blob: a6544187140fe9b50756847390b312470d6f80c7 (
plain) (
blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
|
# Copy of this Vespa config stored here because Vespa config definitions are not
# available in unit tests, and are needed (by DomConfigPayloadBuilder.parseLeaf)
# Alternativ ely, we could make that not need it as it is not strictly necessaery.
namespace=embedding
# Wordpiece tokenizer
tokenizerVocab model
transformerModel model
# 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
|