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

[XLA] Fix rendering of the RngBitGenerator description table

Browse Source 
PiperOrigin-RevId: 312123981
Change-Id: I9d1ecdf88dfb9f5689dcfc26f6243a192ab55dd6
This commit is contained in:
George Karpenkov 2020-05-18 11:50:56 -07:00 committed by TensorFlower Gardener
parent ef45324fc6
commit 6f19d507f4
1 changed files with 19 additions and 13 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

32
tensorflow/compiler/xla/g3doc/operation_semantics.md
View File

@ -2299,20 +2299,26 @@ The output is guaranteed to be a deterministic function of the initial state but
it is *not* guaranteed to be deterministic between backends and different it is *not* guaranteed to be deterministic between backends and different
compiler versions. compiler versions.
<b>`RngBitGenerator(algorithm, key, shape)`</b> | Arguments | Type | Semantics | <b>`RngBitGenerator(algorithm, key, shape)`</b>
|---------------- | ----------------- | ------------------------------------- |
| `algorithm` | `RandomAlgorithm` | PRNG algorithm to be used. | |
`initial_state` | `XlaOp` | Initial state for the PRNG algorithm. | | `shape` |
`Shape` | Output shape for generated data. |
Available values for `algorithm`: * `rng_default`: Backend specific algorithm Arguments | Type | Semantics
with backend specific shape requirements. * `rng_three_fry`: ThreeFry --------------- | ----------------- | -------------------------------------
counter-based PRNG algorithm. The `initial_state` shape is `u64[2]` with `algorithm` | `RandomAlgorithm` | PRNG algorithm to be used.
arbitrary values. `initial_state` | `XlaOp` | Initial state for the PRNG algorithm.
[Salmon et al. SC 2011. Parallel random numbers: as easy as 1, 2, 3.](http://www.thesalmons.org/john/random123/papers/random123sc11.pdf) `shape` | `Shape` | Output shape for generated data.
* `rng_philox`: Philox algorithm to generate random numbers in parallel. The
`initial_state` shape is `u64[3]` with arbitrary values. Available values for `algorithm`:
[Salmon et al. SC 2011. Parallel random numbers: as easy as 1, 2, 3.](http://www.thesalmons.org/john/random123/papers/random123sc11.pdf)
- `rng_default`: Backend specific algorithm with backend specific shape
requirements.
- `rng_three_fry`: ThreeFry counter-based PRNG algorithm. The `initial_state`
shape is `u64[2]` with arbitrary values.
[Salmon et al. SC 2011. Parallel random numbers: as easy as 1, 2, 3.](http://www.thesalmons.org/john/random123/papers/random123sc11.pdf)
- `rng_philox`: Philox algorithm to generate random numbers in parallel. The
`initial_state` shape is `u64[3]` with arbitrary values.
[Salmon et al. SC 2011. Parallel random numbers: as easy as 1, 2, 3.](http://www.thesalmons.org/john/random123/papers/random123sc11.pdf)
## Scatter ## Scatter