Is is possible to pre-load the model for multiple request?

google / uis-rnn

This is the library for the Unbounded Interleaved-State Recurrent Neural Network (UIS-RNN) algorithm, corresponding to the paper Fully Supervised Speaker Diarization.

https://arxiv.org/abs/1810.04719

Apache License 2.0

1.56k stars 319 forks source link

Is is possible to pre-load the model for multiple request? #83

Closed ghost closed 3 years ago

ghost commented 3 years ago

For single file, it works fine. Now I'm trying to process lots of files, for example 10000 audio files. I made a script for loop processing, but it is loading diarization model in each calling. I tried to change it to pre-load model, but not worked.

ValueError: Tensor Tensor("lambda_1/l2_normalize:0", shape=(?, 512), dtype=float32) is not an element of this graph.

Is there a way for it? Thanks for anyone, any help.

wq2012 commented 3 years ago

See https://github.com/google/uis-rnn/issues/32

We added uisrnn.parallel_predict().