6th Feb 2018 – Data Science News Daily Roundup

Tensorflow 1.6.0-rc, RocksDB 5.10.2, Grafana v5.0, the upcoming release of Spark 2.3, and more in today’s top stories around machine learning, deep learning, and data science news.

1. Tensorflow 1.6.0-rc released

Introducing TensorFlow 1.6 release candidate with some breaking changes and other exciting major features and improvements.

• Prebuilt binaries are now built against CUDA 9.0 and cuDNN 7.
• Prebuilt binaries will now use AVX instructions. (This may break TF on older CPUs.)
• tf.estimator.{FinalExporter,LatestExporter} can now export stripped SavedModels. This improves forward compatibility of the SavedModel.
• FFT support added to XLA CPU/GPU.

To know about Bug Fixes and other changes, visit the GitHub repo.

2. Facebook’s RocksDB 5.10.2 is now released

RocksDB, the high performance embedded database for key-value data built by Facebook, has released its version 5.10.2. The new features include:

• CRC32C is now using the 3-way pipelined SSE algorithm crc32c_3way on supported platforms to improve performance.
• It now provides lifetime hints when writing files on Linux. This reduces hardware write-amp on storage devices supporting multiple streams.
• It now has a DB stat, NUMBER_ITER_SKIP, which returns the number of internal keys skipped during iterations.
• PerfContext counters, key_lock_wait_count and key_lock_wait_time are added,  which measure the number of times transactions wait on key locks and total amount of time waiting.

The complete release and changes are available at the official GitHub repo.

3. Grafana v5.0 is out in Beta

Grafana, the open platform for analytics and monitoring, is now available in version 5.0. The major new features and enhancements include

• New Dashboard Layout Engine with easier drag, drop and resize experience and new types of layouts.
• New UX and improvements in UI in both look and function.
• Dashboard Folders for dashboards organization.
• Permissions on folders and dashboards to help manage larger Grafana installations.
• Datasource provisioning, to setup datasources and dashboards via config files.
• Persistent dashboard url makes it possible to rename dashboards without breaking links.

The entire changes can be read at the official documentation.

4. What is expected from the upcoming Apache Spark 2.3 Release

Apache Spark is soon to release their version 2.3.0 in an upcoming live webinar.

The expected changes include:

• New DataSource APIs for helping developers to easily read and write data for Continuous Processing in Structured Streaming.
• PySpark support for vectorization, giving Python developers the ability to run native Python code fast.
• Improved performance by taking advantage of NVMe SSDs.
• Native Kubernetes support.

5. Ian Goodfellow releases code for SN-GAN and the projection discriminator

Ian Goodfellow, the inventor of GANs, has released the code for SN-GAN and the projection discriminator. Spectral Normalization for GANs is a novel weight normalization technique to stabilize the training of the discriminator of GANs. cGANs with Projection Discriminator is a projection based way to incorporate the conditional information into the discriminator of GANs that respects the role of the conditional information in the underlying probabilistic model. Ian has done the chainer implementation for conditional image generation on ILSVRC2012 dataset (ImageNet) with spectral normalization and projection discriminator. The entire code implementation is available on GitHub.