Closed gleblebedev closed 1 year ago
yes manually adding the .pth file in the /checkpoints folder should work. Did you run it locally first? It should download everything however it could be broken as its been a while since i made/updated it ;)
It should download everything however it could be broken as its been a while since i made/updated it ;)
It doesn't download the model file, I had to copy it manually from SD. Also i've noticed that documentation says "imgurl" when it actually expects "imgurls" field. Otherwise seems to work fine...
Would be nice not to reload model on each request too:
D00>]) Loading Model... load checkpoint from ./checkpoints/model_large_caption.pth Model to cuda:0 Captioning started ['a video game scene of two knights fighting in a field
In general it still saved me some time. I'm not familiar with python and this docker image helped me to make a prototype app.
Here is how I run it:
docker run -it --rm -p 5000:5000 --gpus=all py-worker
May worth to add this to readme too.
nice! glad it works now.
When running the powershell file docker build -t py-worker . it fails on
7 [2/7] ADD ./checkpoints /app/checkpoints
7 sha256:66385eb2b9e14ebca5144f4dc9ed2f54ac7cfa4adfaee182aaf2035af0103d3b
7 ERROR: "/checkpoints" not found: not found
the readme says:
It seems that there is step missing somewhere.
Shall I just drop model_base_caption_capfilt_large.pth in the folder?