From 1c06036383d1359fece99d770d269509eb14960d Mon Sep 17 00:00:00 2001 From: Bjørn Christian Seime Date: Thu, 11 May 2023 14:06:43 +0200 Subject: Don't specify both package and namespace --- .../embedding/huggingface/HuggingFaceEmbedder.java | 1 + ...embedding.huggingface.hugging-face-embedder.def | 28 ---------------------- .../configdefinitions/hugging-face-embedder.def | 27 +++++++++++++++++++++ 3 files changed, 28 insertions(+), 28 deletions(-) delete mode 100644 model-integration/src/main/resources/configdefinitions/embedding.huggingface.hugging-face-embedder.def create mode 100644 model-integration/src/main/resources/configdefinitions/hugging-face-embedder.def (limited to 'model-integration') diff --git a/model-integration/src/main/java/ai/vespa/embedding/huggingface/HuggingFaceEmbedder.java b/model-integration/src/main/java/ai/vespa/embedding/huggingface/HuggingFaceEmbedder.java index cc13254385b..365a50f47b5 100644 --- a/model-integration/src/main/java/ai/vespa/embedding/huggingface/HuggingFaceEmbedder.java +++ b/model-integration/src/main/java/ai/vespa/embedding/huggingface/HuggingFaceEmbedder.java @@ -5,6 +5,7 @@ import ai.vespa.modelintegration.evaluator.OnnxEvaluatorOptions; import ai.vespa.modelintegration.evaluator.OnnxRuntime; import com.yahoo.component.AbstractComponent; import com.yahoo.component.annotation.Inject; +import com.yahoo.embedding.huggingface.HuggingFaceEmbedderConfig; import com.yahoo.language.process.Embedder; import com.yahoo.tensor.IndexedTensor; import com.yahoo.tensor.Tensor; diff --git a/model-integration/src/main/resources/configdefinitions/embedding.huggingface.hugging-face-embedder.def b/model-integration/src/main/resources/configdefinitions/embedding.huggingface.hugging-face-embedder.def deleted file mode 100644 index 584f23046ba..00000000000 --- a/model-integration/src/main/resources/configdefinitions/embedding.huggingface.hugging-face-embedder.def +++ /dev/null @@ -1,28 +0,0 @@ -package=ai.vespa.embedding.huggingface -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 - -# 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 -transformerGpuDevice int default=-1 diff --git a/model-integration/src/main/resources/configdefinitions/hugging-face-embedder.def b/model-integration/src/main/resources/configdefinitions/hugging-face-embedder.def new file mode 100644 index 00000000000..1dccea0ddf6 --- /dev/null +++ b/model-integration/src/main/resources/configdefinitions/hugging-face-embedder.def @@ -0,0 +1,27 @@ +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 + +# 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 +transformerGpuDevice int default=-1 -- cgit v1.2.3