vespa
vespa copied to clipboard
vespa-engine
AI + Data, online. https://vespa.ai
- Feature Name: Binomial Coefficient Entropy Coding for Inverted Index Postings - Status: draft - Start Date: 2021-12-13 - Authors: Murage Kibicho # Summary This is a proposal for encoding...
@bratseth Draft PR. For discussion. Ref https://github.com/vespa-engine/vespa/issues/19901. To avoid breaking existing API, this adds a new `MultiFunctionEvaluator` which ensures that ONNX models are only evaluated once for all outputs. Example...
Currently, we don't support expressing query tensors with [hex feed format](https://docs.vespa.ai/en/reference/document-json-format.html#tensor-hex-dump ) (for int8 tensors). It would be great if we could support this format on the query side as...
A model might have several outputs. Current API only supports one output per FunctionEvaluator. ``` FunctionEvaluator evaluator = modelsEvaluator.evaluatorOf("reader", "output_0"); FunctionEvaluator evaluator1 = modelsEvaluator.evaluatorOf("reader", "output_1"); FunctionEvaluator evaluator2 = modelsEvaluator.evaluatorOf("reader", "output_2");...
When using `vespa-fbench`, you need to specifically add the content-type header (`-H "Content-type:application/json"`) when using HTTP POST queries. When you specify post queries on the command line with the `-P`...
The current procedure for using/importing ML models to Vespa is to put them in the application package under the `models` directory (see https://docs.vespa.ai/documentation/reference/application-packages-reference.html). For many applications it would be more...
I confirm that this contribution is made under the terms of the license found in the root directory of this repository's source tree and that I have the authority necessary...
In mixed and sparse tensor attributes we are heap allocating the memory needed for each tensor value. These allocations are managed by _vespamalloc_. When such tensor attribute is downsized (e.g....
Currently if you group on a multi-valued field grouping uses all values. There are use cases where one wants to filter the values and only group over the values which...
E.g BERT base (12 layers, 768 hidden, input max sequence 512) has 110M parameters and the exported ONNX model for a fine-tuned BERT base model for question answering used in...