summaryrefslogtreecommitdiffstats
path: root/configdefinitions
diff options
context:
space:
mode:
authorLester Solbakken <lesters@users.noreply.github.com>2022-04-19 11:35:02 +0200
committerGitHub <noreply@github.com>2022-04-19 11:35:02 +0200
commita0cf7298d39421400ee83dc4e6231465f115659a (patch)
tree4825eced6a81f83cb325e332cc51da700e713af6 /configdefinitions
parent814e8a415dee0ce0735a988932db80b20301b897 (diff)
parent0040e513d9ebf634130615dde62b2908c4aa9aa1 (diff)
Merge pull request #21982 from vespa-engine/lesters/default-url-config
Accept default values for URL types in config
Diffstat (limited to 'configdefinitions')
-rw-r--r--configdefinitions/src/main/java/com/yahoo/embedding/package-info.java5
-rw-r--r--configdefinitions/src/vespa/CMakeLists.txt1
-rw-r--r--configdefinitions/src/vespa/embedding.bert-base-embedder.def27
3 files changed, 33 insertions, 0 deletions
diff --git a/configdefinitions/src/main/java/com/yahoo/embedding/package-info.java b/configdefinitions/src/main/java/com/yahoo/embedding/package-info.java
new file mode 100644
index 00000000000..1842b80aba9
--- /dev/null
+++ b/configdefinitions/src/main/java/com/yahoo/embedding/package-info.java
@@ -0,0 +1,5 @@
+// Copyright Yahoo. Licensed under the terms of the Apache 2.0 license. See LICENSE in the project root.
+@ExportPackage
+package com.yahoo.embedding;
+
+import com.yahoo.osgi.annotation.ExportPackage;
diff --git a/configdefinitions/src/vespa/CMakeLists.txt b/configdefinitions/src/vespa/CMakeLists.txt
index 3735a8ae57a..33ceb5338d3 100644
--- a/configdefinitions/src/vespa/CMakeLists.txt
+++ b/configdefinitions/src/vespa/CMakeLists.txt
@@ -74,3 +74,4 @@ vespa_generate_config(configdefinitions all-clusters-bucket-spaces.def)
install_config_definition(all-clusters-bucket-spaces.def vespa.config.content.all-clusters-bucket-spaces.def)
vespa_generate_config(configdefinitions stateserver.def)
install_config_definition(stateserver.def vespa.config.core.stateserver.def)
+install_config_definition(embedding.bert-base-embedder.def embedding.bert-base-embedder.def)
diff --git a/configdefinitions/src/vespa/embedding.bert-base-embedder.def b/configdefinitions/src/vespa/embedding.bert-base-embedder.def
new file mode 100644
index 00000000000..a37599de411
--- /dev/null
+++ b/configdefinitions/src/vespa/embedding.bert-base-embedder.def
@@ -0,0 +1,27 @@
+
+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