image_embeddings
image_embeddings copied to clipboard
batch mode and multithread
Hello,
Thx for the package ? How to setup batch mode and multithread easily ? (inference on CPU).
- Think one should be able load some small fine tuner head for quick fine tuning and save the model back.
hi, batching is already happening by default https://github.com/rom1504/image_embeddings/blob/master/image_embeddings/inference/inference.py
Hi,
Sure. Thanks.
Thinking of one command to skip or hide (The tfrecords part)
folder of images —> emb in numpy on disk
with model_path =…. (ie typically eff or some fine tuned eff)…
- Is reallt tfrecord necessary ? Why not image —> numpy—> TF tensor
Anwway, thanks for the package.
On Jan 23, 2022, at 1:09, Romain Beaumont @.***> wrote:
hi, batching is already happening by default https://github.com/rom1504/image_embeddings/blob/master/image_embeddings/inference/inference.py
— Reply to this email directly, view it on GitHub, or unsubscribe. Triage notifications on the go with GitHub Mobile for iOS or Android. You are receiving this because you authored the thread.
Does the package use cuda if it's available or is it CPU inference only @rom1504 ?