Create your own GitHub profile
Sign up for your own profile on GitHub, the best place to host code, manage projects, and build software alongside 28 million developers.
Sign up
Pinned repositories
975 contributions in the last year
Contribution activity
November 2018
- Add rfft op
- Add tf prefix to symbols in tf core docs.
- Add integration test with layers/node/converter when we modify `version.ts`
- Optimize pool CPU
- Remove node 8 from travis since 10 is LTS
- Speedup matmul when the matrices are vectors
- support SetDiff1D op
- Support LogSoftmax op
- Add dropout.
- Add TypeScript return type annotation to getCPUBackend
- Fix failing test in travis
- ensure webgl backend read downloads converted typedarray
- Enable lazy unpacking by rearranging data within webGL textures when necessary as part of reshaping
- Add docs for tf.io.browserFiles() and tf.io.browserDownloads()
- add check for window.location.search before using it
- Made tf.abs underflow-safe on CPU & GPU and added test for it.
- Disable uniform uploads when testing on Travis
- Remove unnecessary blob creation in tests
- Fix outdated references to DownloadTrigger class
- Make IS_NODE check more failure-proof
- Support Rank > 1 Tensors for tf.spectral.fft and ifft
- BatchToSpaceND doc fixes
- Add IRFFT ops
- person-segmentation: upgrade parcel, and update demo to use released version
- Fix overlapFactor bug and cleanup a bit
- [speech-command] Add includeEmbeddings; Allow recognize() to draw from WebAudio directly
- [speech-commands] Changes to API and model location
- Improvements to the speech recognition model
- [speech-commands] Load embedding layer only for includeEmbedding=true
- renamed person segmentation model to bodyPix
Created an issue in tensorflow/tfjs that received 6 comments
PoseNet single pose estimation returning NaN on iOS
From a thread on the mailing list: "I am using posenet 2.3 and tjfs 13.2 tfjs: "0.13.2" tfjs-converter: "0.6.4" tfjs-core: "0.13.6" tfjs-layers: "0…
6
comments
- Optimize the tf.rfft / tf.irfft op by calling an rfft kernel.
- tf.softmax() and tf.logSoftmax() should call into a kernel instead of composing high-level ops.
- Use JSON instead of protocol buffers in the TensorFlow saved model converter.
- Once we have a test suite that doesn't run uniforms, update the tf.abs() unit tests to use EPSILON.
1
contribution
in private repositories
Nov 9