[ONNX] Part1: update current onnx export

[felixdittrich92]

This PR:

• updates current onnx export for further processing in Part2

Part2 : add slow tests (only local testing) for each model

Any feedback is welcome 🤗

[codecov]

Codecov Report

Merging #976 (0fd1aef) into main (2e9a4e1) will decrease coverage by 0.03%.
The diff coverage is 77.77%.

@@            Coverage Diff             @@
##             main     #976      +/-   ##
==========================================
- Coverage   95.18%   95.14%   -0.04%     
==========================================
  Files         134      134              
  Lines        5520     5524       +4     
==========================================
+ Hits         5254     5256       +2     
- Misses        266      268       +2     

Flag	Coverage Δ
unittests	95.14% <77.77%> (-0.04%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
doctr/models/utils/tensorflow.py	95.34% <71.42%> (-4.66%)	⬇️
doctr/models/utils/pytorch.py	100.00% <100.00%> (ø)
doctr/transforms/modules/base.py	94.59% <0.00%> (ø)
doctr/models/recognition/master/tensorflow.py	100.00% <0.00%> (ø)
doctr/models/recognition/transformer/tensorflow.py	100.00% <0.00%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 2e9a4e1...0fd1aef. Read the comment docs.

[frgfm]

Thanks :) A few questions in comment

[felixdittrich92]

@frgfm after testing again opset 14 is the minimum version we need to ensure all operators are supported (have added a comment) otherwise the export fails for some models in both frameworks

We should also update torch to a minimum of 1.10 (first support for opset 14 in #791 requirements update)

[frgfm]

Thanks!