search

akoepke / audio-retrieval-benchmark

Implementation of "Audio Retrieval with Natural Language Queries: A Benchmark Study".
https://www.robots.ox.ac.uk/~vgg/research/audio-retrieval/
47 stars 2 forks source link

Training issue for MMT #2

Closed Gokulapriyan-B closed 2 years ago

Gokulapriyan-B commented 2 years ago

When trying to train the MMT model, the error is popping up. What is the solution to this?

1

P.S.. While loading dataset, the expert vggish or vggsound is not available. Is this an issue?

akoepke commented 2 years ago

Hi, It looks as if you did not download the features before running the training. For CLOTHO, you can download the features with python3 misc/sync_experts.py --dataset CLOTHO. The feature download is explained in more detail in our README.