Ayush Chaurasia comments

Results 416 comments of


                                            Ayush Chaurasia

multi-gpu training error

@AlphaNext please provide a command to reproduce

multi-gpu training error

@AlphaNext you don't need the distributed launch manually. The package handles that. Simply run python train.py

Unable to export model ( tflite )

tflite isn't supported yet. We'll work on it later this week

Unable to export model ( tflite )

Yes once we implement it, it'll be similar to yolov5

Onnx inference not working for image instance segmentation, maybe a bug in ONNX model?

@V4A001 This repo only supports yolov5. Are you seeing this error on a yolov5 trained model?

Model layers are fusing between predictions

@SkalskiP I guess you're trying to do this? It doesn't happen to me.. Or maybe I completely missunderstood your question

Model layers are fusing between predictions

@SkalskiP Oh I see what's happening. You don't have to pass video frame by frame. You can just pass the video directly and we'll return the generator if you want...

Model layers are fusing between predictions

@SkalskiP ok I'll fix it in main now. And should be available in tomorrow with next release

Model layers are fusing between predictions

@glenn-jocher yeah. Fuse is called by Autobackend and not by the predictor...`is_fuse` needs to be a model attribute then

Model layers are fusing between predictions

@glenn-jocher If we add the model attribute now, the released ckpts won't get it. Let's move this convo to the PR thread