Incoherent logged val/iou value despite coherent confusion matrix and per-class IoUs when training model in DDP

[CharlesGaydon]

Observed in DDP mode with 2x3 GPUs (not tested on single node training).

This is surprising, because the per-class IoUs are computed with a bespoke method from the confusion matrix of the MultiClassJaccardIndex object that computes mean IoU

myria3d/myria3d/models/model.py

Line 190 in d52810e

self.log("val/iou", iou_epoch, on_step=False, on_epoch=True, prog_bar=True)

Mean IoU:

Per-class:

Might be linked to #108

[CharlesGaydon]

Note: maybe #124 has corrected this since the computation of the IoU is now performed using a different function. Feel free to close if you do not observe this behavior again.