diff options
| author | Bjørn Christian Seime <bjorncs@yahooinc.com> | 2023-06-02 12:10:32 +0200 |
|---|---|---|
| committer | Bjørn Christian Seime <bjorncs@yahooinc.com> | 2023-06-02 12:10:32 +0200 |
| commit | a67788f2b7786a2cfcb9244d1e72a7fb1815425b (patch) | |
| tree | fa34be2f0f13ef4ea116dd12853c734de3bc2eca /model-integration | |
| parent | e757e5ff2e6dadbe31389c7dfeb3f52827a1668b (diff) | |
Introduce services.xml syntax for configuring HuggingFace embedders
Diffstat (limited to 'model-integration')
| -rw-r--r-- | model-integration/pom.xml | 6 | ||||
| -rw-r--r-- | model-integration/src/main/resources/configdefinitions/hugging-face-embedder.def | 29 |
2 files changed, 6 insertions, 29 deletions
diff --git a/model-integration/pom.xml b/model-integration/pom.xml index 681003fdc89..519aebe6f79 100644 --- a/model-integration/pom.xml +++ b/model-integration/pom.xml @@ -81,6 +81,12 @@ <scope>provided</scope> </dependency> <dependency> + <groupId>com.yahoo.vespa</groupId> + <artifactId>configdefinitions</artifactId> + <version>${project.version}</version> + <scope>provided</scope> + </dependency> + <dependency> <groupId>net.java.dev.jna</groupId> <artifactId>jna</artifactId> <scope>provided</scope> diff --git a/model-integration/src/main/resources/configdefinitions/hugging-face-embedder.def b/model-integration/src/main/resources/configdefinitions/hugging-face-embedder.def deleted file mode 100644 index 36957004e02..00000000000 --- a/model-integration/src/main/resources/configdefinitions/hugging-face-embedder.def +++ /dev/null @@ -1,29 +0,0 @@ -namespace=embedding.huggingface - -# Path to tokenizer.json -tokenizerPath model - -# Path to model.onnx -transformerModel model - -# Max length of token sequence model can handle -transformerMaxTokens int default=512 - -# Input names -transformerInputIds string default=input_ids -transformerAttentionMask string default=attention_mask -transformerTokenTypeIds string default=token_type_ids - -# Output name -transformerOutput string default=last_hidden_state - - -# Normalize tensors from tokenizer -normalize bool default=false - -# Settings for ONNX model evaluation -transformerExecutionMode enum { parallel, sequential } default=sequential -transformerInterOpThreads int default=1 -transformerIntraOpThreads int default=-4 -# GPU device id, -1 for CPU -transformerGpuDevice int default=0 |