aboutsummaryrefslogtreecommitdiffstats
path: root/vespajlib/src/main/java/com/yahoo/tensor/IndexedTensor.java
Commit message (Collapse)AuthorAgeFilesLines
* Update 2017 copyright notices.gjoranv2021-10-071-1/+1
|
* Encode to dense tensorJon Bratseth2021-09-161-3/+1
|
* Revert "Revert "Add ONNX-RT evaluator to model-integration module""Lester Solbakken2021-05-191-0/+8
| | | | This reverts commit 97080252fac0ba45b58f9d0efb56603da518428f.
* Revert "Add ONNX-RT evaluator to model-integration module"Lester Solbakken2021-05-191-8/+0
|
* Add ONNX-RT evaluator to model-integration moduleLester Solbakken2021-05-191-0/+8
|
* Return 0.0 rather than nan for non-existing tensor cells in JavaLester Solbakken2021-05-031-2/+2
|
* add api for detecting cell existenceArne Juul2021-04-301-0/+11
| | | | | | | | * new API "has(TensorAddress)" detects if a Tensor has a cell with the given address. * use new API in join and merge. This will give different results for cells that are present but contain NaN versus cells that aren't present at all. * use new API in slice. This gives a different default (0, not NaN) when trying to access cells that aren't present.
* throw exception for unexpected enum value in more placesArne Juul2021-04-091-5/+4
|
* Resolve feedback from PR reviewLester Solbakken2021-04-081-42/+32
|
* Add bfloat16 and int8 tensor cell types in JavaLester Solbakken2021-04-081-0/+16
|
* Use the tensor type to switch tensor binary formatJon Bratseth2020-09-031-3/+1
| | | | | | | | | | The binary format of a tensor should depend on the tensor type, not the implementation type as the API permits the user choosing that (and it may not be 1-1 anyway). This makes this change for sparse tensors using the mixed implementation type but not dense tensors using the mixed implementation type as that would be more work given the unfinished state of the mixed implementation.
* Merge pull request #11637 from vespa-engine/bratseth/tensor-short-form-tostringJon Bratseth2020-01-061-9/+14
|\ | | | | More tensor short forms in Tensor.toString()
| * More tensor short forms in Tensor.toString()Jon Bratseth2020-01-031-9/+14
| |
* | Add mergeJon Bratseth2020-01-021-5/+0
|/
* Interpret dimensions in written orderJon Bratseth2019-12-141-26/+69
|
* Tensor parsing improvementsJon Bratseth2019-12-101-17/+57
| | | | | - Mixed tensor format parsing (outside expressions) - Validate structure of dense tensor strings
* Always serialize dynamic tensors in verbose formJon Bratseth2019-11-081-1/+10
|
* Literal tensors with value expressionsJon Bratseth2019-11-051-2/+0
|
* Document the standard value order one placeJon Bratseth2019-10-011-20/+19
|
* Convenience method to find smallest/largest cellsJon Bratseth2019-07-031-0/+5
|
* Keep writing floats as doubles in toString for nowJon Bratseth2019-06-201-1/+1
|
* Revert "Merge pull request #9849 from ↵Jon Bratseth2019-06-201-1/+51
| | | | | | | vespa-engine/revert-9836-bratseth/output-in-dense-tensor-form" This reverts commit 2663557e0c6d652fd082507878059a69c75e2415, reversing changes made to 46cfd5c26c5dac150efb698d726810ade4364e70.
* Read dense tensor form in documentsJon Bratseth2019-06-181-0/+2
|
* Dense string formJon Bratseth2019-06-071-4/+3
|
* Expect IllegalArgumentException and be consistentJon Bratseth2019-06-041-7/+7
|
* Adjust to changes exception typeJon Bratseth2019-06-041-1/+1
|
* Validate moreJon Bratseth2019-06-041-5/+4
|
* Update ABI spec and add javadocWJon Bratseth2019-06-011-4/+38
|
* Allow passing your own vector without copy to the IndexedTensor.Henning Baldersheim2019-06-011-10/+67
|
* Add a test and correct conditionJon Bratseth2019-04-291-1/+1
|
* Add float accessorJon Bratseth2019-04-261-0/+10
|
* Encode directly as floatJon Bratseth2019-04-261-1/+10
|
* Build tensors purely with floatsJon Bratseth2019-04-261-0/+13
|
* Make float builder when appropriateJon Bratseth2019-04-261-3/+2
|
* Move bound builder double array into double subclassJon Bratseth2019-04-261-57/+22
|
* Split values into IndexedDoubleTensor subclassJon Bratseth2019-04-251-34/+17
|
* TensorRemoveUpdate support for mixed tensorsLester Solbakken2019-02-201-0/+5
|
* TensorAddUpdate support for mixed tensorsLester Solbakken2019-02-201-0/+6
|
* Add faster tensor rename if dimension after rename are in the same orderLester Solbakken2018-10-081-0/+9
|
* Add reduce-join optimization in JavaLester Solbakken2018-09-281-1/+8
|
* Tensor API out of betaJon Bratseth2018-03-121-2/+0
|
* Use longs for dimensions lengths in all API'sJon Bratseth2017-12-191-96/+98
| | | | | This is to be able to support tensor dimensions with more than 2B elements in the future without API change.
* Revert "Revert "Bratseth/tensorflow models""Jon Bratseth2017-12-181-82/+82
|
* Revert "Bratseth/tensorflow models"gjoranv2017-12-171-82/+82
|
* Check agreement between TF and Vespa executionJon Bratseth2017-12-131-82/+82
|
* Add bounds checking for indexed tensorLester Solbakken2017-11-091-2/+10
|
* Update copyright headersJon Bratseth2017-06-141-1/+1
|
* Revert "Update copyright headers"Jon Bratseth2017-06-141-1/+1
|
* Update copyright headersJon Bratseth2017-06-141-1/+1
|
* Revert "Copyright header"Jon Bratseth2017-06-131-1/+1
|