https://github.com/spirius/aws-sagemaker-tensorflow
AWS Sagemaker image with FMA disable tensorflow
https://github.com/spirius/aws-sagemaker-tensorflow
Last synced: 4 months ago
JSON representation
AWS Sagemaker image with FMA disable tensorflow
- Host: GitHub
- URL: https://github.com/spirius/aws-sagemaker-tensorflow
- Owner: spirius
- Created: 2019-03-25T23:16:38.000Z (about 7 years ago)
- Default Branch: master
- Last Pushed: 2019-03-25T23:47:56.000Z (about 7 years ago)
- Last Synced: 2025-10-19T00:06:04.335Z (8 months ago)
- Language: Dockerfile
- Size: 1000 Bytes
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# AWS Sagemaker Tensorflow Container - without FMA
Default tensorflow distribution is compiled with FMA support, but FMA
instructions are not supported on VirtualBox. Therfore it is impossible to running sagemaker tensorflow containers when using docker-machine.
Following docker image builds tansorflow with `-march=-no-fma` flag and
reinstalls tensorflow in AWS provided container.
| Official Image | No FMA image |
|----------------|--------------|
| `sagemaker-tensorflow-scriptmode:1.12.0-cpu-py3` | `spirius/sagemaker-tensorflow-scriptmode:1.12.0-cpu-nofma-py3` |
### Example
```python
import sagemaker
from sagemaker.tensorflow import TensorFlow
estimator = TensorFlow(
entry_point='train.py',
source_dir='code',
model_dir='model',
train_instance_type='local',
train_instance_count=1,
role=sagemaker.get_execution_role(),
script_mode=True,
image_name='spirius/sagemaker-tensorflow-scriptmode:1.12.0-cpu-nofma-py3',
)
```