall_time | Avg_epoch_time | Exp_per_sec |
Distribution_Strategy | | :---: | :--------: | :-------: | :------------:
| :---------: | :-------------------: | | CPU | 256 | 196.52 | 12.19 | 4915.26 |
`off` | | GPU:2 | 256 | 24.5794 | 1.21 | 47899.32 | `mirrored` |
| Batch_size | Wall_time | Avg_epoch_time | Exp_per_sec | Distribution_Strategy
:---: | :--------: | :-------: | :------------: | :---------: | :-------------------:
CPU | 256 | 196.52 | 12.19 | 4915.26 | `off`
GPU:2 | 256 | 24.5794 | 1.21 | 47899.32 | `mirrored`
#### MNIST Hierarchical RNN (HRNN) benchmark
| ----- | Batch_size | Wall_time | Avg_epoch_time | Exp_per_sec |
Distribution_Strategy | | :---: | :--------: | :-------: | :------------:
| :---------: | :-------------------: | | CPU | 256 | 654.05 | 218.68 | 274.24 |
`off` | | GPU:2 | 256 | 20.77 | 3.73 | 15088.06 | `mirrored` |
| Batch_size | Wall_time | Avg_epoch_time | Exp_per_sec | Distribution_Strategy
:---: | :--------: | :-------: | :------------: | :---------: | :-------------------:
CPU | 256 | 654.05 | 218.68 | 274.24 | `off`
GPU:2 | 256 | 20.77 | 3.73 | 15088.06 | `mirrored`
#### Bidirectional LSTM benchmark
| ----- | Batch_size | Wall_time | Avg_epoch_time | Exp_per_sec |
Distribution_Strategy | | :---: | :--------: | :-------: | :------------:
| :---------: | :-------------------: | | CPU | 512 | 225.57 | 72.55 | 344.70 |
`off` | | GPU:2 | 512 | 23.54 | 3.23 | 7532.53 | `mirrored` |
| Batch_size | Wall_time | Avg_epoch_time | Exp_per_sec | Distribution_Strategy
:---: | :--------: | :-------: | :------------: | :---------: | :-------------------:
CPU | 512 | 225.57 | 72.55 | 344.70 | `off`
GPU:2 | 512 | 23.54 | 3.23 | 7532.53 | `mirrored`
#### Text classification with transformer benchmark
| ----- | Batch_size | Wall_time | Avg_epoch_time | Exp_per_sec |
Distribution_Strategy | | :---: | :--------: | :-------: | :------------:
| :---------: | :-------------------: | | CPU | 512 | 109.22 | 35.93 | 698.10 |
`off` | | GPU:2 | 512 | 9.28 | 0.83 | 26567.54 | `mirrored` |
| Batch_size | Wall_time | Avg_epoch_time | Exp_per_sec | Distribution_Strategy
:---: | :--------: | :-------: | :------------: | :---------: | :-------------------:
CPU | 512 | 109.22 | 35.93 | 698.10 | `off`
GPU:2 | 512 | 9.28 | 0.83 | 26567.54 | `mirrored`
#### MLP benchmark
| ----- | Batch_size | Wall_time | Avg_epoch_time | Exp_per_sec |
Distribution_Strategy | | :---: | :--------: | :-------: | :------------:
| :---------: | :-------------------: | | CPU | 128 | 3.76 | 0.54 | 17678.54 |
`off` | | GPU:2 | 128 | 5.91 | 0.30 | 25435.14 | `mirrored` |
| Batch_size | Wall_time | Avg_epoch_time | Exp_per_sec | Distribution_Strategy
:---: | :--------: | :-------: | :------------: | :---------: | :-------------------:
CPU | 128 | 3.76 | 0.54 | 17678.54 | `off`
GPU:2 | 128 | 5.91 | 0.30 | 25435.14 | `mirrored`
#### Antirectifier benchmark
| ----- | Batch_size | Wall_time | Avg_epoch_time | Exp_per_sec |
Distribution_Strategy | | :---: | :--------: | :-------: | :------------:
| :---------: | :-------------------: | | CPU | 512 | 6.77 | 1.79 | 30916.39 |
`off` | | GPU:2 | 512 | 6.81 | 0.66 | 66563.17 | `mirrored` |
| Batch_size | Wall_time | Avg_epoch_time | Exp_per_sec | Distribution_Strategy
:---: | :--------: | :-------: | :------------: | :---------: | :-------------------:
CPU | 512 | 6.77 | 1.79 | 30916.39 | `off`
GPU:2 | 512 | 6.81 | 0.66 | 66563.17 | `mirrored`
#### IRNN benchmark
----- | Batch_size | Wall_time | Avg_epoch_time | Exp_per_sec | Distribution_Strategy
| Batch_size | Wall_time | Avg_epoch_time | Exp_per_sec | Distribution_Strategy
:---: | :--------: | :-------: | :------------: | :---------: | :-------------------:
CPU | 1024 | 213.00 | 69.01 | 868.08 | `off`
GPU:2 | 1024 | 92.71 | 29.12 | 2042.94 | `mirrored`