diff options
Diffstat (limited to 'config-model/src/test/cfg/application/embed/services.xml')
| -rw-r--r-- | config-model/src/test/cfg/application/embed/services.xml | 26 |
1 files changed, 18 insertions, 8 deletions
diff --git a/config-model/src/test/cfg/application/embed/services.xml b/config-model/src/test/cfg/application/embed/services.xml index 99c89bc4324..6823ef900ae 100644 --- a/config-model/src/test/cfg/application/embed/services.xml +++ b/config-model/src/test/cfg/application/embed/services.xml @@ -16,6 +16,7 @@ <onnx-intraop-threads>10</onnx-intraop-threads> <onnx-interop-threads>8</onnx-interop-threads> <onnx-gpu-device>1</onnx-gpu-device> + <pooling-strategy>mean</pooling-strategy> </component> <component id="hf-tokenizer" type="hugging-face-tokenizer"> @@ -25,15 +26,24 @@ <truncation>true</truncation> </component> - <component id="transformer" class="ai.vespa.embedding.BertBaseEmbedder" bundle="model-integration"> - <config name="embedding.bert-base-embedder"> - <!-- model specifics --> - <transformerModel model-id="minilm-l6-v2" url="application-url"/> - <tokenizerVocab path="files/vocab.txt"/> + <component id="bert-embedder" type="bert-embedder"> + <!-- model specifics --> + <transformer-model model-id="minilm-l6-v2" url="application-url"/> + <tokenizer-vocab path="files/vocab.txt"/> + <max-tokens>512</max-tokens> + <transformer-input-ids>my_input_ids</transformer-input-ids> + <transformer-attention-mask>my_attention_mask</transformer-attention-mask> + <transformer-token-type-ids>my_token_type_ids</transformer-token-type-ids> + <transformer-output>my_output</transformer-output> + <transformer-start-sequence-token>101</transformer-start-sequence-token> + <transformer-end-sequence-token>102</transformer-end-sequence-token> - <!-- tunable parameters: number of threads etc --> - <onnxIntraOpThreads>4</onnxIntraOpThreads> - </config> + + <!-- tunable parameters: number of threads etc --> + <onnx-execution-mode>parallel</onnx-execution-mode> + <onnx-intraop-threads>4</onnx-intraop-threads> + <onnx-interop-threads>8</onnx-interop-threads> + <onnx-gpu-device>1</onnx-gpu-device> </component> <nodes> |