| Commit message (Collapse) | Author | Age | Files | Lines | |
|---|---|---|---|---|---|
| * | Update copyright | Jon Bratseth | 2023-10-09 | 1 | -1/+1 |
| | | |||||
| * | Add config options + license | Jo Kristian Bergum | 2023-09-21 | 1 | -0/+19 |
| | | |||||
| * | Actually differentiate missing element from empty value | Bjørn Christian Seime | 2023-06-15 | 1 | -1/+1 |
| | | |||||
| * | Prefer truncation configuration from tokenizer model | Bjørn Christian Seime | 2023-06-12 | 1 | -3/+0 |
| | | | | | | | | Only override truncation if not specified or max length exceeds max tokens accepted by model. Use JNI wrapper directly to determine existing truncation configuration (JSON format is not really documented). Simply configuration for pure tokenizer embedder. Disable DJL usage telemetry. | ||||
| * | Remove/rewrite unit tests with old mechanism for configuring Bert embedder | Bjørn Christian Seime | 2023-06-05 | 1 | -11/+0 |
| | | |||||
| * | Make pooling strategy configurable for Huggingface embedder | Bjørn Christian Seime | 2023-06-05 | 1 | -0/+1 |
| | | |||||
| * | Test that Bert embedder is configured | Bjørn Christian Seime | 2023-06-05 | 1 | -0/+20 |
| | | |||||
| * | Introduce services.xml syntax for configuring HuggingFace embedders | Bjørn Christian Seime | 2023-06-02 | 1 | -0/+21 |
| | | |||||
| * | always use "go" version of vespa-start-container-daemon | Arne Juul | 2022-11-30 | 1 | -1/+1 |
| | | |||||
| * | Add 'model' config type | Jon Bratseth | 2022-08-31 | 1 | -1/+1 |
| | | |||||
| * | Drop special embedder config | Jon Bratseth | 2022-08-24 | 1 | -7/+9 |
| | | |||||
| * | More tests | Jon Bratseth | 2022-08-24 | 1 | -6/+1 |
| | | |||||
| * | Test model id lookup | Jon Bratseth | 2022-08-24 | 1 | -1/+1 |
| | | |||||
| * | Simplify and generalize | Jon Bratseth | 2022-08-23 | 1 | -2/+2 |
| | | | | | | | | | | | | | | | | Let embedder rewrites simply always change <x path="y" url="z"> to <xPath>y</xPath> <xUrl>z</xUrl>, as well as translating the id attirebute to the corresponding path if on hosted. This means that the current built-in embedder accepting "vocab" and "model" is broken as these names are different from the names in the config model, but this isn't documented yet so should be okay. The effect of this is that our built-in models can be used in any embedder and config, and the embedder syntax can be used in application specific embedders. | ||||
| * | Add services.xml syntax for embedders | Lester Solbakken | 2022-05-23 | 1 | -0/+26 |
 |
index : vespa | |
| An engine for low-latency computation over large data sets |
| aboutsummaryrefslogtreecommitdiffstats |