Commit message (Collapse) | Author | Age | Files | Lines | |
---|---|---|---|---|---|
* | Better naming | Henning Baldersheim | 2023-09-04 | 1 | -9/+9 |
| | |||||
* | Probe with optimize as in production. | Henning Baldersheim | 2023-09-04 | 1 | -5/+3 |
| | |||||
* | Report memory usage back when probing types. | Henning Baldersheim | 2023-09-04 | 1 | -13/+48 |
| | |||||
* | experiment with simple instruction-level profiling | Håvard Pettersen | 2022-09-01 | 1 | -12/+30 |
| | |||||
* | use stdout for 'error' result | Håvard Pettersen | 2022-03-07 | 1 | -2/+2 |
| | |||||
* | better error messages | Håvard Pettersen | 2022-02-15 | 1 | -7/+55 |
| | | | | | | | | | | | | | added support for capturing stderr when you expect stuff to fail in order to capture all error messages from the child process. simplify ServerCmd to always be verbose (print stdin/stdout(stderr) interactions). improve ServerCmd to enable explicitly checking the child process exit code. fixed test dependency on binary for eval expr and analyze model tests. | ||||
* | enable probing output types | Håvard Pettersen | 2022-02-11 | 2 | -2/+45 |
| | | | | also added some testing | ||||
* | run onnx model to detect unknown output sizes | Håvard Pettersen | 2022-02-09 | 1 | -0/+1 |
| | |||||
* | Update Verizon Media copyright notices. | gjoranv | 2021-10-07 | 2 | -2/+2 |
| | |||||
* | Update 2017 copyright notices. | gjoranv | 2021-10-07 | 8 | -8/+8 |
| | |||||
* | remove hamming test where Java eval result was ignored | Håvard Pettersen | 2021-09-22 | 1 | -4/+6 |
| | | | | | | | Java tensor conformance evaluation failures will break global invariants by dropping tests in the output. We need to isolate test evaluation better before we are able to ignore expressions that will not even parse. | ||||
* | added new 'hamming' function | Håvard Pettersen | 2021-09-22 | 3 | -5/+26 |
| | |||||
* | Revert "added new 'hamming' function" | Harald Musum | 2021-09-22 | 3 | -26/+5 |
| | |||||
* | added new 'hamming' function | Håvard Pettersen | 2021-09-22 | 3 | -5/+26 |
| | |||||
* | added 'undef' command for completeness | Håvard Pettersen | 2021-09-07 | 1 | -13/+36 |
| | |||||
* | added help and list commands to interactive mode | Håvard Pettersen | 2021-09-02 | 1 | -0/+23 |
| | |||||
* | use common history event | Håvard Pettersen | 2021-09-02 | 1 | -6/+4 |
| | |||||
* | use editline | Håvard Pettersen | 2021-09-02 | 2 | -9/+52 |
| | |||||
* | minor fixes after review | Håvard Pettersen | 2021-08-31 | 1 | -4/+5 |
| | |||||
* | add json-repl (and interactive) modes to vespa-eval-expr | Håvard Pettersen | 2021-08-30 | 1 | -52/+248 |
| | |||||
* | improve vespa-eval-expr | Håvard Pettersen | 2021-08-23 | 1 | -18/+119 |
| | | | | | | Allow multiple expressions to be evaluated. Later expressions may use results from earlier expressions. Also add option to print information about low-level optimizations (how the expressions will be executed). | ||||
* | dry run onnx models on setup | Håvard Pettersen | 2021-06-23 | 1 | -1/+10 |
| | |||||
* | enable conformance testing of bit extraction | Håvard Pettersen | 2021-06-21 | 1 | -6/+4 |
| | |||||
* | add 'bit(a,b)' math function | Håvard Pettersen | 2021-06-18 | 1 | -0/+30 |
| | |||||
* | estimate time and space needed for onnx model evaluation | Håvard Pettersen | 2021-06-14 | 1 | -3/+152 |
| | |||||
* | added very simple onnx model analyzer | Håvard Pettersen | 2021-06-10 | 3 | -0/+68 |
| | | | | just estimating memory needed to load the model for now | ||||
* | un-ignore no longer failing tests | Håvard Pettersen | 2021-05-03 | 1 | -9/+9 |
| | |||||
* | ignore some conformance tests for Java | Håvard Pettersen | 2021-04-30 | 3 | -14/+38 |
| | | | | | peek out-of-bounds nan vs non-existing cells in merge/join | ||||
* | allow ignoring test failures | Håvard Pettersen | 2021-04-29 | 1 | -8/+17 |
| | | | | | + avoid calculating expected result twice + added fail_cnt in test summary | ||||
* | add more conformance testing | Håvard Pettersen | 2021-04-29 | 1 | -2/+28 |
| | |||||
* | stay within defined values for int8 | Håvard Pettersen | 2021-04-27 | 1 | -1/+1 |
| | |||||
* | also test with 0 | Håvard Pettersen | 2021-04-26 | 1 | -1/+1 |
| | | | | and make numbers a bit 'harder' | ||||
* | add more testing of erf function | Håvard Pettersen | 2021-04-23 | 1 | -0/+9 |
| | |||||
* | more conformance testing | Håvard Pettersen | 2021-04-23 | 1 | -0/+56 |
| | |||||
* | print diff of mismatching tensors | Håvard Pettersen | 2021-04-20 | 2 | -51/+64 |
| | | | | also stop using vespalib testkit | ||||
* | remove deprecated tensor model | Håvard Pettersen | 2021-04-16 | 1 | -1/+0 |
| | | | | fix and extend tests to also verify unstable cell types | ||||
* | Merge pull request #17452 from ↵ | Håvard Pettersen | 2021-04-16 | 2 | -111/+243 |
|\ | | | | | | | | | vespa-engine/havardpe/extend-tensor-conformance-test Havardpe/extend tensor conformance test | ||||
| * | extend tensor conformance test | Håvard Pettersen | 2021-04-15 | 2 | -111/+243 |
| | | |||||
* | | Avoid copying map elements when iterating over map. | Tor Egge | 2021-04-14 | 1 | -1/+1 |
|/ | |||||
* | use GenSpec to generate more tests for map/reduce/join | Håvard Pettersen | 2021-04-14 | 3 | -79/+114 |
| | | | | | | - limit the number of cases run in unit test - make pow tests use very limited input values to avoid rounding issues. | ||||
* | add pending stuff (erf/median) | Håvard Pettersen | 2021-04-14 | 1 | -4/+2 |
| | |||||
* | always use reference eval as expected result | Håvard Pettersen | 2021-04-14 | 3 | -120/+72 |
| | |||||
* | remove compare mode | Håvard Pettersen | 2021-04-14 | 1 | -95/+0 |
| | |||||
* | Fix C++ serialization test for new cell types | Lester Solbakken | 2021-04-08 | 1 | -10/+21 |
| | |||||
* | Add bfloat16 and int8 tensor cell types in Java | Lester Solbakken | 2021-04-08 | 1 | -1/+21 |
| | |||||
* | cell meta with both compile-time and run-time resolving | Håvard Pettersen | 2021-03-09 | 1 | -6/+6 |
| | | | | wired into generic join and reduce as examples of usage | ||||
* | all scalars must be double | Håvard Pettersen | 2021-03-04 | 1 | -1/+1 |
| | | | | and all operation results must be at least float | ||||
* | use size literals in eval | Arne Juul | 2021-02-15 | 1 | -9/+10 |
| | |||||
* | added 'display' mode to tensor conformance binary | Håvard Pettersen | 2021-02-04 | 1 | -0/+35 |
| | |||||
* | use GenSpec in tensor_model.cpp | Håvard Pettersen | 2021-02-04 | 1 | -1/+1 |
| |