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 40 million developers.
Sign up-
Facebook
- Menlo Park, California
- https://code.facebook.com
Popular repositories
facebook-github-bot doesn’t have any public repositories yet.
6,407 contributions in the last year
Contribution activity
January 2020
- [ONNX] Sort export w/ negative axes
- [ONNX] Added torchvision tests as part of ORT tests
- Add cummax operator
- Javadoc changes
- [DO NOT REVIEW] TESTING
- docker hub image info
- Add op bitwise_or
- remove THConv code which never be used
- [jit]C++ API parity: isinf
- Move log_normal to Aten(CPU)
- Move geometric to Aten(CPU)
- Move cauchy to Aten(CPU)
- Move lshift to Aten
- Move log_sigmoid to Aten(CPU)
- Port softplus activation to Aten(CPU+CUDA)
- TensorIterator norm update
- Avoid reference invalidation in cuda SpectralOps' plan_caches
- Add device debug info to CUDA build
- [CI] Run a non-quiet submodule update to prevent timeouts on Circle CI
- Upgrade to pybind11 2.4.3
- Fix and add more padding mode support for Conv
- [ROCm] Support PyTorch ROCm CI on Ubuntu18.04
- Add stub for transformer.py and MultiheadAttention Class.
- Use TORCH_CHECK instead of AT_ASSERT in torch::cuda::gather()
- Raise ValueError if CUDA device is specified without specifying the :
- Some pull request reviews not shown.
- Remove inaccurate code comment
- Update file indexer to take timestamp into consideration
- add user nebula
- Stress Test: relax prefix iterator check condition
- Print out non-ok DB::Open status in db_stress
- Skip CancelAllBackgroundWork if DBImpl is already closed
- JMH microbenchmarks for RocksJava
- Implement getfreespace for WinEnv
- Checksum for each SST file and stores in MANIFEST
- Fix test in LITE mode
- parallel occ
- Fix error message
- Add oldest snapshot sequence property
- Fix a data race for cfd->log_number_
- Provide an allocator for new memory type to be used with RocksDB block cache
- Improve error msg for SstFileWriter Merge
- Add range delete function to C-API
- Increase max_log_size in FlushJob to 1024 bytes
- Exclude MergeInProgress from errors in stress tests
- Print before AddErrors in stress tests
- Prevent an incompatible combination of options
- Pass more device data to backend's post lowering graph transformation
- 32b dim_t fixes
- [Graph] Add Flip node.
- [GraphOptz] Add numeric equivalence test for cseRespectsPredicates
- [PyTorch] Remove flaky redundent unit test
- [OpenCL] add an option to cache programs on disk
- [OperatorTest] Add Int8 and Int32 bias versions of RWQ-FC with symmetric quantization
- Updated tensor.assign to only copy unpadded size in the case of an unpadded tensor. (Quick Fix for Repro)
- [Base] Support padding in Tensor::isEqual()
- [easy] Rename GroupwiseQuantizedConvolution OperatorTest
- Update NNPI Backend to v0.4.2.5
- Changed the default tracing level in image-classifier to NONE
- Add flag for enabling extra offset in EmbeddingBag ops
- Fix for Function::clone() not cloning Predicates
- Fix StringRef issue in ExecutionEngine
- [Support] Add LOG_CUSTOM_LOC for specifying custom file/line
- Make GraphOptzTest tests use optimizeFunction helper
- Fix for windows build.
- [torch_glow] Minor change to allow backend selection for pytests
- [GraphOptz] Add helper to check whether a Node is in a Function
6
repositories not shown