1 Name: Tensorflow Hub Models
2 Short Name: tfhub_models
3 URL: https://tfhub.dev/
13 This directory contains models downloaded from https://tfhub.dev/ under its
14 Apache license. These models are not built into Chromium and are only used as
18 1. Put all the models you want uploaded (including existing models to keep) in
19 the existing //third_party/tfhub_models/models directory.
21 2. `cd` to //third_party/tfhub_models.
23 3. Run the following commands:
25 tar -czf models.tar.gz models/*
26 sha1sum models.tar.gz | awk '{print $1}' > models.tar.gz.sha1
29 4. Upload the tarball to the cloud storage bucket:
31 upload_to_google_storage.py --bucket chromium-tfhub-models models.tar.gz
34 5. Overwrite chromium's sha1 models.tar.gz.sha1 file with the new one.
37 Please keep this list and BUILD.gn up to date when new files are added to the
40 * lite-model_universal-sentence-encoder-qa-ondevice_1.tflite - A TFLite
41 Support compatible embedding model trained on Q&A text, pulled from
42 https://tfhub.dev/google/lite-model/universal-sentence-encoder-qa-ondevice/1.
43 Requires sentencepiece tokenization to be built into Chrome, which is not
44 currently supported. See also https://crrev.com/c/4722355.
46 * rehead_embedding_from_mobilebert.tflite - A TFLite Support compatible
47 embedding model which was adapted from this BERT model,
48 https://tfhub.dev/tensorflow/lite-model/mobilebert/1/metadata/1. The base
49 model was re-headed to expose the embedding layer.