Mehran Ghandehari comments

Results 17 comments of


                                            Mehran Ghandehari

Negative loss and IoU is greater than 1

This issue happened for me and the reason was setting logit=False in the loss function by mistake.

Problem with test.py

--save-hybrid leads to MAP of 1 that is a bug!

multi-gpu training error

I have had the same issue and did figure out that there is an issue with my data that leads to this erorr. Have not found out what is the...

multi-gpu training error

I learned that this issue occurs because there are some bounding boxes that slightly go outside of the image border. This might the main reason of WorkNCCL error.

multi-gpu training error

@glenn-jocher Thanks for your response. Yesterday I noted that the bounding box issue can lead to the error. So I did fix it, but faced the WorkNCCL error again when...

multi-gpu training error

@AyushExel @Laughing-q @glenn-jocher It seems that the responses generated by @glenn-jocher is computer generated like using ChatGPT as it does not make any sense!

multi-gpu training error

@AlbertoOddness @glenn-jocher I had the same issue and spent a lot of time explored many potential solutions including checking and fixing data issue, upgrading the pytorch and cuda to the...

stac-to-dc \ --catalog-href='https://earth-search.aws.element84.com/v0/' \ --bbox='25,20,35,30' \ --collections='sentinel-s2-l2a-cogs' \ --datetime='2020-01-01/2020-03-31'...running this command gives exception

it seems that stack api has changed...this works for me. stac-to-dc \ --catalog-href='https://earth-search.aws.element84.com/v1/' \ --bbox='25,20,35,30' \ --collections='sentinel-2-l2a' \ --datetime='2020-01-01/2020-03-31'

load_dataset method returns Unknown split "validation" even if this dir exists

For me it resolved by adding the verification_mode param: ``` imdb_ds = load_dataset( "imdb", verification_mode="no_checks" ) imdb_ds ```

RuntimeError when using "segment/predict.py" with --save-txt

@Robotatron I could run the val.py by making the following change xywh = (xyxy2xywh(torch.tensor(xyxy).view(1, 4)) / gn).view(-1).tolist() -> xywh = torch.tensor(xyxy).view(-1).numpy() This just allows the code to run but is...