clip-as-service
clip-as-service copied to clipboard
feat: support 4 new clip models pretrained on laion2b
This pr supports 4 new clip models introduced by https://laion.ai/blog/large-openclip/
- ViT-B-32::laion2B-s34B-b79K (PyTorch+ONNX)
- ViT-L-14::laion2B-s32B-b82K (PyTorch+ONNX)
- ViT-H-14::laion2B-s32B-b79K (PyTorch)
- ViT-g-14::laion2B-s12B-b42K (PyTorch)
It also sets timeout to 30 min in torch and onnx runtime since we have larger model files now
Codecov Report
Merging #825 (3835f75) into main (c690c24) will decrease coverage by
2.09%
. The diff coverage is62.50%
.
@@ Coverage Diff @@
## main #825 +/- ##
==========================================
- Coverage 84.67% 82.57% -2.10%
==========================================
Files 21 21
Lines 1566 1567 +1
==========================================
- Hits 1326 1294 -32
- Misses 240 273 +33
Flag | Coverage Δ | |
---|---|---|
cas | 82.57% <62.50%> (-2.10%) |
:arrow_down: |
Flags with carried forward coverage won't be shown. Click here to find out more.
Impacted Files | Coverage Δ | |
---|---|---|
server/clip_server/model/clip_onnx.py | 88.88% <ø> (ø) |
|
server/clip_server/model/pretrained_models.py | 98.41% <ø> (ø) |
|
server/clip_server/model/model.py | 78.97% <62.50%> (+0.04%) |
:arrow_up: |
server/clip_server/model/trt_utils.py | 56.04% <0.00%> (-27.48%) |
:arrow_down: |
server/clip_server/model/clip_trt.py | 69.38% <0.00%> (-16.33%) |
:arrow_down: |
Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.
:memo: Docs are deployed on https://ft-large-model--jina-docs.netlify.app :tada:
server/clip_server/model/model.py
This file is synced with https://github.com/mlfoundations/open_clip/blob/main/src/open_clip/model.py
@ZiniuYu Please polish the PR's title. 4 new clip models
is meaningless. Please highlights the contribution in the title