Automatic Inference of attach_index

[kozlov721]

Automatic attach_index

Added a functionality to automatically infer sensible attach index based on the type signature of forward method. Therefore, the attach_index doesn't have to be defined by the user if correct type annotations are present.

If the forward method has an annotated inputs, input, or x parameter, the type annotation is used to infer a sensible value for attach_index. This will be either "all" for inputs: list[Tensor] or -1 for inputs: Tensor. "all" also stays the default if unable to infer or not specified.

[github-actions]

☂️ Python Coverage

current status: ✅

Overall Coverage

Lines	Covered	Coverage	Threshold	Status
4429	3524	80%	0%	🟢

New Files

No new covered files...

Modified Files

File	Coverage	Status
luxonis_train/nodes/base_node.py	75%	🟢
luxonis_train/nodes/bisenet_head.py	50%	🟢
luxonis_train/nodes/classification_head.py	100%	🟢
luxonis_train/nodes/contextspatial.py	25%	🟢
luxonis_train/nodes/efficientrep.py	97%	🟢
luxonis_train/nodes/implicit_keypoint_bbox_head.py	92%	🟢
luxonis_train/nodes/micronet.py	95%	🟢
luxonis_train/nodes/mobilenetv2.py	35%	🟢
luxonis_train/nodes/mobileone.py	12%	🟢
luxonis_train/nodes/resnet.py	96%	🟢
luxonis_train/nodes/rexnetv1.py	15%	🟢
luxonis_train/nodes/segmentation_head.py	100%	🟢
TOTAL	66%	🟢

updated for commit: 04c01be by action🐍

[github-actions]

Test Results

  6 files    6 suites   1h 16m 41s ⏱️
 32 tests  32 ✅ 0 💤 0 ❌
192 runs  192 ✅ 0 💤 0 ❌

Results for commit 04c01be.

[conorsim]

LGTM