diff options
| author | Lester Solbakken <lesters@oath.com> | 2022-04-06 15:42:16 +0200 |
|---|---|---|
| committer | Lester Solbakken <lesters@oath.com> | 2022-04-06 15:42:16 +0200 |
| commit | 0040e513d9ebf634130615dde62b2908c4aa9aa1 (patch) | |
| tree | 9d153ffe7204cf2a7d1b744cb94ed962c68c72e2 /model-integration | |
| parent | 7359332a2a806c84c2e504da60b9c5b087e81747 (diff) | |
Move embedder config def to configdefinitions and serialize as reference
Diffstat (limited to 'model-integration')
| -rw-r--r-- | model-integration/CMakeLists.txt | 2 | ||||
| -rw-r--r-- | model-integration/src/main/resources/configdefinitions/embedding.bert-base-embedder.def | 27 |
2 files changed, 0 insertions, 29 deletions
diff --git a/model-integration/CMakeLists.txt b/model-integration/CMakeLists.txt index b564fa29b1c..4225ac38f89 100644 --- a/model-integration/CMakeLists.txt +++ b/model-integration/CMakeLists.txt @@ -1,8 +1,6 @@ # Copyright Yahoo. Licensed under the terms of the Apache 2.0 license. See LICENSE in the project root. install_fat_java_artifact(model-integration) -install_config_definitions() - vespa_install_script(src/main/python/vespa-convert-tf2onnx.py vespa-convert-tf2onnx bin) install(FILES src/main/config/model-integration.xml DESTINATION conf/configserver-app)
\ No newline at end of file diff --git a/model-integration/src/main/resources/configdefinitions/embedding.bert-base-embedder.def b/model-integration/src/main/resources/configdefinitions/embedding.bert-base-embedder.def deleted file mode 100644 index a37599de411..00000000000 --- a/model-integration/src/main/resources/configdefinitions/embedding.bert-base-embedder.def +++ /dev/null @@ -1,27 +0,0 @@ - -namespace=embedding - -# Transformer model settings -transformerModelUrl url default=https://data.vespa.oath.cloud/onnx_models/sentence_all_MiniLM_L6_v2.onnx - -# Max length of token sequence model can handle -transformerMaxTokens int default=384 - -# Pooling strategy -poolingStrategy enum { cls, mean } default=mean - -# Input names -transformerInputIds string default=input_ids -transformerAttentionMask string default=attention_mask -transformerTokenTypeIds string default=token_type_ids - -# Output name -transformerOutput string default=output_0 - -# Settings for ONNX model evaluation -onnxExecutionMode enum { parallel, sequential } default=sequential -onnxInterOpThreads int default=1 -onnxIntraOpThreads int default=-4 # n=number of threads -> n<0: CPUs/(-n), n==0: CPUs, n>0: n - -# Settings for wordpiece tokenizer -tokenizerVocabUrl url default=https://data.vespa.oath.cloud/onnx_models/bert-base-uncased-vocab.txt |